metadata
dict
answer
sequencelengths
1
389
question_with_context
sequencelengths
102
16.4k
paragraph_indices
sequencelengths
7
531
evidence
sequencelengths
7
531
{ "all_answers": [ { "text": "a vocabulary of positive and negative predicates that helps determine the polarity score of an event", "type": "ABSTRACTIVE" }, { "text": "seed lexicon consists of positive and negative predicates", "type": "EXTRACTIVE" } ], "all_evidence": [ [ "The seed lexicon consists of positive and negative predicates. If the predicate of an extracted event is in the seed lexicon and does not involve complex phenomena like negation, we assign the corresponding polarity score ($+1$ for positive events and $-1$ for negative events) to the event. We expect the model to automatically learn complex phenomena through label propagation. Based on the availability of scores and the types of discourse relations, we classify the extracted event pairs into the following three types." ], [ "The seed lexicon consists of positive and negative predicates. If the predicate of an extracted event is in the seed lexicon and does not involve complex phenomena like negation, we assign the corresponding polarity score ($+1$ for positive events and $-1$ for negative events) to the event. We expect the model to automatically learn complex phenomena through label propagation. Based on the availability of scores and the types of discourse relations, we classify the extracted event pairs into the following three types." ] ], "all_evidence_masks": [ [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ] ], "article_id": "1909.00694", "context_tokens": [ "Introduction", "</s>", "A", "ffect", "ive", "Ġevents", "ĠB", "IB", "REF", "0", "Ġare", "Ġevents", "Ġthat", "Ġtypically", "Ġaffect", "Ġpeople", "Ġin", "Ġpositive", "Ġor", "Ġnegative", "Ġways", ".", "ĠFor", "Ġexample", ",", "Ġgetting", "Ġmoney", "Ġand", "Ġplaying", "Ġsports", "Ġare", "Ġusually", "Ġpositive", "Ġto", "Ġthe", "Ġexperien", "cers", ";", "Ġcatching", "Ġcold", "Ġand", "Ġlosing", "Ġone", "'s", "Ġwallet", "Ġare", "Ġnegative", ".", "ĠUnderstanding", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġimportant", "Ġto", "Ġvarious", "Ġnatural", "Ġlanguage", "Ġprocessing", "Ġ(", "N", "LP", ")", "Ġapplications", "Ġsuch", "Ġas", "Ġdialogue", "Ġsystems", "ĠB", "IB", "REF", "1", ",", "Ġquestion", "-", "ans", "w", "ering", "Ġsystems", "ĠB", "IB", "REF", "2", ",", "Ġand", "Ġhumor", "Ġrecognition", "ĠB", "IB", "REF", "3", ".", "ĠIn", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġwork", "Ġon", "Ġrecognizing", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġaffect", "ive", "Ġevent", "Ġthat", "Ġis", "Ġrepresented", "Ġby", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġ(", "negative", ")", "Ġto", "Ġ1", "Ġ(", "positive", ").", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġchallenging", "Ġbecause", ",", "Ġas", "Ġthe", "Ġexamples", "Ġabove", "Ġsuggest", ",", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġis", "Ġnot", "Ġnecessarily", "Ġpredictable", "Ġfrom", "Ġits", "Ġconstituent", "Ġwords", ".", "ĠCombined", "Ġwith", "Ġthe", "Ġunb", "ounded", "Ġcomb", "inator", "ial", "Ġnature", "Ġof", "Ġlanguage", ",", "Ġthe", "Ġnon", "-", "com", "position", "ality", "Ġof", "Ġaffect", "ive", "Ġpol", "arity", "Ġentails", "Ġthe", "Ġneed", "Ġfor", "Ġlarge", "Ġamounts", "Ġof", "Ġworld", "Ġknowledge", ",", "Ġwhich", "Ġcan", "Ġhardly", "Ġbe", "Ġlearned", "Ġfrom", "Ġsmall", "Ġannot", "ated", "Ġdata", ".", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġpropose", "Ġa", "Ġsimple", "Ġand", "Ġeffective", "Ġmethod", "Ġfor", "Ġlearning", "Ġaffect", "ive", "Ġevents", "Ġthat", "Ġonly", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠAs", "Ġillustrated", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġour", "Ġkey", "Ġidea", "Ġis", "Ġthat", "Ġwe", "Ġcan", "Ġexploit", "Ġdiscourse", "Ġrelations", "ĠB", "IB", "REF", "4", "Ġto", "Ġefficiently", "Ġpropagate", "Ġpol", "arity", "Ġfrom", "Ġseed", "Ġpred", "icates", "Ġthat", "Ġdirectly", "Ġreport", "Ġone", "'s", "Ġemotions", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "to", "Ġbe", "Ġglad", "âĢ", "Ŀ", "Ġis", "Ġpositive", ").", "ĠSuppose", "Ġthat", "Ġevents", "Ġ$", "x", "_", "1", "$", "Ġare", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCause", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "1", "$", "Ġcauses", "Ġ$", "x", "_", "2", "$", ").", "ĠIf", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġsuggests", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġpositive", ",", "Ġ$", "x", "_", "1", "$", "Ġis", "Ġalso", "Ġlikely", "Ġto", "Ġbe", "Ġpositive", "Ġbecause", "Ġit", "Ġtriggers", "Ġthe", "Ġpositive", "Ġemotion", ".", "ĠThe", "Ġfact", "Ġthat", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġknown", "Ġto", "Ġbe", "Ġnegative", "Ġindicates", "Ġthe", "Ġnegative", "Ġpol", "arity", "Ġof", "Ġ$", "x", "_", "1", "$.", "ĠSimilarly", ",", "Ġif", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCon", "cession", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "2", "$", "Ġin", "Ġspite", "Ġof", "Ġ$", "x", "_", "1", "$", "),", "Ġthe", "Ġreverse", "Ġof", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġcan", "Ġbe", "Ġpropag", "ated", "Ġto", "Ġ$", "x", "_", "1", "$.", "ĠEven", "Ġif", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġis", "Ġnot", "Ġknown", "Ġin", "Ġadvance", ",", "Ġwe", "Ġcan", "Ġexploit", "Ġthe", "Ġtendency", "Ġof", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġto", "Ġbe", "Ġof", "Ġthe", "Ġsame", "Ġpol", "arity", "Ġ(", "for", "ĠCause", ")", "Ġor", "Ġof", "Ġthe", "Ġreverse", "Ġpol", "arity", "Ġ(", "for", "ĠCon", "cession", ")", "Ġalthough", "Ġthe", "Ġhe", "uristic", "Ġis", "Ġnot", "Ġexempt", "Ġfrom", "Ġcount", "ere", "x", "amples", ".", "ĠWe", "Ġtransform", "Ġthis", "Ġidea", "Ġinto", "Ġobjective", "Ġfunctions", "Ġand", "Ġtrain", "Ġneural", "Ġnetwork", "Ġmodels", "Ġthat", "Ġpredict", "Ġthe", "Ġpol", "arity", "Ġof", "Ġa", "Ġgiven", "Ġevent", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodels", "Ġusing", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", ".", "ĠGiven", "Ġthe", "Ġminimum", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthey", "Ġperformed", "Ġwell", ".", "ĠIn", "Ġaddition", ",", "Ġthe", "Ġcombination", "Ġof", "Ġannot", "ated", "Ġand", "Ġun", "annot", "ated", "Ġdata", "Ġyielded", "Ġa", "Ġgain", "Ġover", "Ġa", "Ġpurely", "Ġsupervised", "Ġbaseline", "Ġwhen", "Ġlabeled", "Ġdata", "Ġwere", "Ġsmall", ".", "</s>", "Related", "ĠWork", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġclosely", "Ġrelated", "Ġto", "Ġsentiment", "Ġanalysis", ".", "ĠWhereas", "Ġsentiment", "Ġanalysis", "Ġusually", "Ġfocuses", "Ġon", "Ġthe", "Ġpol", "arity", "Ġof", "Ġwhat", "Ġare", "Ġdescribed", "Ġ(", "e", ".", "g", ".,", "Ġmovies", "),", "Ġwe", "Ġwork", "Ġon", "Ġhow", "Ġpeople", "Ġare", "Ġtypically", "Ġaffected", "Ġby", "Ġevents", ".", "ĠIn", "Ġsentiment", "Ġanalysis", ",", "Ġmuch", "Ġattention", "Ġhas", "Ġbeen", "Ġpaid", "Ġto", "Ġcomposition", "ality", ".", "ĠWord", "-", "level", "Ġpol", "arity", "ĠB", "IB", "REF", "5", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "7", "Ġand", "Ġthe", "Ġroles", "Ġof", "Ġneg", "ation", "Ġand", "Ġintens", "ification", "ĠB", "IB", "REF", "8", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "9", "Ġare", "Ġamong", "Ġthe", "Ġmost", "Ġimportant", "Ġtopics", ".", "ĠIn", "Ġcontrast", ",", "Ġwe", "Ġare", "Ġmore", "Ġinterested", "Ġin", "Ġrecognizing", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġthat", "Ġper", "tains", "Ġto", "Ġcommons", "ense", "Ġknowledge", "Ġ(", "e", ".", "g", ".,", "Ġgetting", "Ġmoney", "Ġand", "Ġcatching", "Ġcold", ").", "</s>", "Label", "Ġpropagation", "Ġfrom", "Ġseed", "Ġinstances", "Ġis", "Ġa", "Ġcommon", "Ġapproach", "Ġto", "Ġinducing", "Ġsentiment", "Ġpolar", "ities", ".", "ĠWhile", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġworked", "Ġon", "Ġword", "-", "Ġand", "Ġphrase", "-", "level", "Ġpolar", "ities", ",", "ĠB", "IB", "REF", "0", "Ġdealt", "Ġwith", "Ġevent", "-", "level", "Ġpolar", "ities", ".", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġlinked", "Ġinstances", "Ġusing", "Ġco", "-", "occ", "urrence", "Ġinformation", "Ġand", "/", "or", "Ġphrase", "-", "level", "Ġcoord", "inations", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "$", "A", "$", "Ġand", "Ġ$", "B", "$", "âĢ", "Ŀ", "Ġand", "ĠâĢ", "ľ", "$", "A", "$", "Ġbut", "Ġ$", "B", "$", "âĢ", "Ŀ", ").", "ĠWe", "Ġshift", "Ġour", "Ġscope", "Ġto", "Ġevent", "Ġpairs", "Ġthat", "Ġare", "Ġmore", "Ġcomplex", "Ġthan", "Ġphrase", "Ġpairs", ",", "Ġand", "Ġconsequently", "Ġexploit", "Ġdiscourse", "Ġconnect", "ives", "Ġas", "Ġevent", "-", "level", "Ġcounterparts", "Ġof", "Ġphrase", "-", "level", "Ġconj", "unctions", ".", "</s>", "BI", "B", "REF", "0", "Ġconstructed", "Ġa", "Ġnetwork", "Ġof", "Ġevents", "Ġusing", "Ġword", "Ġembed", "ding", "-", "derived", "Ġsimilarities", ".", "ĠCompared", "Ġwith", "Ġthis", "Ġmethod", ",", "Ġour", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlinking", "Ġof", "Ġevents", "Ġis", "Ġmuch", "Ġsimpler", "Ġand", "Ġmore", "Ġintuitive", ".", "</s>", "Some", "Ġprevious", "Ġstudies", "Ġmade", "Ġuse", "Ġof", "Ġdocument", "Ġstructure", "Ġto", "Ġunderstand", "Ġthe", "Ġsentiment", ".", "ĠB", "IB", "REF", "11", "Ġproposed", "Ġa", "Ġsentiment", "-", "specific", "Ġpre", "-", "training", "Ġstrategy", "Ġusing", "Ġunl", "abel", "ed", "Ġdialog", "Ġdata", "Ġ(", "t", "weet", "-", "reply", "Ġpairs", ").", "ĠB", "IB", "REF", "12", "Ġproposed", "Ġa", "Ġmethod", "Ġof", "Ġbuilding", "Ġa", "Ġpol", "arity", "-", "tag", "ged", "Ġcorpus", "Ġ(", "ACP", "ĠCorpus", ").", "ĠThey", "Ġautomatically", "Ġgathered", "Ġsentences", "Ġthat", "Ġhad", "Ġpositive", "Ġor", "Ġnegative", "Ġopinions", "Ġutilizing", "ĠHTML", "Ġlayout", "Ġstructures", "Ġin", "Ġaddition", "Ġto", "Ġlinguistic", "Ġpatterns", ".", "ĠOur", "Ġmethod", "Ġdepends", "Ġonly", "Ġon", "Ġraw", "Ġtexts", "Ġand", "Ġthus", "Ġhas", "Ġwider", "Ġapplic", "ability", ".", "</s>", "Prop", "osed", "ĠMethod", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠPol", "arity", "ĠFunction", "</s>", "Our", "Ġgoal", "Ġis", "Ġto", "Ġlearn", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$,", "Ġwhich", "Ġpredicts", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġscore", "Ġof", "Ġan", "Ġevent", "Ġ$", "x", "$.", "ĠWe", "Ġapproximate", "Ġ$", "p", "(", "x", ")", "$", "Ġby", "Ġa", "Ġneural", "Ġnetwork", "Ġwith", "Ġthe", "Ġfollowing", "Ġform", ":", "</s>", "$", "{\\", "rm", "ĠEnc", "oder", "}", "$", "Ġoutputs", "Ġa", "Ġvector", "Ġrepresentation", "Ġof", "Ġthe", "Ġevent", "Ġ$", "x", "$.", "Ġ${", "\\", "rm", "ĠLinear", "}", "$", "Ġis", "Ġa", "Ġfully", "-", "connected", "Ġlayer", "Ġand", "Ġtransforms", "Ġthe", "Ġrepresentation", "Ġinto", "Ġa", "Ġscal", "ar", ".", "Ġ${", "\\", "rm", "Ġtan", "h", "}", "$", "Ġis", "Ġthe", "Ġhyper", "b", "olic", "Ġtang", "ent", "Ġand", "Ġtransforms", "Ġthe", "Ġscal", "ar", "Ġinto", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġto", "Ġ1", ".", "ĠIn", "ĠSection", "ĠSEC", "REF", "21", ",", "Ġwe", "Ġconsider", "Ġtwo", "Ġspecific", "Ġimplementations", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$.", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "</s>", "Our", "Ġmethod", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġassume", "Ġthat", "Ġwe", "Ġcan", "Ġautomatically", "Ġextract", "Ġdiscourse", "-", "tag", "ged", "Ġevent", "Ġpairs", ",", "Ġ$(", "x", "_{", "i", "1", "},", "Ġx", "_{", "i", "2", "})", "$", "Ġ($", "i", "=", "1", ",", "Ġ\\", "cd", "ots", "Ġ$", ")", "Ġfrom", "Ġthe", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġrefer", "Ġto", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġas", "Ġformer", "Ġand", "Ġlatter", "Ġevents", ",", "Ġrespectively", ".", "ĠAs", "Ġshown", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġwe", "Ġlimit", "Ġour", "Ġscope", "Ġto", "Ġtwo", "Ġdiscourse", "Ġrelations", ":", "ĠCause", "Ġand", "ĠCon", "cession", ".", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġconsists", "Ġof", "Ġpositive", "Ġand", "Ġnegative", "Ġpred", "icates", ".", "ĠIf", "Ġthe", "Ġpredicate", "Ġof", "Ġan", "Ġextracted", "Ġevent", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġdoes", "Ġnot", "Ġinvolve", "Ġcomplex", "Ġphenomena", "Ġlike", "Ġneg", "ation", ",", "Ġwe", "Ġassign", "Ġthe", "Ġcorresponding", "Ġpol", "arity", "Ġscore", "Ġ($", "+", "1", "$", "Ġfor", "Ġpositive", "Ġevents", "Ġand", "Ġ$", "-", "1", "$", "Ġfor", "Ġnegative", "Ġevents", ")", "Ġto", "Ġthe", "Ġevent", ".", "ĠWe", "Ġexpect", "Ġthe", "Ġmodel", "Ġto", "Ġautomatically", "Ġlearn", "Ġcomplex", "Ġphenomena", "Ġthrough", "Ġlabel", "Ġpropagation", ".", "ĠBased", "Ġon", "Ġthe", "Ġavailability", "Ġof", "Ġscores", "Ġand", "Ġthe", "Ġtypes", "Ġof", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġclassify", "Ġthe", "Ġextracted", "Ġevent", "Ġpairs", "Ġinto", "Ġthe", "Ġfollowing", "Ġthree", "Ġtypes", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠAL", "Ġ(", "Autom", "atically", "ĠLab", "eled", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġ(", "1", ")", "Ġthe", "Ġlatter", "Ġevent", "Ġbut", "Ġ(", "2", ")", "Ġnot", "Ġthe", "Ġformer", "Ġevent", ",", "Ġand", "Ġ(", "3", ")", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", "Ġor", "ĠCon", "cession", ".", "ĠIf", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġsame", "Ġscore", "Ġas", "Ġthe", "Ġlatter", ".", "ĠLikewise", ",", "Ġif", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġopposite", "Ġof", "Ġthe", "Ġlatter", "'s", "Ġscore", ".", "ĠThey", "Ġare", "Ġused", "Ġas", "Ġreference", "Ġscores", "Ġduring", "Ġtraining", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCA", "Ġ(", "Cause", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġsame", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCO", "Ġ(", "Con", "cession", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġreversed", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠLoss", "ĠFunctions", "</s>", "Using", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "Ġdata", ",", "Ġwe", "Ġoptimize", "Ġthe", "Ġparameters", "Ġof", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$.", "ĠWe", "Ġdefine", "Ġa", "Ġloss", "Ġfunction", "Ġfor", "Ġeach", "Ġof", "Ġthe", "Ġthree", "Ġtypes", "Ġof", "Ġevent", "Ġpairs", "Ġand", "Ġsum", "Ġup", "Ġthe", "Ġmultiple", "Ġloss", "Ġfunctions", ".", "</s>", "We", "Ġuse", "Ġmean", "Ġsquared", "Ġerror", "Ġto", "Ġconstruct", "Ġloss", "Ġfunctions", ".", "ĠFor", "Ġthe", "ĠAL", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "where", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠAL", "Ġdata", ".", "Ġ$", "r", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "r", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġautomatically", "-", "ass", "igned", "Ġscores", "Ġof", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$,", "Ġrespectively", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠAL", "Ġpairs", ",", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġa", "Ġhyper", "param", "eter", ".", "</s>", "For", "Ġthe", "ĠCA", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "$", "y", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "y", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠCA", "Ġpairs", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠCA", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠCA", "Ġpairs", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "mu", "Ġ$", "Ġare", "Ġhyper", "param", "eters", ".", "ĠThe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġcloser", "Ġwhile", "Ġthe", "Ġsecond", "Ġterm", "Ġprevents", "Ġthe", "Ġscores", "Ġfrom", "Ġshrinking", "Ġto", "Ġzero", ".", "</s>", "The", "Ġloss", "Ġfunction", "Ġfor", "Ġthe", "ĠCO", "Ġdata", "Ġis", "Ġdefined", "Ġanalog", "ously", ":", "</s>", "The", "Ġdifference", "Ġis", "Ġthat", "Ġthe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġdistant", "Ġfrom", "Ġeach", "Ġother", ".", "</s>", "Exper", "iments", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "</s>", "As", "Ġa", "Ġraw", "Ġcorpus", ",", "Ġwe", "Ġused", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", "Ġthat", "Ġwas", "Ġcompiled", "Ġthrough", "Ġthe", "Ġprocedures", "Ġproposed", "Ġby", "ĠB", "IB", "REF", "13", ".", "ĠTo", "Ġextract", "Ġevent", "Ġpairs", "Ġtagged", "Ġwith", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġused", "Ġthe", "ĠJapanese", "Ġdependency", "Ġparser", "ĠK", "NP", "Ġand", "Ġin", "-", "house", "Ġpost", "processing", "Ġscripts", "ĠB", "IB", "REF", "14", ".", "ĠK", "NP", "Ġused", "Ġhand", "-", "written", "Ġrules", "Ġto", "Ġsegment", "Ġeach", "Ġsentence", "Ġinto", "Ġwhat", "Ġwe", "Ġconvention", "ally", "Ġcalled", "Ġclauses", "Ġ(", "mostly", "Ġconsecutive", "Ġtext", "Ġchunks", "),", "Ġeach", "Ġof", "Ġwhich", "Ġcontained", "Ġone", "Ġmain", "Ġpredicate", ".", "ĠK", "NP", "Ġalso", "Ġidentified", "Ġthe", "Ġdiscourse", "Ġrelations", "Ġof", "Ġevent", "Ġpairs", "Ġif", "Ġexplicit", "Ġdiscourse", "Ġconnect", "ives", "ĠB", "IB", "REF", "4", "Ġsuch", "Ġas", "ĠâĢ", "ľ", "ãģ®", "ãģ§", "âĢ", "Ŀ", "Ġ(", "because", ")", "Ġand", "ĠâĢ", "ľ", "ãģ®", "ãģ«", "âĢ", "Ŀ", "Ġ(", "in", "Ġspite", "Ġof", ")", "Ġwere", "Ġpresent", ".", "ĠWe", "Ġtreated", "ĠCause", "/", "Reason", "Ġ(", "åİ", "Ł", "åĽ", "ł", "ãĥ»", "çIJ", "Ĩ", "çĶ", "±", ")", "Ġand", "ĠCondition", "Ġ(", "æĿ", "¡", "ä»", "¶", ")", "Ġin", "Ġthe", "Ġoriginal", "Ġtag", "set", "ĠB", "IB", "REF", "15", "Ġas", "ĠCause", "Ġand", "ĠCon", "cession", "Ġ(", "éĢ", "Ĩ", "æ", "İ", "¥", ")", "Ġas", "ĠCon", "cession", ",", "Ġrespectively", ".", "ĠHere", "Ġis", "Ġan", "Ġexample", "Ġof", "Ġevent", "Ġpair", "Ġextraction", ".", "</s>", ".", "Ġé", "ĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĹ", "ãģŁ", "ãģ®", "ãģ§", "ãĢģ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãģ£", "ãģŁ", "ãĢĤ", "</s>", "Because", "Ġ[", "I", "]", "Ġmade", "Ġa", "Ġserious", "Ġmistake", ",", "Ġ[", "I", "]", "Ġgot", "Ġfired", ".", "</s>", "From", "Ġthis", "Ġsentence", ",", "Ġwe", "Ġextracted", "Ġthe", "Ġevent", "Ġpair", "Ġof", "ĠâĢ", "ľ", "éĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĻ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġmake", "Ġa", "Ġserious", "Ġmistake", ")", "Ġand", "ĠâĢ", "ľ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãĤĭ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġget", "Ġfired", "),", "Ġand", "Ġtagged", "Ġit", "Ġwith", "ĠCause", ".", "</s>", "We", "Ġconstructed", "Ġour", "Ġseed", "Ġlex", "icon", "Ġconsisting", "Ġof", "Ġ15", "Ġpositive", "Ġwords", "Ġand", "Ġ15", "Ġnegative", "Ġwords", ",", "Ġas", "Ġshown", "Ġin", "ĠSection", "ĠSEC", "REF", "27", ".", "ĠFrom", "Ġthe", "Ġcorpus", "Ġof", "Ġabout", "Ġ100", "Ġmillion", "Ġsentences", ",", "Ġwe", "Ġobtained", "Ġ1", ".", "4", "Ġmillions", "Ġevent", "Ġpairs", "Ġfor", "ĠAL", ",", "Ġ41", "Ġmillions", "Ġfor", "ĠCA", ",", "Ġand", "Ġ6", "Ġmillions", "Ġfor", "ĠCO", ".", "ĠWe", "Ġrandomly", "Ġselected", "Ġsubs", "ets", "Ġof", "ĠAL", "Ġevent", "Ġpairs", "Ġsuch", "Ġthat", "Ġpositive", "Ġand", "Ġnegative", "Ġlatter", "Ġevents", "Ġwere", "Ġequal", "Ġin", "Ġsize", ".", "ĠWe", "Ġalso", "Ġsampled", "Ġevent", "Ġpairs", "Ġfor", "Ġeach", "Ġof", "ĠCA", "Ġand", "ĠCO", "Ġsuch", "Ġthat", "Ġit", "Ġwas", "Ġfive", "Ġtimes", "Ġlarger", "Ġthan", "ĠAL", ".", "ĠThe", "Ġresults", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "16", ".", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAC", "P", "Ġ(", "ACP", "ĠCorpus", ")", "</s>", "We", "Ġused", "Ġthe", "Ġlatest", "Ġversion", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", "ĠB", "IB", "REF", "12", "Ġfor", "Ġevaluation", ".", "ĠIt", "Ġwas", "Ġused", "Ġfor", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġtraining", "Ġas", "Ġwell", ".", "ĠExt", "racted", "Ġfrom", "ĠJapanese", "Ġwebsites", "Ġusing", "ĠHTML", "Ġlayouts", "Ġand", "Ġlinguistic", "Ġpatterns", ",", "Ġthe", "Ġdataset", "Ġcovered", "Ġvarious", "Ġgenres", ".", "ĠFor", "Ġexample", ",", "Ġthe", "Ġfollowing", "Ġtwo", "Ġsentences", "Ġwere", "Ġlabeled", "Ġpositive", "Ġand", "Ġnegative", ",", "Ġrespectively", ":", "</s>", ".", "Ġ", "ä½ľ", "æ", "¥", "Ń", "ãģĮ", "æ", "¥", "½", "ãģł", "ãĢĤ", "</s>", "The", "Ġwork", "Ġis", "Ġeasy", ".", "</s>", ".", "Ġé", "§", "IJ", "è»", "Ĭ", "å", "ł", "´", "ãģĮ", "ãģª", "ãģĦ", "ãĢĤ", "</s>", "There", "Ġis", "Ġno", "Ġparking", "Ġlot", ".", "</s>", "Although", "Ġthe", "ĠAC", "P", "Ġcorpus", "Ġwas", "Ġoriginally", "Ġconstructed", "Ġin", "Ġthe", "Ġcontext", "Ġof", "Ġsentiment", "Ġanalysis", ",", "Ġwe", "Ġfound", "Ġthat", "Ġit", "Ġcould", "Ġroughly", "Ġbe", "Ġregarded", "Ġas", "Ġa", "Ġcollection", "Ġof", "Ġaffect", "ive", "Ġevents", ".", "ĠWe", "Ġparsed", "Ġeach", "Ġsentence", "Ġand", "Ġextracted", "Ġthe", "Ġlast", "Ġclause", "Ġin", "Ġit", ".", "ĠThe", "Ġtrain", "/", "dev", "/", "test", "Ġsplit", "Ġof", "Ġthe", "Ġdata", "Ġis", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "19", ".", "</s>", "The", "Ġobjective", "Ġfunction", "Ġfor", "Ġsupervised", "Ġtraining", "Ġis", ":", "</s>", "where", "Ġ$", "v", "_", "i", "$", "Ġis", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġevent", ",", "Ġ$", "R", "_", "i", "$", "Ġis", "Ġthe", "Ġreference", "Ġscore", "Ġof", "Ġ$", "v", "_", "i", "$,", "Ġand", "Ġ$", "N", "_{", "\\", "rm", "ĠAC", "P", "}", "$", "Ġis", "Ġthe", "Ġnumber", "Ġof", "Ġthe", "Ġevents", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "</s>", "To", "Ġoptimize", "Ġthe", "Ġhyper", "param", "eters", ",", "Ġwe", "Ġused", "Ġthe", "Ġdev", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠFor", "Ġthe", "Ġevaluation", ",", "Ġwe", "Ġused", "Ġthe", "Ġtest", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠThe", "Ġmodel", "Ġoutput", "Ġwas", "Ġclassified", "Ġas", "Ġpositive", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ>", "Ġ0", "$", "Ġand", "Ġnegative", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ\\", "le", "Ġ0", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠModel", "ĠConfig", "urations", "</s>", "As", "Ġfor", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġwe", "Ġcompared", "Ġtwo", "Ġtypes", "Ġof", "Ġneural", "Ġnetworks", ":", "ĠBi", "GR", "U", "Ġand", "ĠB", "ERT", ".", "ĠGR", "U", "ĠB", "IB", "REF", "16", "Ġis", "Ġa", "Ġrecurrent", "Ġneural", "Ġnetwork", "Ġsequence", "Ġenc", "oder", ".", "ĠBi", "GR", "U", "Ġreads", "Ġan", "Ġinput", "Ġsequence", "Ġforward", "Ġand", "Ġbackward", "Ġand", "Ġthe", "Ġoutput", "Ġis", "Ġthe", "Ġconc", "aten", "ation", "Ġof", "Ġthe", "Ġfinal", "Ġforward", "Ġand", "Ġbackward", "Ġhidden", "Ġstates", ".", "</s>", "BER", "T", "ĠB", "IB", "REF", "17", "Ġis", "Ġa", "Ġpre", "-", "trained", "Ġmulti", "-", "layer", "Ġbid", "irection", "al", "ĠTrans", "former", "ĠB", "IB", "REF", "18", "Ġenc", "oder", ".", "ĠIts", "Ġoutput", "Ġis", "Ġthe", "Ġfinal", "Ġhidden", "Ġstate", "Ġcorresponding", "Ġto", "Ġthe", "Ġspecial", "Ġclassification", "Ġtag", "Ġ([", "CL", "S", "]).", "ĠFor", "Ġthe", "Ġdetails", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġsee", "ĠSections", "ĠSEC", "REF", "30", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodel", "Ġwith", "Ġthe", "Ġfollowing", "Ġfour", "Ġcombinations", "Ġof", "Ġthe", "Ġdatasets", ":", "ĠAL", ",", "ĠAL", "+", "CA", "+", "CO", "Ġ(", "two", "Ġproposed", "Ġmodels", "),", "ĠAC", "P", "Ġ(", "super", "vised", "),", "Ġand", "ĠAC", "P", "+", "AL", "+", "CA", "+", "CO", "Ġ(", "se", "mi", "-", "super", "vised", ").", "ĠThe", "Ġcorresponding", "Ġobjective", "Ġfunctions", "Ġwere", ":", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "$,", "Ġand", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠResults", "Ġand", "ĠDiscussion", "</s>", "Table", "ĠT", "AB", "REF", "23", "Ġshows", "Ġaccuracy", ".", "ĠAs", "Ġthe", "ĠRandom", "Ġbaseline", "Ġsuggests", ",", "Ġpositive", "Ġand", "Ġnegative", "Ġlabels", "Ġwere", "Ġdistributed", "Ġevenly", ".", "ĠThe", "ĠRandom", "+", "S", "eed", "Ġbaseline", "Ġmade", "Ġuse", "Ġof", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġoutput", "Ġthe", "Ġcorresponding", "Ġlabel", "Ġ(", "or", "Ġthe", "Ġreverse", "Ġof", "Ġit", "Ġfor", "Ġneg", "ation", ")", "Ġif", "Ġthe", "Ġevent", "'s", "Ġpredicate", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", ".", "ĠWe", "Ġcan", "Ġsee", "Ġthat", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġitself", "Ġhad", "Ġpractically", "Ġno", "Ġimpact", "Ġon", "Ġprediction", ".", "</s>", "The", "Ġmodels", "Ġin", "Ġthe", "Ġtop", "Ġblock", "Ġperformed", "Ġconsiderably", "Ġbetter", "Ġthan", "Ġthe", "Ġrandom", "Ġbas", "elines", ".", "ĠThe", "Ġperformance", "Ġgaps", "Ġwith", "Ġtheir", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġcounterparts", ",", "Ġshown", "Ġin", "Ġthe", "Ġmiddle", "Ġblock", ",", "Ġwere", "Ġless", "Ġthan", "Ġ7", "%.", "ĠThis", "Ġdemonstrates", "Ġthe", "Ġeffectiveness", "Ġof", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlabel", "Ġpropagation", ".", "</s>", "Comp", "aring", "Ġthe", "Ġmodel", "Ġvariants", ",", "Ġwe", "Ġobtained", "Ġthe", "Ġhighest", "Ġscore", "Ġwith", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġtrained", "Ġwith", "Ġthe", "ĠAL", "+", "CA", "+", "CO", "Ġdataset", ".", "ĠB", "ERT", "Ġwas", "Ġcompetitive", "Ġbut", "Ġits", "Ġperformance", "Ġwent", "Ġdown", "Ġif", "ĠCA", "Ġand", "ĠCO", "Ġwere", "Ġused", "Ġin", "Ġaddition", "Ġto", "ĠAL", ".", "ĠWe", "Ġconjecture", "Ġthat", "ĠB", "ERT", "Ġwas", "Ġmore", "Ġsensitive", "Ġto", "Ġnoises", "Ġfound", "Ġmore", "Ġfrequently", "Ġin", "ĠCA", "Ġand", "ĠCO", ".", "</s>", "Cont", "rary", "Ġto", "Ġour", "Ġexpectations", ",", "Ġsupervised", "Ġmodels", "Ġ(", "ACP", ")", "Ġoutper", "formed", "Ġsemi", "-", "super", "vised", "Ġmodels", "Ġ(", "ACP", "+", "AL", "+", "CA", "+", "CO", ").", "ĠThis", "Ġsuggests", "Ġthat", "Ġthe", "Ġtraining", "Ġset", "Ġof", "Ġ0", ".", "6", "Ġmillion", "Ġevents", "Ġis", "Ġsufficiently", "Ġlarge", "Ġfor", "Ġtraining", "Ġthe", "Ġmodels", ".", "ĠFor", "Ġcomparison", ",", "Ġwe", "Ġtrained", "Ġthe", "Ġmodels", "Ġwith", "Ġa", "Ġsubset", "Ġ(", "6", ",", "000", "Ġevents", ")", "Ġof", "Ġthe", "ĠAC", "P", "Ġdataset", ".", "ĠAs", "Ġthe", "Ġresults", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "24", "Ġdemonstrate", ",", "Ġour", "Ġmethod", "Ġis", "Ġeffective", "Ġwhen", "Ġlabeled", "Ġdata", "Ġare", "Ġsmall", ".", "</s>", "The", "Ġresult", "Ġof", "Ġhyper", "param", "eter", "Ġoptimization", "Ġfor", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġwas", "Ġas", "Ġfollows", ":", "</s>", "As", "Ġthe", "ĠCA", "Ġand", "ĠCO", "Ġpairs", "Ġwere", "Ġequal", "Ġin", "Ġsize", "Ġ(", "Table", "ĠT", "AB", "REF", "16", "),", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$", "Ġwere", "Ġcomparable", "Ġvalues", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġwas", "Ġabout", "Ġone", "-", "third", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$,", "Ġand", "Ġthis", "Ġindicated", "Ġthat", "Ġthe", "ĠCA", "Ġpairs", "Ġwere", "Ġno", "is", "ier", "Ġthan", "Ġthe", "ĠCO", "Ġpairs", ".", "ĠA", "Ġmajor", "Ġtype", "Ġof", "ĠCA", "Ġpairs", "Ġthat", "Ġviolates", "Ġour", "Ġassumption", "Ġwas", "Ġin", "Ġthe", "Ġform", "Ġof", "ĠâĢ", "ľ", "$", "\\", "text", "it", "Ġ{", "problem", "}", "_{", "\\", "text", "{", "negative", "}}", "$", "Ġcauses", "Ġ$\\", "text", "it", "Ġ{", "s", "olution", "}", "_{", "\\", "text", "{", "positive", "}}", "$", "âĢ", "Ŀ", ":", "</s>", ".", "Ġ(", "æ", "Ĥª", "ãģĦ", "ãģ¨", "ãģĵ", "ãĤ", "į", "ãģĮ", "ãģĤ", "ãĤĭ", ",", "ĠãĤ", "Ī", "ãģı", "ãģª", "ãĤĭ", "ãĤ", "Ī", "ãģĨ", "ãģ«", "åĬ", "ª", "åĬ", "Ľ", "ãģĻ", "ãĤĭ", ")", "</s>", "(", "there", "Ġis", "Ġa", "Ġbad", "Ġpoint", ",", "Ġ[", "I", "]", "Ġtry", "Ġto", "Ġimprove", "Ġ[", "it", "])", "</s>", "The", "Ġpolar", "ities", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġwere", "Ġreversed", "Ġin", "Ġspite", "Ġof", "Ġthe", "ĠCause", "Ġrelation", ",", "Ġand", "Ġthis", "Ġlowered", "Ġthe", "Ġvalue", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$.", "</s>", "Some", "Ġexamples", "Ġof", "Ġmodel", "Ġoutputs", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "26", ".", "ĠThe", "Ġfirst", "Ġtwo", "Ġexamples", "Ġsuggest", "Ġthat", "Ġour", "Ġmodel", "Ġsuccessfully", "Ġlearned", "Ġneg", "ation", "Ġwithout", "Ġexplicit", "Ġsupervision", ".", "ĠSimilarly", ",", "Ġthe", "Ġnext", "Ġtwo", "Ġexamples", "Ġdiffer", "Ġonly", "Ġin", "Ġvoice", "Ġbut", "Ġthe", "Ġmodel", "Ġcorrectly", "Ġrecognized", "Ġthat", "Ġthey", "Ġhad", "Ġopposite", "Ġpolar", "ities", ".", "ĠThe", "Ġlast", "Ġtwo", "Ġexamples", "Ġshare", "Ġthe", "Ġpredicate", "ĠâĢ", "ľ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "drop", ")", "Ġand", "Ġonly", "Ġthe", "Ġobjects", "Ġare", "Ġdifferent", ".", "ĠThe", "Ġsecond", "Ġevent", "ĠâĢ", "ľ", "è", "Ĥ", "©", "ãĤĴ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "lit", ".", "Ġdrop", "Ġone", "'s", "Ġshoulders", ")", "Ġis", "Ġan", "Ġid", "iom", "Ġthat", "Ġexpresses", "Ġa", "Ġdisappointed", "Ġfeeling", ".", "ĠThe", "Ġexamples", "Ġdemonstrate", "Ġthat", "Ġour", "Ġmodel", "Ġcorrectly", "Ġlearned", "Ġnon", "-", "com", "pos", "itional", "Ġexpressions", ".", "</s>", "Conclusion", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġproposed", "Ġto", "Ġuse", "Ġdiscourse", "Ġrelations", "Ġto", "Ġeffectively", "Ġpropagate", "Ġpolar", "ities", "Ġof", "Ġaffect", "ive", "Ġevents", "Ġfrom", "Ġseeds", ".", "ĠExper", "iments", "Ġshow", "Ġthat", ",", "Ġeven", "Ġwith", "Ġa", "Ġminimal", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthe", "Ġproposed", "Ġmethod", "Ġperformed", "Ġwell", ".", "</s>", "Although", "Ġevent", "Ġpairs", "Ġlinked", "Ġby", "Ġdiscourse", "Ġanalysis", "Ġare", "Ġshown", "Ġto", "Ġbe", "Ġuseful", ",", "Ġthey", "Ġnevertheless", "Ġcontain", "Ġnoises", ".", "ĠAdding", "Ġlingu", "istically", "-", "mot", "ivated", "Ġfiltering", "Ġrules", "Ġwould", "Ġhelp", "Ġimprove", "Ġthe", "Ġperformance", ".", "</s>", "Acknowled", "gments", "</s>", "We", "Ġthank", "ĠNob", "u", "hiro", "ĠK", "aji", "Ġfor", "Ġproviding", "Ġthe", "ĠAC", "P", "ĠCorpus", "Ġand", "ĠHiro", "k", "az", "u", "ĠK", "iy", "om", "aru", "Ġand", "ĠY", "ud", "ai", "ĠK", "ish", "imoto", "Ġfor", "Ġtheir", "Ġhelp", "Ġin", "Ġextracting", "Ġevent", "Ġpairs", ".", "ĠThis", "Ġwork", "Ġwas", "Ġpartially", "Ġsupported", "Ġby", "ĠYahoo", "!", "ĠJapan", "ĠCorporation", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠPositive", "ĠWords", "</s>", "å", "ĸ", "ľ", "ãģ", "¶", "Ġ(", "re", "joice", "),", "Ġå", "¬", "ī", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġglad", "),", "Ġæ", "¥", "½", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġpleasant", "),", "Ġå", "¹", "¸", "ãģ", "Ľ", "Ġ(", "be", "Ġhappy", "),", "Ġæ", "Ħ", "Ł", "åĭ", "ķ", "Ġ(", "be", "Ġimpressed", "),", "Ġè", "Ī", "Ī", "å¥", "®", "Ġ(", "be", "Ġexcited", "),", "Ġæ", "ĩ", "IJ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "feel", "Ġnostalgic", "),", "Ġå", "¥", "½", "ãģį", "Ġ(", "like", "),", "Ġå", "°", "Ĭ", "æķ", "¬", "Ġ(", "respect", "),", "Ġå", "®", "ī", "å¿", "ĥ", "Ġ(", "be", "Ġrelieved", "),", "Ġæ", "Ħ", "Ł", "å¿", "ĥ", "Ġ(", "ad", "mire", "),", "Ġè", "IJ", "½", "ãģ", "¡", "ç", "Ŀ", "Ģ", "ãģı", "Ġ(", "be", "Ġcalm", "),", "Ġæ", "º", "Ģ", "è", "¶", "³", "Ġ(", "be", "Ġsatisfied", "),", "Ġç", "Ļ", "Ĵ", "ãģķ", "ãĤĮ", "ãĤĭ", "Ġ(", "be", "Ġhealed", "),", "Ġand", "Ġ", "ãĤ¹", "ãĥĥ", "ãĤŃ", "ãĥª", "Ġ(", "be", "Ġrefreshed", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠNegative", "ĠWords", "</s>", "æĢ", "Ĵ", "ãĤĭ", "Ġ(", "get", "Ġangry", "),", "Ġæ", "Ĥ", "²", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġsad", "),", "Ġå", "¯", "Ĥ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġlonely", "),", "Ġæ", "Ģ", "ĸ", "ãģĦ", "Ġ(", "be", "Ġscared", "),", "Ġ", "ä¸į", "å®", "ī", "Ġ(", "feel", "Ġanxious", "),", "Ġæ", "ģ", "¥", "ãģ", "ļ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġembarrassed", "),", "Ġå", "«", "Į", "Ġ(", "hate", "),", "Ġè", "IJ", "½", "ãģ", "¡", "è", "¾", "¼", "ãĤ", "Ģ", "Ġ(", "feel", "Ġdown", "),", "Ġé", "Ģ", "Ģ", "å", "±", "Ī", "Ġ(", "be", "Ġbored", "),", "Ġç", "µ", "¶", "æľ", "Ľ", "Ġ(", "feel", "Ġhopeless", "),", "Ġè", "¾", "Ľ", "ãģĦ", "Ġ(", "have", "Ġa", "Ġhard", "Ġtime", "),", "Ġå", "Ľ", "°", "ãĤĭ", "Ġ(", "have", "Ġtrouble", "),", "Ġæ", "Ĩ", "Ĥ", "é", "¬", "±", "Ġ(", "be", "Ġdepressed", "),", "Ġå", "¿", "ĥ", "é", "ħ", "į", "Ġ(", "be", "Ġworried", "),", "Ġand", "Ġæ", "ĥ", "ħ", "ãģ", "ij", "ãģª", "ãģĦ", "Ġ(", "be", "Ġsorry", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠBi", "GR", "U", "</s>", "The", "Ġdimension", "Ġof", "Ġthe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġ256", ".", "ĠThe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġinitialized", "Ġwith", "Ġthe", "Ġword", "Ġembed", "d", "ings", "Ġpret", "rained", "Ġusing", "Ġthe", "ĠWeb", "Ġcorpus", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "Ġthe", "Ġmorph", "ological", "Ġanaly", "zer", "ĠJ", "uman", "++", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ100", ",", "000", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ2", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ256", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠMoment", "um", "ĠSG", "D", "ĠB", "IB", "REF", "21", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ1024", ".", "ĠWe", "Ġran", "Ġ100", "Ġepoch", "s", "Ġand", "Ġselected", "Ġthe", "Ġsnapshot", "Ġthat", "Ġachieved", "Ġthe", "Ġhighest", "Ġscore", "Ġfor", "Ġthe", "Ġdev", "Ġset", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠB", "ERT", "</s>", "We", "Ġused", "Ġa", "ĠJapanese", "ĠB", "ERT", "Ġmodel", "Ġpret", "rained", "Ġwith", "ĠJapanese", "ĠWikipedia", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "ĠJ", "uman", "++", ",", "Ġand", "Ġwords", "Ġwere", "Ġbroken", "Ġinto", "Ġsub", "words", "Ġby", "Ġapplying", "ĠB", "PE", "ĠB", "IB", "REF", "20", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ32", ",", "000", ".", "ĠThe", "Ġmaximum", "Ġlength", "Ġof", "Ġan", "Ġinput", "Ġsequence", "Ġwas", "Ġ128", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ12", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ768", ".", "ĠThe", "Ġnumber", "Ġof", "Ġself", "-", "att", "ention", "Ġheads", "Ġwas", "Ġ12", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠAdam", "ĠB", "IB", "REF", "19", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ32", ".", "ĠWe", "Ġran", "Ġ1", "Ġepoch", "." ], "paragraphs": [ "Introduction", "Affective events BIBREF0 are events that typically affect people in positive or negative ways. For example, getting money and playing sports are usually positive to the experiencers; catching cold and losing one's wallet are negative. Understanding affective events is important to various natural language processing (NLP) applications such as dialogue systems BIBREF1, question-answering systems BIBREF2, and humor recognition BIBREF3. In this paper, we work on recognizing the polarity of an affective event that is represented by a score ranging from $-1$ (negative) to 1 (positive).", "Learning affective events is challenging because, as the examples above suggest, the polarity of an event is not necessarily predictable from its constituent words. Combined with the unbounded combinatorial nature of language, the non-compositionality of affective polarity entails the need for large amounts of world knowledge, which can hardly be learned from small annotated data.", "In this paper, we propose a simple and effective method for learning affective events that only requires a very small seed lexicon and a large raw corpus. As illustrated in Figure FIGREF1, our key idea is that we can exploit discourse relations BIBREF4 to efficiently propagate polarity from seed predicates that directly report one's emotions (e.g., “to be glad” is positive). Suppose that events $x_1$ are $x_2$ are in the discourse relation of Cause (i.e., $x_1$ causes $x_2$). If the seed lexicon suggests $x_2$ is positive, $x_1$ is also likely to be positive because it triggers the positive emotion. The fact that $x_2$ is known to be negative indicates the negative polarity of $x_1$. Similarly, if $x_1$ and $x_2$ are in the discourse relation of Concession (i.e., $x_2$ in spite of $x_1$), the reverse of $x_2$'s polarity can be propagated to $x_1$. Even if $x_2$'s polarity is not known in advance, we can exploit the tendency of $x_1$ and $x_2$ to be of the same polarity (for Cause) or of the reverse polarity (for Concession) although the heuristic is not exempt from counterexamples. We transform this idea into objective functions and train neural network models that predict the polarity of a given event.", "We trained the models using a Japanese web corpus. Given the minimum amount of supervision, they performed well. In addition, the combination of annotated and unannotated data yielded a gain over a purely supervised baseline when labeled data were small.", "Related Work", "Learning affective events is closely related to sentiment analysis. Whereas sentiment analysis usually focuses on the polarity of what are described (e.g., movies), we work on how people are typically affected by events. In sentiment analysis, much attention has been paid to compositionality. Word-level polarity BIBREF5, BIBREF6, BIBREF7 and the roles of negation and intensification BIBREF8, BIBREF6, BIBREF9 are among the most important topics. In contrast, we are more interested in recognizing the sentiment polarity of an event that pertains to commonsense knowledge (e.g., getting money and catching cold).", "Label propagation from seed instances is a common approach to inducing sentiment polarities. While BIBREF5 and BIBREF10 worked on word- and phrase-level polarities, BIBREF0 dealt with event-level polarities. BIBREF5 and BIBREF10 linked instances using co-occurrence information and/or phrase-level coordinations (e.g., “$A$ and $B$” and “$A$ but $B$”). We shift our scope to event pairs that are more complex than phrase pairs, and consequently exploit discourse connectives as event-level counterparts of phrase-level conjunctions.", "BIBREF0 constructed a network of events using word embedding-derived similarities. Compared with this method, our discourse relation-based linking of events is much simpler and more intuitive.", "Some previous studies made use of document structure to understand the sentiment. BIBREF11 proposed a sentiment-specific pre-training strategy using unlabeled dialog data (tweet-reply pairs). BIBREF12 proposed a method of building a polarity-tagged corpus (ACP Corpus). They automatically gathered sentences that had positive or negative opinions utilizing HTML layout structures in addition to linguistic patterns. Our method depends only on raw texts and thus has wider applicability.", "Proposed Method", "Proposed Method ::: Polarity Function", "Our goal is to learn the polarity function $p(x)$, which predicts the sentiment polarity score of an event $x$. We approximate $p(x)$ by a neural network with the following form:", "${\\rm Encoder}$ outputs a vector representation of the event $x$. ${\\rm Linear}$ is a fully-connected layer and transforms the representation into a scalar. ${\\rm tanh}$ is the hyperbolic tangent and transforms the scalar into a score ranging from $-1$ to 1. In Section SECREF21, we consider two specific implementations of ${\\rm Encoder}$.", "Proposed Method ::: Discourse Relation-Based Event Pairs", "Our method requires a very small seed lexicon and a large raw corpus. We assume that we can automatically extract discourse-tagged event pairs, $(x_{i1}, x_{i2})$ ($i=1, \\cdots $) from the raw corpus. We refer to $x_{i1}$ and $x_{i2}$ as former and latter events, respectively. As shown in Figure FIGREF1, we limit our scope to two discourse relations: Cause and Concession.", "The seed lexicon consists of positive and negative predicates. If the predicate of an extracted event is in the seed lexicon and does not involve complex phenomena like negation, we assign the corresponding polarity score ($+1$ for positive events and $-1$ for negative events) to the event. We expect the model to automatically learn complex phenomena through label propagation. Based on the availability of scores and the types of discourse relations, we classify the extracted event pairs into the following three types.", "Proposed Method ::: Discourse Relation-Based Event Pairs ::: AL (Automatically Labeled Pairs)", "The seed lexicon matches (1) the latter event but (2) not the former event, and (3) their discourse relation type is Cause or Concession. If the discourse relation type is Cause, the former event is given the same score as the latter. Likewise, if the discourse relation type is Concession, the former event is given the opposite of the latter's score. They are used as reference scores during training.", "Proposed Method ::: Discourse Relation-Based Event Pairs ::: CA (Cause Pairs)", "The seed lexicon matches neither the former nor the latter event, and their discourse relation type is Cause. We assume the two events have the same polarities.", "Proposed Method ::: Discourse Relation-Based Event Pairs ::: CO (Concession Pairs)", "The seed lexicon matches neither the former nor the latter event, and their discourse relation type is Concession. We assume the two events have the reversed polarities.", "Proposed Method ::: Loss Functions", "Using AL, CA, and CO data, we optimize the parameters of the polarity function $p(x)$. We define a loss function for each of the three types of event pairs and sum up the multiple loss functions.", "We use mean squared error to construct loss functions. For the AL data, the loss function is defined as:", "where $x_{i1}$ and $x_{i2}$ are the $i$-th pair of the AL data. $r_{i1}$ and $r_{i2}$ are the automatically-assigned scores of $x_{i1}$ and $x_{i2}$, respectively. $N_{\\rm AL}$ is the total number of AL pairs, and $\\lambda _{\\rm AL}$ is a hyperparameter.", "For the CA data, the loss function is defined as:", "$y_{i1}$ and $y_{i2}$ are the $i$-th pair of the CA pairs. $N_{\\rm CA}$ is the total number of CA pairs. $\\lambda _{\\rm CA}$ and $\\mu $ are hyperparameters. The first term makes the scores of the two events closer while the second term prevents the scores from shrinking to zero.", "The loss function for the CO data is defined analogously:", "The difference is that the first term makes the scores of the two events distant from each other.", "Experiments", "Experiments ::: Dataset", "Experiments ::: Dataset ::: AL, CA, and CO", "As a raw corpus, we used a Japanese web corpus that was compiled through the procedures proposed by BIBREF13. To extract event pairs tagged with discourse relations, we used the Japanese dependency parser KNP and in-house postprocessing scripts BIBREF14. KNP used hand-written rules to segment each sentence into what we conventionally called clauses (mostly consecutive text chunks), each of which contained one main predicate. KNP also identified the discourse relations of event pairs if explicit discourse connectives BIBREF4 such as “ので” (because) and “のに” (in spite of) were present. We treated Cause/Reason (原因・理由) and Condition (条件) in the original tagset BIBREF15 as Cause and Concession (逆接) as Concession, respectively. Here is an example of event pair extraction.", ". 重大な失敗を犯したので、仕事をクビになった。", "Because [I] made a serious mistake, [I] got fired.", "From this sentence, we extracted the event pair of “重大な失敗を犯す” ([I] make a serious mistake) and “仕事をクビになる” ([I] get fired), and tagged it with Cause.", "We constructed our seed lexicon consisting of 15 positive words and 15 negative words, as shown in Section SECREF27. From the corpus of about 100 million sentences, we obtained 1.4 millions event pairs for AL, 41 millions for CA, and 6 millions for CO. We randomly selected subsets of AL event pairs such that positive and negative latter events were equal in size. We also sampled event pairs for each of CA and CO such that it was five times larger than AL. The results are shown in Table TABREF16.", "Experiments ::: Dataset ::: ACP (ACP Corpus)", "We used the latest version of the ACP Corpus BIBREF12 for evaluation. It was used for (semi-)supervised training as well. Extracted from Japanese websites using HTML layouts and linguistic patterns, the dataset covered various genres. For example, the following two sentences were labeled positive and negative, respectively:", ". 作業が楽だ。", "The work is easy.", ". 駐車場がない。", "There is no parking lot.", "Although the ACP corpus was originally constructed in the context of sentiment analysis, we found that it could roughly be regarded as a collection of affective events. We parsed each sentence and extracted the last clause in it. The train/dev/test split of the data is shown in Table TABREF19.", "The objective function for supervised training is:", "where $v_i$ is the $i$-th event, $R_i$ is the reference score of $v_i$, and $N_{\\rm ACP}$ is the number of the events of the ACP Corpus.", "To optimize the hyperparameters, we used the dev set of the ACP Corpus. For the evaluation, we used the test set of the ACP Corpus. The model output was classified as positive if $p(x) > 0$ and negative if $p(x) \\le 0$.", "Experiments ::: Model Configurations", "As for ${\\rm Encoder}$, we compared two types of neural networks: BiGRU and BERT. GRU BIBREF16 is a recurrent neural network sequence encoder. BiGRU reads an input sequence forward and backward and the output is the concatenation of the final forward and backward hidden states.", "BERT BIBREF17 is a pre-trained multi-layer bidirectional Transformer BIBREF18 encoder. Its output is the final hidden state corresponding to the special classification tag ([CLS]). For the details of ${\\rm Encoder}$, see Sections SECREF30.", "We trained the model with the following four combinations of the datasets: AL, AL+CA+CO (two proposed models), ACP (supervised), and ACP+AL+CA+CO (semi-supervised). The corresponding objective functions were: $\\mathcal {L}_{\\rm AL}$, $\\mathcal {L}_{\\rm AL} + \\mathcal {L}_{\\rm CA} + \\mathcal {L}_{\\rm CO}$, $\\mathcal {L}_{\\rm ACP}$, and $\\mathcal {L}_{\\rm ACP} + \\mathcal {L}_{\\rm AL} + \\mathcal {L}_{\\rm CA} + \\mathcal {L}_{\\rm CO}$.", "Experiments ::: Results and Discussion", "Table TABREF23 shows accuracy. As the Random baseline suggests, positive and negative labels were distributed evenly. The Random+Seed baseline made use of the seed lexicon and output the corresponding label (or the reverse of it for negation) if the event's predicate is in the seed lexicon. We can see that the seed lexicon itself had practically no impact on prediction.", "The models in the top block performed considerably better than the random baselines. The performance gaps with their (semi-)supervised counterparts, shown in the middle block, were less than 7%. This demonstrates the effectiveness of discourse relation-based label propagation.", "Comparing the model variants, we obtained the highest score with the BiGRU encoder trained with the AL+CA+CO dataset. BERT was competitive but its performance went down if CA and CO were used in addition to AL. We conjecture that BERT was more sensitive to noises found more frequently in CA and CO.", "Contrary to our expectations, supervised models (ACP) outperformed semi-supervised models (ACP+AL+CA+CO). This suggests that the training set of 0.6 million events is sufficiently large for training the models. For comparison, we trained the models with a subset (6,000 events) of the ACP dataset. As the results shown in Table TABREF24 demonstrate, our method is effective when labeled data are small.", "The result of hyperparameter optimization for the BiGRU encoder was as follows:", "As the CA and CO pairs were equal in size (Table TABREF16), $\\lambda _{\\rm CA}$ and $\\lambda _{\\rm CO}$ were comparable values. $\\lambda _{\\rm CA}$ was about one-third of $\\lambda _{\\rm CO}$, and this indicated that the CA pairs were noisier than the CO pairs. A major type of CA pairs that violates our assumption was in the form of “$\\textit {problem}_{\\text{negative}}$ causes $\\textit {solution}_{\\text{positive}}$”:", ". (悪いところがある, よくなるように努力する)", "(there is a bad point, [I] try to improve [it])", "The polarities of the two events were reversed in spite of the Cause relation, and this lowered the value of $\\lambda _{\\rm CA}$.", "Some examples of model outputs are shown in Table TABREF26. The first two examples suggest that our model successfully learned negation without explicit supervision. Similarly, the next two examples differ only in voice but the model correctly recognized that they had opposite polarities. The last two examples share the predicate “落とす\" (drop) and only the objects are different. The second event “肩を落とす\" (lit. drop one's shoulders) is an idiom that expresses a disappointed feeling. The examples demonstrate that our model correctly learned non-compositional expressions.", "Conclusion", "In this paper, we proposed to use discourse relations to effectively propagate polarities of affective events from seeds. Experiments show that, even with a minimal amount of supervision, the proposed method performed well.", "Although event pairs linked by discourse analysis are shown to be useful, they nevertheless contain noises. Adding linguistically-motivated filtering rules would help improve the performance.", "Acknowledgments", "We thank Nobuhiro Kaji for providing the ACP Corpus and Hirokazu Kiyomaru and Yudai Kishimoto for their help in extracting event pairs. This work was partially supported by Yahoo! Japan Corporation.", "Appendices ::: Seed Lexicon ::: Positive Words", "喜ぶ (rejoice), 嬉しい (be glad), 楽しい (be pleasant), 幸せ (be happy), 感動 (be impressed), 興奮 (be excited), 懐かしい (feel nostalgic), 好き (like), 尊敬 (respect), 安心 (be relieved), 感心 (admire), 落ち着く (be calm), 満足 (be satisfied), 癒される (be healed), and スッキリ (be refreshed).", "Appendices ::: Seed Lexicon ::: Negative Words", "怒る (get angry), 悲しい (be sad), 寂しい (be lonely), 怖い (be scared), 不安 (feel anxious), 恥ずかしい (be embarrassed), 嫌 (hate), 落ち込む (feel down), 退屈 (be bored), 絶望 (feel hopeless), 辛い (have a hard time), 困る (have trouble), 憂鬱 (be depressed), 心配 (be worried), and 情けない (be sorry).", "Appendices ::: Settings of Encoder ::: BiGRU", "The dimension of the embedding layer was 256. The embedding layer was initialized with the word embeddings pretrained using the Web corpus. The input sentences were segmented into words by the morphological analyzer Juman++. The vocabulary size was 100,000. The number of hidden layers was 2. The dimension of hidden units was 256. The optimizer was Momentum SGD BIBREF21. The mini-batch size was 1024. We ran 100 epochs and selected the snapshot that achieved the highest score for the dev set.", "Appendices ::: Settings of Encoder ::: BERT", "We used a Japanese BERT model pretrained with Japanese Wikipedia. The input sentences were segmented into words by Juman++, and words were broken into subwords by applying BPE BIBREF20. The vocabulary size was 32,000. The maximum length of an input sequence was 128. The number of hidden layers was 12. The dimension of hidden units was 768. The number of self-attention heads was 12. The optimizer was Adam BIBREF19. The mini-batch size was 32. We ran 1 epoch." ], "question": "What is the seed lexicon?", "question_id": "753990d0b621d390ed58f20c4d9e4f065f0dc672", "question_tokens": [ "What", "Ġis", "Ġthe", "Ġseed", "Ġlex", "icon", "?" ] }
[ "a", "Ġvocabulary", "Ġof", "Ġpositive", "Ġand", "Ġnegative", "Ġpred", "icates", "Ġthat", "Ġhelps", "Ġdetermine", "Ġthe", "Ġpol", "arity", "Ġscore", "Ġof", "Ġan", "Ġevent" ]
[ "<s>", "What", "Ġis", "Ġthe", "Ġseed", "Ġlex", "icon", "?", "</s>", "Introduction", "</s>", "A", "ffect", "ive", "Ġevents", "ĠB", "IB", "REF", "0", "Ġare", "Ġevents", "Ġthat", "Ġtypically", "Ġaffect", "Ġpeople", "Ġin", "Ġpositive", "Ġor", "Ġnegative", "Ġways", ".", "ĠFor", "Ġexample", ",", "Ġgetting", "Ġmoney", "Ġand", "Ġplaying", "Ġsports", "Ġare", "Ġusually", "Ġpositive", "Ġto", "Ġthe", "Ġexperien", "cers", ";", "Ġcatching", "Ġcold", "Ġand", "Ġlosing", "Ġone", "'s", "Ġwallet", "Ġare", "Ġnegative", ".", "ĠUnderstanding", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġimportant", "Ġto", "Ġvarious", "Ġnatural", "Ġlanguage", "Ġprocessing", "Ġ(", "N", "LP", ")", "Ġapplications", "Ġsuch", "Ġas", "Ġdialogue", "Ġsystems", "ĠB", "IB", "REF", "1", ",", "Ġquestion", "-", "ans", "w", "ering", "Ġsystems", "ĠB", "IB", "REF", "2", ",", "Ġand", "Ġhumor", "Ġrecognition", "ĠB", "IB", "REF", "3", ".", "ĠIn", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġwork", "Ġon", "Ġrecognizing", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġaffect", "ive", "Ġevent", "Ġthat", "Ġis", "Ġrepresented", "Ġby", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġ(", "negative", ")", "Ġto", "Ġ1", "Ġ(", "positive", ").", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġchallenging", "Ġbecause", ",", "Ġas", "Ġthe", "Ġexamples", "Ġabove", "Ġsuggest", ",", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġis", "Ġnot", "Ġnecessarily", "Ġpredictable", "Ġfrom", "Ġits", "Ġconstituent", "Ġwords", ".", "ĠCombined", "Ġwith", "Ġthe", "Ġunb", "ounded", "Ġcomb", "inator", "ial", "Ġnature", "Ġof", "Ġlanguage", ",", "Ġthe", "Ġnon", "-", "com", "position", "ality", "Ġof", "Ġaffect", "ive", "Ġpol", "arity", "Ġentails", "Ġthe", "Ġneed", "Ġfor", "Ġlarge", "Ġamounts", "Ġof", "Ġworld", "Ġknowledge", ",", "Ġwhich", "Ġcan", "Ġhardly", "Ġbe", "Ġlearned", "Ġfrom", "Ġsmall", "Ġannot", "ated", "Ġdata", ".", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġpropose", "Ġa", "Ġsimple", "Ġand", "Ġeffective", "Ġmethod", "Ġfor", "Ġlearning", "Ġaffect", "ive", "Ġevents", "Ġthat", "Ġonly", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠAs", "Ġillustrated", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġour", "Ġkey", "Ġidea", "Ġis", "Ġthat", "Ġwe", "Ġcan", "Ġexploit", "Ġdiscourse", "Ġrelations", "ĠB", "IB", "REF", "4", "Ġto", "Ġefficiently", "Ġpropagate", "Ġpol", "arity", "Ġfrom", "Ġseed", "Ġpred", "icates", "Ġthat", "Ġdirectly", "Ġreport", "Ġone", "'s", "Ġemotions", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "to", "Ġbe", "Ġglad", "âĢ", "Ŀ", "Ġis", "Ġpositive", ").", "ĠSuppose", "Ġthat", "Ġevents", "Ġ$", "x", "_", "1", "$", "Ġare", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCause", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "1", "$", "Ġcauses", "Ġ$", "x", "_", "2", "$", ").", "ĠIf", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġsuggests", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġpositive", ",", "Ġ$", "x", "_", "1", "$", "Ġis", "Ġalso", "Ġlikely", "Ġto", "Ġbe", "Ġpositive", "Ġbecause", "Ġit", "Ġtriggers", "Ġthe", "Ġpositive", "Ġemotion", ".", "ĠThe", "Ġfact", "Ġthat", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġknown", "Ġto", "Ġbe", "Ġnegative", "Ġindicates", "Ġthe", "Ġnegative", "Ġpol", "arity", "Ġof", "Ġ$", "x", "_", "1", "$.", "ĠSimilarly", ",", "Ġif", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCon", "cession", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "2", "$", "Ġin", "Ġspite", "Ġof", "Ġ$", "x", "_", "1", "$", "),", "Ġthe", "Ġreverse", "Ġof", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġcan", "Ġbe", "Ġpropag", "ated", "Ġto", "Ġ$", "x", "_", "1", "$.", "ĠEven", "Ġif", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġis", "Ġnot", "Ġknown", "Ġin", "Ġadvance", ",", "Ġwe", "Ġcan", "Ġexploit", "Ġthe", "Ġtendency", "Ġof", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġto", "Ġbe", "Ġof", "Ġthe", "Ġsame", "Ġpol", "arity", "Ġ(", "for", "ĠCause", ")", "Ġor", "Ġof", "Ġthe", "Ġreverse", "Ġpol", "arity", "Ġ(", "for", "ĠCon", "cession", ")", "Ġalthough", "Ġthe", "Ġhe", "uristic", "Ġis", "Ġnot", "Ġexempt", "Ġfrom", "Ġcount", "ere", "x", "amples", ".", "ĠWe", "Ġtransform", "Ġthis", "Ġidea", "Ġinto", "Ġobjective", "Ġfunctions", "Ġand", "Ġtrain", "Ġneural", "Ġnetwork", "Ġmodels", "Ġthat", "Ġpredict", "Ġthe", "Ġpol", "arity", "Ġof", "Ġa", "Ġgiven", "Ġevent", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodels", "Ġusing", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", ".", "ĠGiven", "Ġthe", "Ġminimum", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthey", "Ġperformed", "Ġwell", ".", "ĠIn", "Ġaddition", ",", "Ġthe", "Ġcombination", "Ġof", "Ġannot", "ated", "Ġand", "Ġun", "annot", "ated", "Ġdata", "Ġyielded", "Ġa", "Ġgain", "Ġover", "Ġa", "Ġpurely", "Ġsupervised", "Ġbaseline", "Ġwhen", "Ġlabeled", "Ġdata", "Ġwere", "Ġsmall", ".", "</s>", "Related", "ĠWork", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġclosely", "Ġrelated", "Ġto", "Ġsentiment", "Ġanalysis", ".", "ĠWhereas", "Ġsentiment", "Ġanalysis", "Ġusually", "Ġfocuses", "Ġon", "Ġthe", "Ġpol", "arity", "Ġof", "Ġwhat", "Ġare", "Ġdescribed", "Ġ(", "e", ".", "g", ".,", "Ġmovies", "),", "Ġwe", "Ġwork", "Ġon", "Ġhow", "Ġpeople", "Ġare", "Ġtypically", "Ġaffected", "Ġby", "Ġevents", ".", "ĠIn", "Ġsentiment", "Ġanalysis", ",", "Ġmuch", "Ġattention", "Ġhas", "Ġbeen", "Ġpaid", "Ġto", "Ġcomposition", "ality", ".", "ĠWord", "-", "level", "Ġpol", "arity", "ĠB", "IB", "REF", "5", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "7", "Ġand", "Ġthe", "Ġroles", "Ġof", "Ġneg", "ation", "Ġand", "Ġintens", "ification", "ĠB", "IB", "REF", "8", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "9", "Ġare", "Ġamong", "Ġthe", "Ġmost", "Ġimportant", "Ġtopics", ".", "ĠIn", "Ġcontrast", ",", "Ġwe", "Ġare", "Ġmore", "Ġinterested", "Ġin", "Ġrecognizing", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġthat", "Ġper", "tains", "Ġto", "Ġcommons", "ense", "Ġknowledge", "Ġ(", "e", ".", "g", ".,", "Ġgetting", "Ġmoney", "Ġand", "Ġcatching", "Ġcold", ").", "</s>", "Label", "Ġpropagation", "Ġfrom", "Ġseed", "Ġinstances", "Ġis", "Ġa", "Ġcommon", "Ġapproach", "Ġto", "Ġinducing", "Ġsentiment", "Ġpolar", "ities", ".", "ĠWhile", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġworked", "Ġon", "Ġword", "-", "Ġand", "Ġphrase", "-", "level", "Ġpolar", "ities", ",", "ĠB", "IB", "REF", "0", "Ġdealt", "Ġwith", "Ġevent", "-", "level", "Ġpolar", "ities", ".", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġlinked", "Ġinstances", "Ġusing", "Ġco", "-", "occ", "urrence", "Ġinformation", "Ġand", "/", "or", "Ġphrase", "-", "level", "Ġcoord", "inations", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "$", "A", "$", "Ġand", "Ġ$", "B", "$", "âĢ", "Ŀ", "Ġand", "ĠâĢ", "ľ", "$", "A", "$", "Ġbut", "Ġ$", "B", "$", "âĢ", "Ŀ", ").", "ĠWe", "Ġshift", "Ġour", "Ġscope", "Ġto", "Ġevent", "Ġpairs", "Ġthat", "Ġare", "Ġmore", "Ġcomplex", "Ġthan", "Ġphrase", "Ġpairs", ",", "Ġand", "Ġconsequently", "Ġexploit", "Ġdiscourse", "Ġconnect", "ives", "Ġas", "Ġevent", "-", "level", "Ġcounterparts", "Ġof", "Ġphrase", "-", "level", "Ġconj", "unctions", ".", "</s>", "BI", "B", "REF", "0", "Ġconstructed", "Ġa", "Ġnetwork", "Ġof", "Ġevents", "Ġusing", "Ġword", "Ġembed", "ding", "-", "derived", "Ġsimilarities", ".", "ĠCompared", "Ġwith", "Ġthis", "Ġmethod", ",", "Ġour", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlinking", "Ġof", "Ġevents", "Ġis", "Ġmuch", "Ġsimpler", "Ġand", "Ġmore", "Ġintuitive", ".", "</s>", "Some", "Ġprevious", "Ġstudies", "Ġmade", "Ġuse", "Ġof", "Ġdocument", "Ġstructure", "Ġto", "Ġunderstand", "Ġthe", "Ġsentiment", ".", "ĠB", "IB", "REF", "11", "Ġproposed", "Ġa", "Ġsentiment", "-", "specific", "Ġpre", "-", "training", "Ġstrategy", "Ġusing", "Ġunl", "abel", "ed", "Ġdialog", "Ġdata", "Ġ(", "t", "weet", "-", "reply", "Ġpairs", ").", "ĠB", "IB", "REF", "12", "Ġproposed", "Ġa", "Ġmethod", "Ġof", "Ġbuilding", "Ġa", "Ġpol", "arity", "-", "tag", "ged", "Ġcorpus", "Ġ(", "ACP", "ĠCorpus", ").", "ĠThey", "Ġautomatically", "Ġgathered", "Ġsentences", "Ġthat", "Ġhad", "Ġpositive", "Ġor", "Ġnegative", "Ġopinions", "Ġutilizing", "ĠHTML", "Ġlayout", "Ġstructures", "Ġin", "Ġaddition", "Ġto", "Ġlinguistic", "Ġpatterns", ".", "ĠOur", "Ġmethod", "Ġdepends", "Ġonly", "Ġon", "Ġraw", "Ġtexts", "Ġand", "Ġthus", "Ġhas", "Ġwider", "Ġapplic", "ability", ".", "</s>", "Prop", "osed", "ĠMethod", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠPol", "arity", "ĠFunction", "</s>", "Our", "Ġgoal", "Ġis", "Ġto", "Ġlearn", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$,", "Ġwhich", "Ġpredicts", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġscore", "Ġof", "Ġan", "Ġevent", "Ġ$", "x", "$.", "ĠWe", "Ġapproximate", "Ġ$", "p", "(", "x", ")", "$", "Ġby", "Ġa", "Ġneural", "Ġnetwork", "Ġwith", "Ġthe", "Ġfollowing", "Ġform", ":", "</s>", "$", "{\\", "rm", "ĠEnc", "oder", "}", "$", "Ġoutputs", "Ġa", "Ġvector", "Ġrepresentation", "Ġof", "Ġthe", "Ġevent", "Ġ$", "x", "$.", "Ġ${", "\\", "rm", "ĠLinear", "}", "$", "Ġis", "Ġa", "Ġfully", "-", "connected", "Ġlayer", "Ġand", "Ġtransforms", "Ġthe", "Ġrepresentation", "Ġinto", "Ġa", "Ġscal", "ar", ".", "Ġ${", "\\", "rm", "Ġtan", "h", "}", "$", "Ġis", "Ġthe", "Ġhyper", "b", "olic", "Ġtang", "ent", "Ġand", "Ġtransforms", "Ġthe", "Ġscal", "ar", "Ġinto", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġto", "Ġ1", ".", "ĠIn", "ĠSection", "ĠSEC", "REF", "21", ",", "Ġwe", "Ġconsider", "Ġtwo", "Ġspecific", "Ġimplementations", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$.", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "</s>", "Our", "Ġmethod", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġassume", "Ġthat", "Ġwe", "Ġcan", "Ġautomatically", "Ġextract", "Ġdiscourse", "-", "tag", "ged", "Ġevent", "Ġpairs", ",", "Ġ$(", "x", "_{", "i", "1", "},", "Ġx", "_{", "i", "2", "})", "$", "Ġ($", "i", "=", "1", ",", "Ġ\\", "cd", "ots", "Ġ$", ")", "Ġfrom", "Ġthe", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġrefer", "Ġto", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġas", "Ġformer", "Ġand", "Ġlatter", "Ġevents", ",", "Ġrespectively", ".", "ĠAs", "Ġshown", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġwe", "Ġlimit", "Ġour", "Ġscope", "Ġto", "Ġtwo", "Ġdiscourse", "Ġrelations", ":", "ĠCause", "Ġand", "ĠCon", "cession", ".", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġconsists", "Ġof", "Ġpositive", "Ġand", "Ġnegative", "Ġpred", "icates", ".", "ĠIf", "Ġthe", "Ġpredicate", "Ġof", "Ġan", "Ġextracted", "Ġevent", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġdoes", "Ġnot", "Ġinvolve", "Ġcomplex", "Ġphenomena", "Ġlike", "Ġneg", "ation", ",", "Ġwe", "Ġassign", "Ġthe", "Ġcorresponding", "Ġpol", "arity", "Ġscore", "Ġ($", "+", "1", "$", "Ġfor", "Ġpositive", "Ġevents", "Ġand", "Ġ$", "-", "1", "$", "Ġfor", "Ġnegative", "Ġevents", ")", "Ġto", "Ġthe", "Ġevent", ".", "ĠWe", "Ġexpect", "Ġthe", "Ġmodel", "Ġto", "Ġautomatically", "Ġlearn", "Ġcomplex", "Ġphenomena", "Ġthrough", "Ġlabel", "Ġpropagation", ".", "ĠBased", "Ġon", "Ġthe", "Ġavailability", "Ġof", "Ġscores", "Ġand", "Ġthe", "Ġtypes", "Ġof", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġclassify", "Ġthe", "Ġextracted", "Ġevent", "Ġpairs", "Ġinto", "Ġthe", "Ġfollowing", "Ġthree", "Ġtypes", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠAL", "Ġ(", "Autom", "atically", "ĠLab", "eled", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġ(", "1", ")", "Ġthe", "Ġlatter", "Ġevent", "Ġbut", "Ġ(", "2", ")", "Ġnot", "Ġthe", "Ġformer", "Ġevent", ",", "Ġand", "Ġ(", "3", ")", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", "Ġor", "ĠCon", "cession", ".", "ĠIf", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġsame", "Ġscore", "Ġas", "Ġthe", "Ġlatter", ".", "ĠLikewise", ",", "Ġif", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġopposite", "Ġof", "Ġthe", "Ġlatter", "'s", "Ġscore", ".", "ĠThey", "Ġare", "Ġused", "Ġas", "Ġreference", "Ġscores", "Ġduring", "Ġtraining", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCA", "Ġ(", "Cause", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġsame", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCO", "Ġ(", "Con", "cession", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġreversed", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠLoss", "ĠFunctions", "</s>", "Using", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "Ġdata", ",", "Ġwe", "Ġoptimize", "Ġthe", "Ġparameters", "Ġof", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$.", "ĠWe", "Ġdefine", "Ġa", "Ġloss", "Ġfunction", "Ġfor", "Ġeach", "Ġof", "Ġthe", "Ġthree", "Ġtypes", "Ġof", "Ġevent", "Ġpairs", "Ġand", "Ġsum", "Ġup", "Ġthe", "Ġmultiple", "Ġloss", "Ġfunctions", ".", "</s>", "We", "Ġuse", "Ġmean", "Ġsquared", "Ġerror", "Ġto", "Ġconstruct", "Ġloss", "Ġfunctions", ".", "ĠFor", "Ġthe", "ĠAL", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "where", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠAL", "Ġdata", ".", "Ġ$", "r", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "r", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġautomatically", "-", "ass", "igned", "Ġscores", "Ġof", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$,", "Ġrespectively", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠAL", "Ġpairs", ",", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġa", "Ġhyper", "param", "eter", ".", "</s>", "For", "Ġthe", "ĠCA", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "$", "y", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "y", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠCA", "Ġpairs", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠCA", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠCA", "Ġpairs", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "mu", "Ġ$", "Ġare", "Ġhyper", "param", "eters", ".", "ĠThe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġcloser", "Ġwhile", "Ġthe", "Ġsecond", "Ġterm", "Ġprevents", "Ġthe", "Ġscores", "Ġfrom", "Ġshrinking", "Ġto", "Ġzero", ".", "</s>", "The", "Ġloss", "Ġfunction", "Ġfor", "Ġthe", "ĠCO", "Ġdata", "Ġis", "Ġdefined", "Ġanalog", "ously", ":", "</s>", "The", "Ġdifference", "Ġis", "Ġthat", "Ġthe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġdistant", "Ġfrom", "Ġeach", "Ġother", ".", "</s>", "Exper", "iments", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "</s>", "As", "Ġa", "Ġraw", "Ġcorpus", ",", "Ġwe", "Ġused", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", "Ġthat", "Ġwas", "Ġcompiled", "Ġthrough", "Ġthe", "Ġprocedures", "Ġproposed", "Ġby", "ĠB", "IB", "REF", "13", ".", "ĠTo", "Ġextract", "Ġevent", "Ġpairs", "Ġtagged", "Ġwith", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġused", "Ġthe", "ĠJapanese", "Ġdependency", "Ġparser", "ĠK", "NP", "Ġand", "Ġin", "-", "house", "Ġpost", "processing", "Ġscripts", "ĠB", "IB", "REF", "14", ".", "ĠK", "NP", "Ġused", "Ġhand", "-", "written", "Ġrules", "Ġto", "Ġsegment", "Ġeach", "Ġsentence", "Ġinto", "Ġwhat", "Ġwe", "Ġconvention", "ally", "Ġcalled", "Ġclauses", "Ġ(", "mostly", "Ġconsecutive", "Ġtext", "Ġchunks", "),", "Ġeach", "Ġof", "Ġwhich", "Ġcontained", "Ġone", "Ġmain", "Ġpredicate", ".", "ĠK", "NP", "Ġalso", "Ġidentified", "Ġthe", "Ġdiscourse", "Ġrelations", "Ġof", "Ġevent", "Ġpairs", "Ġif", "Ġexplicit", "Ġdiscourse", "Ġconnect", "ives", "ĠB", "IB", "REF", "4", "Ġsuch", "Ġas", "ĠâĢ", "ľ", "ãģ®", "ãģ§", "âĢ", "Ŀ", "Ġ(", "because", ")", "Ġand", "ĠâĢ", "ľ", "ãģ®", "ãģ«", "âĢ", "Ŀ", "Ġ(", "in", "Ġspite", "Ġof", ")", "Ġwere", "Ġpresent", ".", "ĠWe", "Ġtreated", "ĠCause", "/", "Reason", "Ġ(", "åİ", "Ł", "åĽ", "ł", "ãĥ»", "çIJ", "Ĩ", "çĶ", "±", ")", "Ġand", "ĠCondition", "Ġ(", "æĿ", "¡", "ä»", "¶", ")", "Ġin", "Ġthe", "Ġoriginal", "Ġtag", "set", "ĠB", "IB", "REF", "15", "Ġas", "ĠCause", "Ġand", "ĠCon", "cession", "Ġ(", "éĢ", "Ĩ", "æ", "İ", "¥", ")", "Ġas", "ĠCon", "cession", ",", "Ġrespectively", ".", "ĠHere", "Ġis", "Ġan", "Ġexample", "Ġof", "Ġevent", "Ġpair", "Ġextraction", ".", "</s>", ".", "Ġé", "ĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĹ", "ãģŁ", "ãģ®", "ãģ§", "ãĢģ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãģ£", "ãģŁ", "ãĢĤ", "</s>", "Because", "Ġ[", "I", "]", "Ġmade", "Ġa", "Ġserious", "Ġmistake", ",", "Ġ[", "I", "]", "Ġgot", "Ġfired", ".", "</s>", "From", "Ġthis", "Ġsentence", ",", "Ġwe", "Ġextracted", "Ġthe", "Ġevent", "Ġpair", "Ġof", "ĠâĢ", "ľ", "éĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĻ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġmake", "Ġa", "Ġserious", "Ġmistake", ")", "Ġand", "ĠâĢ", "ľ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãĤĭ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġget", "Ġfired", "),", "Ġand", "Ġtagged", "Ġit", "Ġwith", "ĠCause", ".", "</s>", "We", "Ġconstructed", "Ġour", "Ġseed", "Ġlex", "icon", "Ġconsisting", "Ġof", "Ġ15", "Ġpositive", "Ġwords", "Ġand", "Ġ15", "Ġnegative", "Ġwords", ",", "Ġas", "Ġshown", "Ġin", "ĠSection", "ĠSEC", "REF", "27", ".", "ĠFrom", "Ġthe", "Ġcorpus", "Ġof", "Ġabout", "Ġ100", "Ġmillion", "Ġsentences", ",", "Ġwe", "Ġobtained", "Ġ1", ".", "4", "Ġmillions", "Ġevent", "Ġpairs", "Ġfor", "ĠAL", ",", "Ġ41", "Ġmillions", "Ġfor", "ĠCA", ",", "Ġand", "Ġ6", "Ġmillions", "Ġfor", "ĠCO", ".", "ĠWe", "Ġrandomly", "Ġselected", "Ġsubs", "ets", "Ġof", "ĠAL", "Ġevent", "Ġpairs", "Ġsuch", "Ġthat", "Ġpositive", "Ġand", "Ġnegative", "Ġlatter", "Ġevents", "Ġwere", "Ġequal", "Ġin", "Ġsize", ".", "ĠWe", "Ġalso", "Ġsampled", "Ġevent", "Ġpairs", "Ġfor", "Ġeach", "Ġof", "ĠCA", "Ġand", "ĠCO", "Ġsuch", "Ġthat", "Ġit", "Ġwas", "Ġfive", "Ġtimes", "Ġlarger", "Ġthan", "ĠAL", ".", "ĠThe", "Ġresults", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "16", ".", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAC", "P", "Ġ(", "ACP", "ĠCorpus", ")", "</s>", "We", "Ġused", "Ġthe", "Ġlatest", "Ġversion", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", "ĠB", "IB", "REF", "12", "Ġfor", "Ġevaluation", ".", "ĠIt", "Ġwas", "Ġused", "Ġfor", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġtraining", "Ġas", "Ġwell", ".", "ĠExt", "racted", "Ġfrom", "ĠJapanese", "Ġwebsites", "Ġusing", "ĠHTML", "Ġlayouts", "Ġand", "Ġlinguistic", "Ġpatterns", ",", "Ġthe", "Ġdataset", "Ġcovered", "Ġvarious", "Ġgenres", ".", "ĠFor", "Ġexample", ",", "Ġthe", "Ġfollowing", "Ġtwo", "Ġsentences", "Ġwere", "Ġlabeled", "Ġpositive", "Ġand", "Ġnegative", ",", "Ġrespectively", ":", "</s>", ".", "Ġ", "ä½ľ", "æ", "¥", "Ń", "ãģĮ", "æ", "¥", "½", "ãģł", "ãĢĤ", "</s>", "The", "Ġwork", "Ġis", "Ġeasy", ".", "</s>", ".", "Ġé", "§", "IJ", "è»", "Ĭ", "å", "ł", "´", "ãģĮ", "ãģª", "ãģĦ", "ãĢĤ", "</s>", "There", "Ġis", "Ġno", "Ġparking", "Ġlot", ".", "</s>", "Although", "Ġthe", "ĠAC", "P", "Ġcorpus", "Ġwas", "Ġoriginally", "Ġconstructed", "Ġin", "Ġthe", "Ġcontext", "Ġof", "Ġsentiment", "Ġanalysis", ",", "Ġwe", "Ġfound", "Ġthat", "Ġit", "Ġcould", "Ġroughly", "Ġbe", "Ġregarded", "Ġas", "Ġa", "Ġcollection", "Ġof", "Ġaffect", "ive", "Ġevents", ".", "ĠWe", "Ġparsed", "Ġeach", "Ġsentence", "Ġand", "Ġextracted", "Ġthe", "Ġlast", "Ġclause", "Ġin", "Ġit", ".", "ĠThe", "Ġtrain", "/", "dev", "/", "test", "Ġsplit", "Ġof", "Ġthe", "Ġdata", "Ġis", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "19", ".", "</s>", "The", "Ġobjective", "Ġfunction", "Ġfor", "Ġsupervised", "Ġtraining", "Ġis", ":", "</s>", "where", "Ġ$", "v", "_", "i", "$", "Ġis", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġevent", ",", "Ġ$", "R", "_", "i", "$", "Ġis", "Ġthe", "Ġreference", "Ġscore", "Ġof", "Ġ$", "v", "_", "i", "$,", "Ġand", "Ġ$", "N", "_{", "\\", "rm", "ĠAC", "P", "}", "$", "Ġis", "Ġthe", "Ġnumber", "Ġof", "Ġthe", "Ġevents", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "</s>", "To", "Ġoptimize", "Ġthe", "Ġhyper", "param", "eters", ",", "Ġwe", "Ġused", "Ġthe", "Ġdev", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠFor", "Ġthe", "Ġevaluation", ",", "Ġwe", "Ġused", "Ġthe", "Ġtest", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠThe", "Ġmodel", "Ġoutput", "Ġwas", "Ġclassified", "Ġas", "Ġpositive", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ>", "Ġ0", "$", "Ġand", "Ġnegative", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ\\", "le", "Ġ0", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠModel", "ĠConfig", "urations", "</s>", "As", "Ġfor", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġwe", "Ġcompared", "Ġtwo", "Ġtypes", "Ġof", "Ġneural", "Ġnetworks", ":", "ĠBi", "GR", "U", "Ġand", "ĠB", "ERT", ".", "ĠGR", "U", "ĠB", "IB", "REF", "16", "Ġis", "Ġa", "Ġrecurrent", "Ġneural", "Ġnetwork", "Ġsequence", "Ġenc", "oder", ".", "ĠBi", "GR", "U", "Ġreads", "Ġan", "Ġinput", "Ġsequence", "Ġforward", "Ġand", "Ġbackward", "Ġand", "Ġthe", "Ġoutput", "Ġis", "Ġthe", "Ġconc", "aten", "ation", "Ġof", "Ġthe", "Ġfinal", "Ġforward", "Ġand", "Ġbackward", "Ġhidden", "Ġstates", ".", "</s>", "BER", "T", "ĠB", "IB", "REF", "17", "Ġis", "Ġa", "Ġpre", "-", "trained", "Ġmulti", "-", "layer", "Ġbid", "irection", "al", "ĠTrans", "former", "ĠB", "IB", "REF", "18", "Ġenc", "oder", ".", "ĠIts", "Ġoutput", "Ġis", "Ġthe", "Ġfinal", "Ġhidden", "Ġstate", "Ġcorresponding", "Ġto", "Ġthe", "Ġspecial", "Ġclassification", "Ġtag", "Ġ([", "CL", "S", "]).", "ĠFor", "Ġthe", "Ġdetails", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġsee", "ĠSections", "ĠSEC", "REF", "30", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodel", "Ġwith", "Ġthe", "Ġfollowing", "Ġfour", "Ġcombinations", "Ġof", "Ġthe", "Ġdatasets", ":", "ĠAL", ",", "ĠAL", "+", "CA", "+", "CO", "Ġ(", "two", "Ġproposed", "Ġmodels", "),", "ĠAC", "P", "Ġ(", "super", "vised", "),", "Ġand", "ĠAC", "P", "+", "AL", "+", "CA", "+", "CO", "Ġ(", "se", "mi", "-", "super", "vised", ").", "ĠThe", "Ġcorresponding", "Ġobjective", "Ġfunctions", "Ġwere", ":", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "$,", "Ġand", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠResults", "Ġand", "ĠDiscussion", "</s>", "Table", "ĠT", "AB", "REF", "23", "Ġshows", "Ġaccuracy", ".", "ĠAs", "Ġthe", "ĠRandom", "Ġbaseline", "Ġsuggests", ",", "Ġpositive", "Ġand", "Ġnegative", "Ġlabels", "Ġwere", "Ġdistributed", "Ġevenly", ".", "ĠThe", "ĠRandom", "+", "S", "eed", "Ġbaseline", "Ġmade", "Ġuse", "Ġof", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġoutput", "Ġthe", "Ġcorresponding", "Ġlabel", "Ġ(", "or", "Ġthe", "Ġreverse", "Ġof", "Ġit", "Ġfor", "Ġneg", "ation", ")", "Ġif", "Ġthe", "Ġevent", "'s", "Ġpredicate", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", ".", "ĠWe", "Ġcan", "Ġsee", "Ġthat", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġitself", "Ġhad", "Ġpractically", "Ġno", "Ġimpact", "Ġon", "Ġprediction", ".", "</s>", "The", "Ġmodels", "Ġin", "Ġthe", "Ġtop", "Ġblock", "Ġperformed", "Ġconsiderably", "Ġbetter", "Ġthan", "Ġthe", "Ġrandom", "Ġbas", "elines", ".", "ĠThe", "Ġperformance", "Ġgaps", "Ġwith", "Ġtheir", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġcounterparts", ",", "Ġshown", "Ġin", "Ġthe", "Ġmiddle", "Ġblock", ",", "Ġwere", "Ġless", "Ġthan", "Ġ7", "%.", "ĠThis", "Ġdemonstrates", "Ġthe", "Ġeffectiveness", "Ġof", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlabel", "Ġpropagation", ".", "</s>", "Comp", "aring", "Ġthe", "Ġmodel", "Ġvariants", ",", "Ġwe", "Ġobtained", "Ġthe", "Ġhighest", "Ġscore", "Ġwith", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġtrained", "Ġwith", "Ġthe", "ĠAL", "+", "CA", "+", "CO", "Ġdataset", ".", "ĠB", "ERT", "Ġwas", "Ġcompetitive", "Ġbut", "Ġits", "Ġperformance", "Ġwent", "Ġdown", "Ġif", "ĠCA", "Ġand", "ĠCO", "Ġwere", "Ġused", "Ġin", "Ġaddition", "Ġto", "ĠAL", ".", "ĠWe", "Ġconjecture", "Ġthat", "ĠB", "ERT", "Ġwas", "Ġmore", "Ġsensitive", "Ġto", "Ġnoises", "Ġfound", "Ġmore", "Ġfrequently", "Ġin", "ĠCA", "Ġand", "ĠCO", ".", "</s>", "Cont", "rary", "Ġto", "Ġour", "Ġexpectations", ",", "Ġsupervised", "Ġmodels", "Ġ(", "ACP", ")", "Ġoutper", "formed", "Ġsemi", "-", "super", "vised", "Ġmodels", "Ġ(", "ACP", "+", "AL", "+", "CA", "+", "CO", ").", "ĠThis", "Ġsuggests", "Ġthat", "Ġthe", "Ġtraining", "Ġset", "Ġof", "Ġ0", ".", "6", "Ġmillion", "Ġevents", "Ġis", "Ġsufficiently", "Ġlarge", "Ġfor", "Ġtraining", "Ġthe", "Ġmodels", ".", "ĠFor", "Ġcomparison", ",", "Ġwe", "Ġtrained", "Ġthe", "Ġmodels", "Ġwith", "Ġa", "Ġsubset", "Ġ(", "6", ",", "000", "Ġevents", ")", "Ġof", "Ġthe", "ĠAC", "P", "Ġdataset", ".", "ĠAs", "Ġthe", "Ġresults", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "24", "Ġdemonstrate", ",", "Ġour", "Ġmethod", "Ġis", "Ġeffective", "Ġwhen", "Ġlabeled", "Ġdata", "Ġare", "Ġsmall", ".", "</s>", "The", "Ġresult", "Ġof", "Ġhyper", "param", "eter", "Ġoptimization", "Ġfor", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġwas", "Ġas", "Ġfollows", ":", "</s>", "As", "Ġthe", "ĠCA", "Ġand", "ĠCO", "Ġpairs", "Ġwere", "Ġequal", "Ġin", "Ġsize", "Ġ(", "Table", "ĠT", "AB", "REF", "16", "),", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$", "Ġwere", "Ġcomparable", "Ġvalues", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġwas", "Ġabout", "Ġone", "-", "third", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$,", "Ġand", "Ġthis", "Ġindicated", "Ġthat", "Ġthe", "ĠCA", "Ġpairs", "Ġwere", "Ġno", "is", "ier", "Ġthan", "Ġthe", "ĠCO", "Ġpairs", ".", "ĠA", "Ġmajor", "Ġtype", "Ġof", "ĠCA", "Ġpairs", "Ġthat", "Ġviolates", "Ġour", "Ġassumption", "Ġwas", "Ġin", "Ġthe", "Ġform", "Ġof", "ĠâĢ", "ľ", "$", "\\", "text", "it", "Ġ{", "problem", "}", "_{", "\\", "text", "{", "negative", "}}", "$", "Ġcauses", "Ġ$\\", "text", "it", "Ġ{", "s", "olution", "}", "_{", "\\", "text", "{", "positive", "}}", "$", "âĢ", "Ŀ", ":", "</s>", ".", "Ġ(", "æ", "Ĥª", "ãģĦ", "ãģ¨", "ãģĵ", "ãĤ", "į", "ãģĮ", "ãģĤ", "ãĤĭ", ",", "ĠãĤ", "Ī", "ãģı", "ãģª", "ãĤĭ", "ãĤ", "Ī", "ãģĨ", "ãģ«", "åĬ", "ª", "åĬ", "Ľ", "ãģĻ", "ãĤĭ", ")", "</s>", "(", "there", "Ġis", "Ġa", "Ġbad", "Ġpoint", ",", "Ġ[", "I", "]", "Ġtry", "Ġto", "Ġimprove", "Ġ[", "it", "])", "</s>", "The", "Ġpolar", "ities", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġwere", "Ġreversed", "Ġin", "Ġspite", "Ġof", "Ġthe", "ĠCause", "Ġrelation", ",", "Ġand", "Ġthis", "Ġlowered", "Ġthe", "Ġvalue", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$.", "</s>", "Some", "Ġexamples", "Ġof", "Ġmodel", "Ġoutputs", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "26", ".", "ĠThe", "Ġfirst", "Ġtwo", "Ġexamples", "Ġsuggest", "Ġthat", "Ġour", "Ġmodel", "Ġsuccessfully", "Ġlearned", "Ġneg", "ation", "Ġwithout", "Ġexplicit", "Ġsupervision", ".", "ĠSimilarly", ",", "Ġthe", "Ġnext", "Ġtwo", "Ġexamples", "Ġdiffer", "Ġonly", "Ġin", "Ġvoice", "Ġbut", "Ġthe", "Ġmodel", "Ġcorrectly", "Ġrecognized", "Ġthat", "Ġthey", "Ġhad", "Ġopposite", "Ġpolar", "ities", ".", "ĠThe", "Ġlast", "Ġtwo", "Ġexamples", "Ġshare", "Ġthe", "Ġpredicate", "ĠâĢ", "ľ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "drop", ")", "Ġand", "Ġonly", "Ġthe", "Ġobjects", "Ġare", "Ġdifferent", ".", "ĠThe", "Ġsecond", "Ġevent", "ĠâĢ", "ľ", "è", "Ĥ", "©", "ãĤĴ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "lit", ".", "Ġdrop", "Ġone", "'s", "Ġshoulders", ")", "Ġis", "Ġan", "Ġid", "iom", "Ġthat", "Ġexpresses", "Ġa", "Ġdisappointed", "Ġfeeling", ".", "ĠThe", "Ġexamples", "Ġdemonstrate", "Ġthat", "Ġour", "Ġmodel", "Ġcorrectly", "Ġlearned", "Ġnon", "-", "com", "pos", "itional", "Ġexpressions", ".", "</s>", "Conclusion", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġproposed", "Ġto", "Ġuse", "Ġdiscourse", "Ġrelations", "Ġto", "Ġeffectively", "Ġpropagate", "Ġpolar", "ities", "Ġof", "Ġaffect", "ive", "Ġevents", "Ġfrom", "Ġseeds", ".", "ĠExper", "iments", "Ġshow", "Ġthat", ",", "Ġeven", "Ġwith", "Ġa", "Ġminimal", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthe", "Ġproposed", "Ġmethod", "Ġperformed", "Ġwell", ".", "</s>", "Although", "Ġevent", "Ġpairs", "Ġlinked", "Ġby", "Ġdiscourse", "Ġanalysis", "Ġare", "Ġshown", "Ġto", "Ġbe", "Ġuseful", ",", "Ġthey", "Ġnevertheless", "Ġcontain", "Ġnoises", ".", "ĠAdding", "Ġlingu", "istically", "-", "mot", "ivated", "Ġfiltering", "Ġrules", "Ġwould", "Ġhelp", "Ġimprove", "Ġthe", "Ġperformance", ".", "</s>", "Acknowled", "gments", "</s>", "We", "Ġthank", "ĠNob", "u", "hiro", "ĠK", "aji", "Ġfor", "Ġproviding", "Ġthe", "ĠAC", "P", "ĠCorpus", "Ġand", "ĠHiro", "k", "az", "u", "ĠK", "iy", "om", "aru", "Ġand", "ĠY", "ud", "ai", "ĠK", "ish", "imoto", "Ġfor", "Ġtheir", "Ġhelp", "Ġin", "Ġextracting", "Ġevent", "Ġpairs", ".", "ĠThis", "Ġwork", "Ġwas", "Ġpartially", "Ġsupported", "Ġby", "ĠYahoo", "!", "ĠJapan", "ĠCorporation", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠPositive", "ĠWords", "</s>", "å", "ĸ", "ľ", "ãģ", "¶", "Ġ(", "re", "joice", "),", "Ġå", "¬", "ī", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġglad", "),", "Ġæ", "¥", "½", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġpleasant", "),", "Ġå", "¹", "¸", "ãģ", "Ľ", "Ġ(", "be", "Ġhappy", "),", "Ġæ", "Ħ", "Ł", "åĭ", "ķ", "Ġ(", "be", "Ġimpressed", "),", "Ġè", "Ī", "Ī", "å¥", "®", "Ġ(", "be", "Ġexcited", "),", "Ġæ", "ĩ", "IJ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "feel", "Ġnostalgic", "),", "Ġå", "¥", "½", "ãģį", "Ġ(", "like", "),", "Ġå", "°", "Ĭ", "æķ", "¬", "Ġ(", "respect", "),", "Ġå", "®", "ī", "å¿", "ĥ", "Ġ(", "be", "Ġrelieved", "),", "Ġæ", "Ħ", "Ł", "å¿", "ĥ", "Ġ(", "ad", "mire", "),", "Ġè", "IJ", "½", "ãģ", "¡", "ç", "Ŀ", "Ģ", "ãģı", "Ġ(", "be", "Ġcalm", "),", "Ġæ", "º", "Ģ", "è", "¶", "³", "Ġ(", "be", "Ġsatisfied", "),", "Ġç", "Ļ", "Ĵ", "ãģķ", "ãĤĮ", "ãĤĭ", "Ġ(", "be", "Ġhealed", "),", "Ġand", "Ġ", "ãĤ¹", "ãĥĥ", "ãĤŃ", "ãĥª", "Ġ(", "be", "Ġrefreshed", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠNegative", "ĠWords", "</s>", "æĢ", "Ĵ", "ãĤĭ", "Ġ(", "get", "Ġangry", "),", "Ġæ", "Ĥ", "²", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġsad", "),", "Ġå", "¯", "Ĥ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġlonely", "),", "Ġæ", "Ģ", "ĸ", "ãģĦ", "Ġ(", "be", "Ġscared", "),", "Ġ", "ä¸į", "å®", "ī", "Ġ(", "feel", "Ġanxious", "),", "Ġæ", "ģ", "¥", "ãģ", "ļ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġembarrassed", "),", "Ġå", "«", "Į", "Ġ(", "hate", "),", "Ġè", "IJ", "½", "ãģ", "¡", "è", "¾", "¼", "ãĤ", "Ģ", "Ġ(", "feel", "Ġdown", "),", "Ġé", "Ģ", "Ģ", "å", "±", "Ī", "Ġ(", "be", "Ġbored", "),", "Ġç", "µ", "¶", "æľ", "Ľ", "Ġ(", "feel", "Ġhopeless", "),", "Ġè", "¾", "Ľ", "ãģĦ", "Ġ(", "have", "Ġa", "Ġhard", "Ġtime", "),", "Ġå", "Ľ", "°", "ãĤĭ", "Ġ(", "have", "Ġtrouble", "),", "Ġæ", "Ĩ", "Ĥ", "é", "¬", "±", "Ġ(", "be", "Ġdepressed", "),", "Ġå", "¿", "ĥ", "é", "ħ", "į", "Ġ(", "be", "Ġworried", "),", "Ġand", "Ġæ", "ĥ", "ħ", "ãģ", "ij", "ãģª", "ãģĦ", "Ġ(", "be", "Ġsorry", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠBi", "GR", "U", "</s>", "The", "Ġdimension", "Ġof", "Ġthe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġ256", ".", "ĠThe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġinitialized", "Ġwith", "Ġthe", "Ġword", "Ġembed", "d", "ings", "Ġpret", "rained", "Ġusing", "Ġthe", "ĠWeb", "Ġcorpus", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "Ġthe", "Ġmorph", "ological", "Ġanaly", "zer", "ĠJ", "uman", "++", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ100", ",", "000", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ2", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ256", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠMoment", "um", "ĠSG", "D", "ĠB", "IB", "REF", "21", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ1024", ".", "ĠWe", "Ġran", "Ġ100", "Ġepoch", "s", "Ġand", "Ġselected", "Ġthe", "Ġsnapshot", "Ġthat", "Ġachieved", "Ġthe", "Ġhighest", "Ġscore", "Ġfor", "Ġthe", "Ġdev", "Ġset", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠB", "ERT", "</s>", "We", "Ġused", "Ġa", "ĠJapanese", "ĠB", "ERT", "Ġmodel", "Ġpret", "rained", "Ġwith", "ĠJapanese", "ĠWikipedia", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "ĠJ", "uman", "++", ",", "Ġand", "Ġwords", "Ġwere", "Ġbroken", "Ġinto", "Ġsub", "words", "Ġby", "Ġapplying", "ĠB", "PE", "ĠB", "IB", "REF", "20", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ32", ",", "000", ".", "ĠThe", "Ġmaximum", "Ġlength", "Ġof", "Ġan", "Ġinput", "Ġsequence", "Ġwas", "Ġ128", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ12", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ768", ".", "ĠThe", "Ġnumber", "Ġof", "Ġself", "-", "att", "ention", "Ġheads", "Ġwas", "Ġ12", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠAdam", "ĠB", "IB", "REF", "19", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ32", ".", "ĠWe", "Ġran", "Ġ1", "Ġepoch", "." ]
[ 8, 10, 137, 211, 543, 592, 595, 734, 870, 908, 1002, 1006, 1015, 1061, 1150, 1165, 1270, 1371, 1397, 1485, 1508, 1541, 1565, 1599, 1607, 1654, 1677, 1778, 1791, 1876, 1889, 1909, 1912, 1920, 1936, 2127, 2159, 2175, 2238, 2347, 2363, 2428, 2441, 2447, 2461, 2468, 2531, 2540, 2593, 2655, 2663, 2730, 2791, 2956, 2964, 3043, 3095, 3162, 3254, 3273, 3399, 3429, 3446, 3477, 3603, 3605, 3647, 3680, 3683, 3732, 3745, 3886, 3899, 4042, 4057, 4169, 4183 ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
{ "all_answers": [ { "text": "Using all data to train: AL -- BiGRU achieved 0.843 accuracy, AL -- BERT achieved 0.863 accuracy, AL+CA+CO -- BiGRU achieved 0.866 accuracy, AL+CA+CO -- BERT achieved 0.835, accuracy, ACP -- BiGRU achieved 0.919 accuracy, ACP -- BERT achived 0.933, accuracy, ACP+AL+CA+CO -- BiGRU achieved 0.917 accuracy, ACP+AL+CA+CO -- BERT achieved 0.913 accuracy. \nUsing a subset to train: BERT achieved 0.876 accuracy using ACP (6K), BERT achieved 0.886 accuracy using ACP (6K) + AL, BiGRU achieved 0.830 accuracy using ACP (6K), BiGRU achieved 0.879 accuracy using ACP (6K) + AL + CA + CO.", "type": "ABSTRACTIVE" } ], "all_evidence": [ [ "FLOAT SELECTED: Table 3: Performance of various models on the ACP test set.", "FLOAT SELECTED: Table 4: Results for small labeled training data. Given the performance with the full dataset, we show BERT trained only with the AL data.", "As for ${\\rm Encoder}$, we compared two types of neural networks: BiGRU and BERT. GRU BIBREF16 is a recurrent neural network sequence encoder. BiGRU reads an input sequence forward and backward and the output is the concatenation of the final forward and backward hidden states.", "We trained the model with the following four combinations of the datasets: AL, AL+CA+CO (two proposed models), ACP (supervised), and ACP+AL+CA+CO (semi-supervised). The corresponding objective functions were: $\\mathcal {L}_{\\rm AL}$, $\\mathcal {L}_{\\rm AL} + \\mathcal {L}_{\\rm CA} + \\mathcal {L}_{\\rm CO}$, $\\mathcal {L}_{\\rm ACP}$, and $\\mathcal {L}_{\\rm ACP} + \\mathcal {L}_{\\rm AL} + \\mathcal {L}_{\\rm CA} + \\mathcal {L}_{\\rm CO}$." ] ], "all_evidence_masks": [ [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ] ], "article_id": "1909.00694", "context_tokens": [ "Introduction", "</s>", "A", "ffect", "ive", "Ġevents", "ĠB", "IB", "REF", "0", "Ġare", "Ġevents", "Ġthat", "Ġtypically", "Ġaffect", "Ġpeople", "Ġin", "Ġpositive", "Ġor", "Ġnegative", "Ġways", ".", "ĠFor", "Ġexample", ",", "Ġgetting", "Ġmoney", "Ġand", "Ġplaying", "Ġsports", "Ġare", "Ġusually", "Ġpositive", "Ġto", "Ġthe", "Ġexperien", "cers", ";", "Ġcatching", "Ġcold", "Ġand", "Ġlosing", "Ġone", "'s", "Ġwallet", "Ġare", "Ġnegative", ".", "ĠUnderstanding", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġimportant", "Ġto", "Ġvarious", "Ġnatural", "Ġlanguage", "Ġprocessing", "Ġ(", "N", "LP", ")", "Ġapplications", "Ġsuch", "Ġas", "Ġdialogue", "Ġsystems", "ĠB", "IB", "REF", "1", ",", "Ġquestion", "-", "ans", "w", "ering", "Ġsystems", "ĠB", "IB", "REF", "2", ",", "Ġand", "Ġhumor", "Ġrecognition", "ĠB", "IB", "REF", "3", ".", "ĠIn", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġwork", "Ġon", "Ġrecognizing", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġaffect", "ive", "Ġevent", "Ġthat", "Ġis", "Ġrepresented", "Ġby", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġ(", "negative", ")", "Ġto", "Ġ1", "Ġ(", "positive", ").", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġchallenging", "Ġbecause", ",", "Ġas", "Ġthe", "Ġexamples", "Ġabove", "Ġsuggest", ",", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġis", "Ġnot", "Ġnecessarily", "Ġpredictable", "Ġfrom", "Ġits", "Ġconstituent", "Ġwords", ".", "ĠCombined", "Ġwith", "Ġthe", "Ġunb", "ounded", "Ġcomb", "inator", "ial", "Ġnature", "Ġof", "Ġlanguage", ",", "Ġthe", "Ġnon", "-", "com", "position", "ality", "Ġof", "Ġaffect", "ive", "Ġpol", "arity", "Ġentails", "Ġthe", "Ġneed", "Ġfor", "Ġlarge", "Ġamounts", "Ġof", "Ġworld", "Ġknowledge", ",", "Ġwhich", "Ġcan", "Ġhardly", "Ġbe", "Ġlearned", "Ġfrom", "Ġsmall", "Ġannot", "ated", "Ġdata", ".", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġpropose", "Ġa", "Ġsimple", "Ġand", "Ġeffective", "Ġmethod", "Ġfor", "Ġlearning", "Ġaffect", "ive", "Ġevents", "Ġthat", "Ġonly", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠAs", "Ġillustrated", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġour", "Ġkey", "Ġidea", "Ġis", "Ġthat", "Ġwe", "Ġcan", "Ġexploit", "Ġdiscourse", "Ġrelations", "ĠB", "IB", "REF", "4", "Ġto", "Ġefficiently", "Ġpropagate", "Ġpol", "arity", "Ġfrom", "Ġseed", "Ġpred", "icates", "Ġthat", "Ġdirectly", "Ġreport", "Ġone", "'s", "Ġemotions", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "to", "Ġbe", "Ġglad", "âĢ", "Ŀ", "Ġis", "Ġpositive", ").", "ĠSuppose", "Ġthat", "Ġevents", "Ġ$", "x", "_", "1", "$", "Ġare", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCause", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "1", "$", "Ġcauses", "Ġ$", "x", "_", "2", "$", ").", "ĠIf", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġsuggests", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġpositive", ",", "Ġ$", "x", "_", "1", "$", "Ġis", "Ġalso", "Ġlikely", "Ġto", "Ġbe", "Ġpositive", "Ġbecause", "Ġit", "Ġtriggers", "Ġthe", "Ġpositive", "Ġemotion", ".", "ĠThe", "Ġfact", "Ġthat", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġknown", "Ġto", "Ġbe", "Ġnegative", "Ġindicates", "Ġthe", "Ġnegative", "Ġpol", "arity", "Ġof", "Ġ$", "x", "_", "1", "$.", "ĠSimilarly", ",", "Ġif", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCon", "cession", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "2", "$", "Ġin", "Ġspite", "Ġof", "Ġ$", "x", "_", "1", "$", "),", "Ġthe", "Ġreverse", "Ġof", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġcan", "Ġbe", "Ġpropag", "ated", "Ġto", "Ġ$", "x", "_", "1", "$.", "ĠEven", "Ġif", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġis", "Ġnot", "Ġknown", "Ġin", "Ġadvance", ",", "Ġwe", "Ġcan", "Ġexploit", "Ġthe", "Ġtendency", "Ġof", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġto", "Ġbe", "Ġof", "Ġthe", "Ġsame", "Ġpol", "arity", "Ġ(", "for", "ĠCause", ")", "Ġor", "Ġof", "Ġthe", "Ġreverse", "Ġpol", "arity", "Ġ(", "for", "ĠCon", "cession", ")", "Ġalthough", "Ġthe", "Ġhe", "uristic", "Ġis", "Ġnot", "Ġexempt", "Ġfrom", "Ġcount", "ere", "x", "amples", ".", "ĠWe", "Ġtransform", "Ġthis", "Ġidea", "Ġinto", "Ġobjective", "Ġfunctions", "Ġand", "Ġtrain", "Ġneural", "Ġnetwork", "Ġmodels", "Ġthat", "Ġpredict", "Ġthe", "Ġpol", "arity", "Ġof", "Ġa", "Ġgiven", "Ġevent", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodels", "Ġusing", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", ".", "ĠGiven", "Ġthe", "Ġminimum", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthey", "Ġperformed", "Ġwell", ".", "ĠIn", "Ġaddition", ",", "Ġthe", "Ġcombination", "Ġof", "Ġannot", "ated", "Ġand", "Ġun", "annot", "ated", "Ġdata", "Ġyielded", "Ġa", "Ġgain", "Ġover", "Ġa", "Ġpurely", "Ġsupervised", "Ġbaseline", "Ġwhen", "Ġlabeled", "Ġdata", "Ġwere", "Ġsmall", ".", "</s>", "Related", "ĠWork", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġclosely", "Ġrelated", "Ġto", "Ġsentiment", "Ġanalysis", ".", "ĠWhereas", "Ġsentiment", "Ġanalysis", "Ġusually", "Ġfocuses", "Ġon", "Ġthe", "Ġpol", "arity", "Ġof", "Ġwhat", "Ġare", "Ġdescribed", "Ġ(", "e", ".", "g", ".,", "Ġmovies", "),", "Ġwe", "Ġwork", "Ġon", "Ġhow", "Ġpeople", "Ġare", "Ġtypically", "Ġaffected", "Ġby", "Ġevents", ".", "ĠIn", "Ġsentiment", "Ġanalysis", ",", "Ġmuch", "Ġattention", "Ġhas", "Ġbeen", "Ġpaid", "Ġto", "Ġcomposition", "ality", ".", "ĠWord", "-", "level", "Ġpol", "arity", "ĠB", "IB", "REF", "5", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "7", "Ġand", "Ġthe", "Ġroles", "Ġof", "Ġneg", "ation", "Ġand", "Ġintens", "ification", "ĠB", "IB", "REF", "8", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "9", "Ġare", "Ġamong", "Ġthe", "Ġmost", "Ġimportant", "Ġtopics", ".", "ĠIn", "Ġcontrast", ",", "Ġwe", "Ġare", "Ġmore", "Ġinterested", "Ġin", "Ġrecognizing", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġthat", "Ġper", "tains", "Ġto", "Ġcommons", "ense", "Ġknowledge", "Ġ(", "e", ".", "g", ".,", "Ġgetting", "Ġmoney", "Ġand", "Ġcatching", "Ġcold", ").", "</s>", "Label", "Ġpropagation", "Ġfrom", "Ġseed", "Ġinstances", "Ġis", "Ġa", "Ġcommon", "Ġapproach", "Ġto", "Ġinducing", "Ġsentiment", "Ġpolar", "ities", ".", "ĠWhile", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġworked", "Ġon", "Ġword", "-", "Ġand", "Ġphrase", "-", "level", "Ġpolar", "ities", ",", "ĠB", "IB", "REF", "0", "Ġdealt", "Ġwith", "Ġevent", "-", "level", "Ġpolar", "ities", ".", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġlinked", "Ġinstances", "Ġusing", "Ġco", "-", "occ", "urrence", "Ġinformation", "Ġand", "/", "or", "Ġphrase", "-", "level", "Ġcoord", "inations", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "$", "A", "$", "Ġand", "Ġ$", "B", "$", "âĢ", "Ŀ", "Ġand", "ĠâĢ", "ľ", "$", "A", "$", "Ġbut", "Ġ$", "B", "$", "âĢ", "Ŀ", ").", "ĠWe", "Ġshift", "Ġour", "Ġscope", "Ġto", "Ġevent", "Ġpairs", "Ġthat", "Ġare", "Ġmore", "Ġcomplex", "Ġthan", "Ġphrase", "Ġpairs", ",", "Ġand", "Ġconsequently", "Ġexploit", "Ġdiscourse", "Ġconnect", "ives", "Ġas", "Ġevent", "-", "level", "Ġcounterparts", "Ġof", "Ġphrase", "-", "level", "Ġconj", "unctions", ".", "</s>", "BI", "B", "REF", "0", "Ġconstructed", "Ġa", "Ġnetwork", "Ġof", "Ġevents", "Ġusing", "Ġword", "Ġembed", "ding", "-", "derived", "Ġsimilarities", ".", "ĠCompared", "Ġwith", "Ġthis", "Ġmethod", ",", "Ġour", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlinking", "Ġof", "Ġevents", "Ġis", "Ġmuch", "Ġsimpler", "Ġand", "Ġmore", "Ġintuitive", ".", "</s>", "Some", "Ġprevious", "Ġstudies", "Ġmade", "Ġuse", "Ġof", "Ġdocument", "Ġstructure", "Ġto", "Ġunderstand", "Ġthe", "Ġsentiment", ".", "ĠB", "IB", "REF", "11", "Ġproposed", "Ġa", "Ġsentiment", "-", "specific", "Ġpre", "-", "training", "Ġstrategy", "Ġusing", "Ġunl", "abel", "ed", "Ġdialog", "Ġdata", "Ġ(", "t", "weet", "-", "reply", "Ġpairs", ").", "ĠB", "IB", "REF", "12", "Ġproposed", "Ġa", "Ġmethod", "Ġof", "Ġbuilding", "Ġa", "Ġpol", "arity", "-", "tag", "ged", "Ġcorpus", "Ġ(", "ACP", "ĠCorpus", ").", "ĠThey", "Ġautomatically", "Ġgathered", "Ġsentences", "Ġthat", "Ġhad", "Ġpositive", "Ġor", "Ġnegative", "Ġopinions", "Ġutilizing", "ĠHTML", "Ġlayout", "Ġstructures", "Ġin", "Ġaddition", "Ġto", "Ġlinguistic", "Ġpatterns", ".", "ĠOur", "Ġmethod", "Ġdepends", "Ġonly", "Ġon", "Ġraw", "Ġtexts", "Ġand", "Ġthus", "Ġhas", "Ġwider", "Ġapplic", "ability", ".", "</s>", "Prop", "osed", "ĠMethod", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠPol", "arity", "ĠFunction", "</s>", "Our", "Ġgoal", "Ġis", "Ġto", "Ġlearn", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$,", "Ġwhich", "Ġpredicts", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġscore", "Ġof", "Ġan", "Ġevent", "Ġ$", "x", "$.", "ĠWe", "Ġapproximate", "Ġ$", "p", "(", "x", ")", "$", "Ġby", "Ġa", "Ġneural", "Ġnetwork", "Ġwith", "Ġthe", "Ġfollowing", "Ġform", ":", "</s>", "$", "{\\", "rm", "ĠEnc", "oder", "}", "$", "Ġoutputs", "Ġa", "Ġvector", "Ġrepresentation", "Ġof", "Ġthe", "Ġevent", "Ġ$", "x", "$.", "Ġ${", "\\", "rm", "ĠLinear", "}", "$", "Ġis", "Ġa", "Ġfully", "-", "connected", "Ġlayer", "Ġand", "Ġtransforms", "Ġthe", "Ġrepresentation", "Ġinto", "Ġa", "Ġscal", "ar", ".", "Ġ${", "\\", "rm", "Ġtan", "h", "}", "$", "Ġis", "Ġthe", "Ġhyper", "b", "olic", "Ġtang", "ent", "Ġand", "Ġtransforms", "Ġthe", "Ġscal", "ar", "Ġinto", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġto", "Ġ1", ".", "ĠIn", "ĠSection", "ĠSEC", "REF", "21", ",", "Ġwe", "Ġconsider", "Ġtwo", "Ġspecific", "Ġimplementations", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$.", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "</s>", "Our", "Ġmethod", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġassume", "Ġthat", "Ġwe", "Ġcan", "Ġautomatically", "Ġextract", "Ġdiscourse", "-", "tag", "ged", "Ġevent", "Ġpairs", ",", "Ġ$(", "x", "_{", "i", "1", "},", "Ġx", "_{", "i", "2", "})", "$", "Ġ($", "i", "=", "1", ",", "Ġ\\", "cd", "ots", "Ġ$", ")", "Ġfrom", "Ġthe", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġrefer", "Ġto", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġas", "Ġformer", "Ġand", "Ġlatter", "Ġevents", ",", "Ġrespectively", ".", "ĠAs", "Ġshown", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġwe", "Ġlimit", "Ġour", "Ġscope", "Ġto", "Ġtwo", "Ġdiscourse", "Ġrelations", ":", "ĠCause", "Ġand", "ĠCon", "cession", ".", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġconsists", "Ġof", "Ġpositive", "Ġand", "Ġnegative", "Ġpred", "icates", ".", "ĠIf", "Ġthe", "Ġpredicate", "Ġof", "Ġan", "Ġextracted", "Ġevent", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġdoes", "Ġnot", "Ġinvolve", "Ġcomplex", "Ġphenomena", "Ġlike", "Ġneg", "ation", ",", "Ġwe", "Ġassign", "Ġthe", "Ġcorresponding", "Ġpol", "arity", "Ġscore", "Ġ($", "+", "1", "$", "Ġfor", "Ġpositive", "Ġevents", "Ġand", "Ġ$", "-", "1", "$", "Ġfor", "Ġnegative", "Ġevents", ")", "Ġto", "Ġthe", "Ġevent", ".", "ĠWe", "Ġexpect", "Ġthe", "Ġmodel", "Ġto", "Ġautomatically", "Ġlearn", "Ġcomplex", "Ġphenomena", "Ġthrough", "Ġlabel", "Ġpropagation", ".", "ĠBased", "Ġon", "Ġthe", "Ġavailability", "Ġof", "Ġscores", "Ġand", "Ġthe", "Ġtypes", "Ġof", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġclassify", "Ġthe", "Ġextracted", "Ġevent", "Ġpairs", "Ġinto", "Ġthe", "Ġfollowing", "Ġthree", "Ġtypes", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠAL", "Ġ(", "Autom", "atically", "ĠLab", "eled", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġ(", "1", ")", "Ġthe", "Ġlatter", "Ġevent", "Ġbut", "Ġ(", "2", ")", "Ġnot", "Ġthe", "Ġformer", "Ġevent", ",", "Ġand", "Ġ(", "3", ")", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", "Ġor", "ĠCon", "cession", ".", "ĠIf", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġsame", "Ġscore", "Ġas", "Ġthe", "Ġlatter", ".", "ĠLikewise", ",", "Ġif", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġopposite", "Ġof", "Ġthe", "Ġlatter", "'s", "Ġscore", ".", "ĠThey", "Ġare", "Ġused", "Ġas", "Ġreference", "Ġscores", "Ġduring", "Ġtraining", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCA", "Ġ(", "Cause", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġsame", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCO", "Ġ(", "Con", "cession", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġreversed", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠLoss", "ĠFunctions", "</s>", "Using", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "Ġdata", ",", "Ġwe", "Ġoptimize", "Ġthe", "Ġparameters", "Ġof", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$.", "ĠWe", "Ġdefine", "Ġa", "Ġloss", "Ġfunction", "Ġfor", "Ġeach", "Ġof", "Ġthe", "Ġthree", "Ġtypes", "Ġof", "Ġevent", "Ġpairs", "Ġand", "Ġsum", "Ġup", "Ġthe", "Ġmultiple", "Ġloss", "Ġfunctions", ".", "</s>", "We", "Ġuse", "Ġmean", "Ġsquared", "Ġerror", "Ġto", "Ġconstruct", "Ġloss", "Ġfunctions", ".", "ĠFor", "Ġthe", "ĠAL", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "where", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠAL", "Ġdata", ".", "Ġ$", "r", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "r", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġautomatically", "-", "ass", "igned", "Ġscores", "Ġof", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$,", "Ġrespectively", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠAL", "Ġpairs", ",", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġa", "Ġhyper", "param", "eter", ".", "</s>", "For", "Ġthe", "ĠCA", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "$", "y", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "y", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠCA", "Ġpairs", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠCA", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠCA", "Ġpairs", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "mu", "Ġ$", "Ġare", "Ġhyper", "param", "eters", ".", "ĠThe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġcloser", "Ġwhile", "Ġthe", "Ġsecond", "Ġterm", "Ġprevents", "Ġthe", "Ġscores", "Ġfrom", "Ġshrinking", "Ġto", "Ġzero", ".", "</s>", "The", "Ġloss", "Ġfunction", "Ġfor", "Ġthe", "ĠCO", "Ġdata", "Ġis", "Ġdefined", "Ġanalog", "ously", ":", "</s>", "The", "Ġdifference", "Ġis", "Ġthat", "Ġthe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġdistant", "Ġfrom", "Ġeach", "Ġother", ".", "</s>", "Exper", "iments", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "</s>", "As", "Ġa", "Ġraw", "Ġcorpus", ",", "Ġwe", "Ġused", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", "Ġthat", "Ġwas", "Ġcompiled", "Ġthrough", "Ġthe", "Ġprocedures", "Ġproposed", "Ġby", "ĠB", "IB", "REF", "13", ".", "ĠTo", "Ġextract", "Ġevent", "Ġpairs", "Ġtagged", "Ġwith", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġused", "Ġthe", "ĠJapanese", "Ġdependency", "Ġparser", "ĠK", "NP", "Ġand", "Ġin", "-", "house", "Ġpost", "processing", "Ġscripts", "ĠB", "IB", "REF", "14", ".", "ĠK", "NP", "Ġused", "Ġhand", "-", "written", "Ġrules", "Ġto", "Ġsegment", "Ġeach", "Ġsentence", "Ġinto", "Ġwhat", "Ġwe", "Ġconvention", "ally", "Ġcalled", "Ġclauses", "Ġ(", "mostly", "Ġconsecutive", "Ġtext", "Ġchunks", "),", "Ġeach", "Ġof", "Ġwhich", "Ġcontained", "Ġone", "Ġmain", "Ġpredicate", ".", "ĠK", "NP", "Ġalso", "Ġidentified", "Ġthe", "Ġdiscourse", "Ġrelations", "Ġof", "Ġevent", "Ġpairs", "Ġif", "Ġexplicit", "Ġdiscourse", "Ġconnect", "ives", "ĠB", "IB", "REF", "4", "Ġsuch", "Ġas", "ĠâĢ", "ľ", "ãģ®", "ãģ§", "âĢ", "Ŀ", "Ġ(", "because", ")", "Ġand", "ĠâĢ", "ľ", "ãģ®", "ãģ«", "âĢ", "Ŀ", "Ġ(", "in", "Ġspite", "Ġof", ")", "Ġwere", "Ġpresent", ".", "ĠWe", "Ġtreated", "ĠCause", "/", "Reason", "Ġ(", "åİ", "Ł", "åĽ", "ł", "ãĥ»", "çIJ", "Ĩ", "çĶ", "±", ")", "Ġand", "ĠCondition", "Ġ(", "æĿ", "¡", "ä»", "¶", ")", "Ġin", "Ġthe", "Ġoriginal", "Ġtag", "set", "ĠB", "IB", "REF", "15", "Ġas", "ĠCause", "Ġand", "ĠCon", "cession", "Ġ(", "éĢ", "Ĩ", "æ", "İ", "¥", ")", "Ġas", "ĠCon", "cession", ",", "Ġrespectively", ".", "ĠHere", "Ġis", "Ġan", "Ġexample", "Ġof", "Ġevent", "Ġpair", "Ġextraction", ".", "</s>", ".", "Ġé", "ĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĹ", "ãģŁ", "ãģ®", "ãģ§", "ãĢģ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãģ£", "ãģŁ", "ãĢĤ", "</s>", "Because", "Ġ[", "I", "]", "Ġmade", "Ġa", "Ġserious", "Ġmistake", ",", "Ġ[", "I", "]", "Ġgot", "Ġfired", ".", "</s>", "From", "Ġthis", "Ġsentence", ",", "Ġwe", "Ġextracted", "Ġthe", "Ġevent", "Ġpair", "Ġof", "ĠâĢ", "ľ", "éĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĻ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġmake", "Ġa", "Ġserious", "Ġmistake", ")", "Ġand", "ĠâĢ", "ľ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãĤĭ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġget", "Ġfired", "),", "Ġand", "Ġtagged", "Ġit", "Ġwith", "ĠCause", ".", "</s>", "We", "Ġconstructed", "Ġour", "Ġseed", "Ġlex", "icon", "Ġconsisting", "Ġof", "Ġ15", "Ġpositive", "Ġwords", "Ġand", "Ġ15", "Ġnegative", "Ġwords", ",", "Ġas", "Ġshown", "Ġin", "ĠSection", "ĠSEC", "REF", "27", ".", "ĠFrom", "Ġthe", "Ġcorpus", "Ġof", "Ġabout", "Ġ100", "Ġmillion", "Ġsentences", ",", "Ġwe", "Ġobtained", "Ġ1", ".", "4", "Ġmillions", "Ġevent", "Ġpairs", "Ġfor", "ĠAL", ",", "Ġ41", "Ġmillions", "Ġfor", "ĠCA", ",", "Ġand", "Ġ6", "Ġmillions", "Ġfor", "ĠCO", ".", "ĠWe", "Ġrandomly", "Ġselected", "Ġsubs", "ets", "Ġof", "ĠAL", "Ġevent", "Ġpairs", "Ġsuch", "Ġthat", "Ġpositive", "Ġand", "Ġnegative", "Ġlatter", "Ġevents", "Ġwere", "Ġequal", "Ġin", "Ġsize", ".", "ĠWe", "Ġalso", "Ġsampled", "Ġevent", "Ġpairs", "Ġfor", "Ġeach", "Ġof", "ĠCA", "Ġand", "ĠCO", "Ġsuch", "Ġthat", "Ġit", "Ġwas", "Ġfive", "Ġtimes", "Ġlarger", "Ġthan", "ĠAL", ".", "ĠThe", "Ġresults", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "16", ".", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAC", "P", "Ġ(", "ACP", "ĠCorpus", ")", "</s>", "We", "Ġused", "Ġthe", "Ġlatest", "Ġversion", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", "ĠB", "IB", "REF", "12", "Ġfor", "Ġevaluation", ".", "ĠIt", "Ġwas", "Ġused", "Ġfor", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġtraining", "Ġas", "Ġwell", ".", "ĠExt", "racted", "Ġfrom", "ĠJapanese", "Ġwebsites", "Ġusing", "ĠHTML", "Ġlayouts", "Ġand", "Ġlinguistic", "Ġpatterns", ",", "Ġthe", "Ġdataset", "Ġcovered", "Ġvarious", "Ġgenres", ".", "ĠFor", "Ġexample", ",", "Ġthe", "Ġfollowing", "Ġtwo", "Ġsentences", "Ġwere", "Ġlabeled", "Ġpositive", "Ġand", "Ġnegative", ",", "Ġrespectively", ":", "</s>", ".", "Ġ", "ä½ľ", "æ", "¥", "Ń", "ãģĮ", "æ", "¥", "½", "ãģł", "ãĢĤ", "</s>", "The", "Ġwork", "Ġis", "Ġeasy", ".", "</s>", ".", "Ġé", "§", "IJ", "è»", "Ĭ", "å", "ł", "´", "ãģĮ", "ãģª", "ãģĦ", "ãĢĤ", "</s>", "There", "Ġis", "Ġno", "Ġparking", "Ġlot", ".", "</s>", "Although", "Ġthe", "ĠAC", "P", "Ġcorpus", "Ġwas", "Ġoriginally", "Ġconstructed", "Ġin", "Ġthe", "Ġcontext", "Ġof", "Ġsentiment", "Ġanalysis", ",", "Ġwe", "Ġfound", "Ġthat", "Ġit", "Ġcould", "Ġroughly", "Ġbe", "Ġregarded", "Ġas", "Ġa", "Ġcollection", "Ġof", "Ġaffect", "ive", "Ġevents", ".", "ĠWe", "Ġparsed", "Ġeach", "Ġsentence", "Ġand", "Ġextracted", "Ġthe", "Ġlast", "Ġclause", "Ġin", "Ġit", ".", "ĠThe", "Ġtrain", "/", "dev", "/", "test", "Ġsplit", "Ġof", "Ġthe", "Ġdata", "Ġis", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "19", ".", "</s>", "The", "Ġobjective", "Ġfunction", "Ġfor", "Ġsupervised", "Ġtraining", "Ġis", ":", "</s>", "where", "Ġ$", "v", "_", "i", "$", "Ġis", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġevent", ",", "Ġ$", "R", "_", "i", "$", "Ġis", "Ġthe", "Ġreference", "Ġscore", "Ġof", "Ġ$", "v", "_", "i", "$,", "Ġand", "Ġ$", "N", "_{", "\\", "rm", "ĠAC", "P", "}", "$", "Ġis", "Ġthe", "Ġnumber", "Ġof", "Ġthe", "Ġevents", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "</s>", "To", "Ġoptimize", "Ġthe", "Ġhyper", "param", "eters", ",", "Ġwe", "Ġused", "Ġthe", "Ġdev", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠFor", "Ġthe", "Ġevaluation", ",", "Ġwe", "Ġused", "Ġthe", "Ġtest", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠThe", "Ġmodel", "Ġoutput", "Ġwas", "Ġclassified", "Ġas", "Ġpositive", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ>", "Ġ0", "$", "Ġand", "Ġnegative", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ\\", "le", "Ġ0", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠModel", "ĠConfig", "urations", "</s>", "As", "Ġfor", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġwe", "Ġcompared", "Ġtwo", "Ġtypes", "Ġof", "Ġneural", "Ġnetworks", ":", "ĠBi", "GR", "U", "Ġand", "ĠB", "ERT", ".", "ĠGR", "U", "ĠB", "IB", "REF", "16", "Ġis", "Ġa", "Ġrecurrent", "Ġneural", "Ġnetwork", "Ġsequence", "Ġenc", "oder", ".", "ĠBi", "GR", "U", "Ġreads", "Ġan", "Ġinput", "Ġsequence", "Ġforward", "Ġand", "Ġbackward", "Ġand", "Ġthe", "Ġoutput", "Ġis", "Ġthe", "Ġconc", "aten", "ation", "Ġof", "Ġthe", "Ġfinal", "Ġforward", "Ġand", "Ġbackward", "Ġhidden", "Ġstates", ".", "</s>", "BER", "T", "ĠB", "IB", "REF", "17", "Ġis", "Ġa", "Ġpre", "-", "trained", "Ġmulti", "-", "layer", "Ġbid", "irection", "al", "ĠTrans", "former", "ĠB", "IB", "REF", "18", "Ġenc", "oder", ".", "ĠIts", "Ġoutput", "Ġis", "Ġthe", "Ġfinal", "Ġhidden", "Ġstate", "Ġcorresponding", "Ġto", "Ġthe", "Ġspecial", "Ġclassification", "Ġtag", "Ġ([", "CL", "S", "]).", "ĠFor", "Ġthe", "Ġdetails", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġsee", "ĠSections", "ĠSEC", "REF", "30", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodel", "Ġwith", "Ġthe", "Ġfollowing", "Ġfour", "Ġcombinations", "Ġof", "Ġthe", "Ġdatasets", ":", "ĠAL", ",", "ĠAL", "+", "CA", "+", "CO", "Ġ(", "two", "Ġproposed", "Ġmodels", "),", "ĠAC", "P", "Ġ(", "super", "vised", "),", "Ġand", "ĠAC", "P", "+", "AL", "+", "CA", "+", "CO", "Ġ(", "se", "mi", "-", "super", "vised", ").", "ĠThe", "Ġcorresponding", "Ġobjective", "Ġfunctions", "Ġwere", ":", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "$,", "Ġand", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠResults", "Ġand", "ĠDiscussion", "</s>", "Table", "ĠT", "AB", "REF", "23", "Ġshows", "Ġaccuracy", ".", "ĠAs", "Ġthe", "ĠRandom", "Ġbaseline", "Ġsuggests", ",", "Ġpositive", "Ġand", "Ġnegative", "Ġlabels", "Ġwere", "Ġdistributed", "Ġevenly", ".", "ĠThe", "ĠRandom", "+", "S", "eed", "Ġbaseline", "Ġmade", "Ġuse", "Ġof", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġoutput", "Ġthe", "Ġcorresponding", "Ġlabel", "Ġ(", "or", "Ġthe", "Ġreverse", "Ġof", "Ġit", "Ġfor", "Ġneg", "ation", ")", "Ġif", "Ġthe", "Ġevent", "'s", "Ġpredicate", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", ".", "ĠWe", "Ġcan", "Ġsee", "Ġthat", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġitself", "Ġhad", "Ġpractically", "Ġno", "Ġimpact", "Ġon", "Ġprediction", ".", "</s>", "The", "Ġmodels", "Ġin", "Ġthe", "Ġtop", "Ġblock", "Ġperformed", "Ġconsiderably", "Ġbetter", "Ġthan", "Ġthe", "Ġrandom", "Ġbas", "elines", ".", "ĠThe", "Ġperformance", "Ġgaps", "Ġwith", "Ġtheir", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġcounterparts", ",", "Ġshown", "Ġin", "Ġthe", "Ġmiddle", "Ġblock", ",", "Ġwere", "Ġless", "Ġthan", "Ġ7", "%.", "ĠThis", "Ġdemonstrates", "Ġthe", "Ġeffectiveness", "Ġof", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlabel", "Ġpropagation", ".", "</s>", "Comp", "aring", "Ġthe", "Ġmodel", "Ġvariants", ",", "Ġwe", "Ġobtained", "Ġthe", "Ġhighest", "Ġscore", "Ġwith", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġtrained", "Ġwith", "Ġthe", "ĠAL", "+", "CA", "+", "CO", "Ġdataset", ".", "ĠB", "ERT", "Ġwas", "Ġcompetitive", "Ġbut", "Ġits", "Ġperformance", "Ġwent", "Ġdown", "Ġif", "ĠCA", "Ġand", "ĠCO", "Ġwere", "Ġused", "Ġin", "Ġaddition", "Ġto", "ĠAL", ".", "ĠWe", "Ġconjecture", "Ġthat", "ĠB", "ERT", "Ġwas", "Ġmore", "Ġsensitive", "Ġto", "Ġnoises", "Ġfound", "Ġmore", "Ġfrequently", "Ġin", "ĠCA", "Ġand", "ĠCO", ".", "</s>", "Cont", "rary", "Ġto", "Ġour", "Ġexpectations", ",", "Ġsupervised", "Ġmodels", "Ġ(", "ACP", ")", "Ġoutper", "formed", "Ġsemi", "-", "super", "vised", "Ġmodels", "Ġ(", "ACP", "+", "AL", "+", "CA", "+", "CO", ").", "ĠThis", "Ġsuggests", "Ġthat", "Ġthe", "Ġtraining", "Ġset", "Ġof", "Ġ0", ".", "6", "Ġmillion", "Ġevents", "Ġis", "Ġsufficiently", "Ġlarge", "Ġfor", "Ġtraining", "Ġthe", "Ġmodels", ".", "ĠFor", "Ġcomparison", ",", "Ġwe", "Ġtrained", "Ġthe", "Ġmodels", "Ġwith", "Ġa", "Ġsubset", "Ġ(", "6", ",", "000", "Ġevents", ")", "Ġof", "Ġthe", "ĠAC", "P", "Ġdataset", ".", "ĠAs", "Ġthe", "Ġresults", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "24", "Ġdemonstrate", ",", "Ġour", "Ġmethod", "Ġis", "Ġeffective", "Ġwhen", "Ġlabeled", "Ġdata", "Ġare", "Ġsmall", ".", "</s>", "The", "Ġresult", "Ġof", "Ġhyper", "param", "eter", "Ġoptimization", "Ġfor", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġwas", "Ġas", "Ġfollows", ":", "</s>", "As", "Ġthe", "ĠCA", "Ġand", "ĠCO", "Ġpairs", "Ġwere", "Ġequal", "Ġin", "Ġsize", "Ġ(", "Table", "ĠT", "AB", "REF", "16", "),", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$", "Ġwere", "Ġcomparable", "Ġvalues", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġwas", "Ġabout", "Ġone", "-", "third", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$,", "Ġand", "Ġthis", "Ġindicated", "Ġthat", "Ġthe", "ĠCA", "Ġpairs", "Ġwere", "Ġno", "is", "ier", "Ġthan", "Ġthe", "ĠCO", "Ġpairs", ".", "ĠA", "Ġmajor", "Ġtype", "Ġof", "ĠCA", "Ġpairs", "Ġthat", "Ġviolates", "Ġour", "Ġassumption", "Ġwas", "Ġin", "Ġthe", "Ġform", "Ġof", "ĠâĢ", "ľ", "$", "\\", "text", "it", "Ġ{", "problem", "}", "_{", "\\", "text", "{", "negative", "}}", "$", "Ġcauses", "Ġ$\\", "text", "it", "Ġ{", "s", "olution", "}", "_{", "\\", "text", "{", "positive", "}}", "$", "âĢ", "Ŀ", ":", "</s>", ".", "Ġ(", "æ", "Ĥª", "ãģĦ", "ãģ¨", "ãģĵ", "ãĤ", "į", "ãģĮ", "ãģĤ", "ãĤĭ", ",", "ĠãĤ", "Ī", "ãģı", "ãģª", "ãĤĭ", "ãĤ", "Ī", "ãģĨ", "ãģ«", "åĬ", "ª", "åĬ", "Ľ", "ãģĻ", "ãĤĭ", ")", "</s>", "(", "there", "Ġis", "Ġa", "Ġbad", "Ġpoint", ",", "Ġ[", "I", "]", "Ġtry", "Ġto", "Ġimprove", "Ġ[", "it", "])", "</s>", "The", "Ġpolar", "ities", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġwere", "Ġreversed", "Ġin", "Ġspite", "Ġof", "Ġthe", "ĠCause", "Ġrelation", ",", "Ġand", "Ġthis", "Ġlowered", "Ġthe", "Ġvalue", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$.", "</s>", "Some", "Ġexamples", "Ġof", "Ġmodel", "Ġoutputs", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "26", ".", "ĠThe", "Ġfirst", "Ġtwo", "Ġexamples", "Ġsuggest", "Ġthat", "Ġour", "Ġmodel", "Ġsuccessfully", "Ġlearned", "Ġneg", "ation", "Ġwithout", "Ġexplicit", "Ġsupervision", ".", "ĠSimilarly", ",", "Ġthe", "Ġnext", "Ġtwo", "Ġexamples", "Ġdiffer", "Ġonly", "Ġin", "Ġvoice", "Ġbut", "Ġthe", "Ġmodel", "Ġcorrectly", "Ġrecognized", "Ġthat", "Ġthey", "Ġhad", "Ġopposite", "Ġpolar", "ities", ".", "ĠThe", "Ġlast", "Ġtwo", "Ġexamples", "Ġshare", "Ġthe", "Ġpredicate", "ĠâĢ", "ľ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "drop", ")", "Ġand", "Ġonly", "Ġthe", "Ġobjects", "Ġare", "Ġdifferent", ".", "ĠThe", "Ġsecond", "Ġevent", "ĠâĢ", "ľ", "è", "Ĥ", "©", "ãĤĴ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "lit", ".", "Ġdrop", "Ġone", "'s", "Ġshoulders", ")", "Ġis", "Ġan", "Ġid", "iom", "Ġthat", "Ġexpresses", "Ġa", "Ġdisappointed", "Ġfeeling", ".", "ĠThe", "Ġexamples", "Ġdemonstrate", "Ġthat", "Ġour", "Ġmodel", "Ġcorrectly", "Ġlearned", "Ġnon", "-", "com", "pos", "itional", "Ġexpressions", ".", "</s>", "Conclusion", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġproposed", "Ġto", "Ġuse", "Ġdiscourse", "Ġrelations", "Ġto", "Ġeffectively", "Ġpropagate", "Ġpolar", "ities", "Ġof", "Ġaffect", "ive", "Ġevents", "Ġfrom", "Ġseeds", ".", "ĠExper", "iments", "Ġshow", "Ġthat", ",", "Ġeven", "Ġwith", "Ġa", "Ġminimal", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthe", "Ġproposed", "Ġmethod", "Ġperformed", "Ġwell", ".", "</s>", "Although", "Ġevent", "Ġpairs", "Ġlinked", "Ġby", "Ġdiscourse", "Ġanalysis", "Ġare", "Ġshown", "Ġto", "Ġbe", "Ġuseful", ",", "Ġthey", "Ġnevertheless", "Ġcontain", "Ġnoises", ".", "ĠAdding", "Ġlingu", "istically", "-", "mot", "ivated", "Ġfiltering", "Ġrules", "Ġwould", "Ġhelp", "Ġimprove", "Ġthe", "Ġperformance", ".", "</s>", "Acknowled", "gments", "</s>", "We", "Ġthank", "ĠNob", "u", "hiro", "ĠK", "aji", "Ġfor", "Ġproviding", "Ġthe", "ĠAC", "P", "ĠCorpus", "Ġand", "ĠHiro", "k", "az", "u", "ĠK", "iy", "om", "aru", "Ġand", "ĠY", "ud", "ai", "ĠK", "ish", "imoto", "Ġfor", "Ġtheir", "Ġhelp", "Ġin", "Ġextracting", "Ġevent", "Ġpairs", ".", "ĠThis", "Ġwork", "Ġwas", "Ġpartially", "Ġsupported", "Ġby", "ĠYahoo", "!", "ĠJapan", "ĠCorporation", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠPositive", "ĠWords", "</s>", "å", "ĸ", "ľ", "ãģ", "¶", "Ġ(", "re", "joice", "),", "Ġå", "¬", "ī", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġglad", "),", "Ġæ", "¥", "½", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġpleasant", "),", "Ġå", "¹", "¸", "ãģ", "Ľ", "Ġ(", "be", "Ġhappy", "),", "Ġæ", "Ħ", "Ł", "åĭ", "ķ", "Ġ(", "be", "Ġimpressed", "),", "Ġè", "Ī", "Ī", "å¥", "®", "Ġ(", "be", "Ġexcited", "),", "Ġæ", "ĩ", "IJ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "feel", "Ġnostalgic", "),", "Ġå", "¥", "½", "ãģį", "Ġ(", "like", "),", "Ġå", "°", "Ĭ", "æķ", "¬", "Ġ(", "respect", "),", "Ġå", "®", "ī", "å¿", "ĥ", "Ġ(", "be", "Ġrelieved", "),", "Ġæ", "Ħ", "Ł", "å¿", "ĥ", "Ġ(", "ad", "mire", "),", "Ġè", "IJ", "½", "ãģ", "¡", "ç", "Ŀ", "Ģ", "ãģı", "Ġ(", "be", "Ġcalm", "),", "Ġæ", "º", "Ģ", "è", "¶", "³", "Ġ(", "be", "Ġsatisfied", "),", "Ġç", "Ļ", "Ĵ", "ãģķ", "ãĤĮ", "ãĤĭ", "Ġ(", "be", "Ġhealed", "),", "Ġand", "Ġ", "ãĤ¹", "ãĥĥ", "ãĤŃ", "ãĥª", "Ġ(", "be", "Ġrefreshed", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠNegative", "ĠWords", "</s>", "æĢ", "Ĵ", "ãĤĭ", "Ġ(", "get", "Ġangry", "),", "Ġæ", "Ĥ", "²", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġsad", "),", "Ġå", "¯", "Ĥ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġlonely", "),", "Ġæ", "Ģ", "ĸ", "ãģĦ", "Ġ(", "be", "Ġscared", "),", "Ġ", "ä¸į", "å®", "ī", "Ġ(", "feel", "Ġanxious", "),", "Ġæ", "ģ", "¥", "ãģ", "ļ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġembarrassed", "),", "Ġå", "«", "Į", "Ġ(", "hate", "),", "Ġè", "IJ", "½", "ãģ", "¡", "è", "¾", "¼", "ãĤ", "Ģ", "Ġ(", "feel", "Ġdown", "),", "Ġé", "Ģ", "Ģ", "å", "±", "Ī", "Ġ(", "be", "Ġbored", "),", "Ġç", "µ", "¶", "æľ", "Ľ", "Ġ(", "feel", "Ġhopeless", "),", "Ġè", "¾", "Ľ", "ãģĦ", "Ġ(", "have", "Ġa", "Ġhard", "Ġtime", "),", "Ġå", "Ľ", "°", "ãĤĭ", "Ġ(", "have", "Ġtrouble", "),", "Ġæ", "Ĩ", "Ĥ", "é", "¬", "±", "Ġ(", "be", "Ġdepressed", "),", "Ġå", "¿", "ĥ", "é", "ħ", "į", "Ġ(", "be", "Ġworried", "),", "Ġand", "Ġæ", "ĥ", "ħ", "ãģ", "ij", "ãģª", "ãģĦ", "Ġ(", "be", "Ġsorry", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠBi", "GR", "U", "</s>", "The", "Ġdimension", "Ġof", "Ġthe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġ256", ".", "ĠThe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġinitialized", "Ġwith", "Ġthe", "Ġword", "Ġembed", "d", "ings", "Ġpret", "rained", "Ġusing", "Ġthe", "ĠWeb", "Ġcorpus", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "Ġthe", "Ġmorph", "ological", "Ġanaly", "zer", "ĠJ", "uman", "++", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ100", ",", "000", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ2", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ256", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠMoment", "um", "ĠSG", "D", "ĠB", "IB", "REF", "21", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ1024", ".", "ĠWe", "Ġran", "Ġ100", "Ġepoch", "s", "Ġand", "Ġselected", "Ġthe", "Ġsnapshot", "Ġthat", "Ġachieved", "Ġthe", "Ġhighest", "Ġscore", "Ġfor", "Ġthe", "Ġdev", "Ġset", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠB", "ERT", "</s>", "We", "Ġused", "Ġa", "ĠJapanese", "ĠB", "ERT", "Ġmodel", "Ġpret", "rained", "Ġwith", "ĠJapanese", "ĠWikipedia", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "ĠJ", "uman", "++", ",", "Ġand", "Ġwords", "Ġwere", "Ġbroken", "Ġinto", "Ġsub", "words", "Ġby", "Ġapplying", "ĠB", "PE", "ĠB", "IB", "REF", "20", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ32", ",", "000", ".", "ĠThe", "Ġmaximum", "Ġlength", "Ġof", "Ġan", "Ġinput", "Ġsequence", "Ġwas", "Ġ128", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ12", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ768", ".", "ĠThe", "Ġnumber", "Ġof", "Ġself", "-", "att", "ention", "Ġheads", "Ġwas", "Ġ12", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠAdam", "ĠB", "IB", "REF", "19", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ32", ".", "ĠWe", "Ġran", "Ġ1", "Ġepoch", "." ], "paragraphs": [ "Introduction", "Affective events BIBREF0 are events that typically affect people in positive or negative ways. For example, getting money and playing sports are usually positive to the experiencers; catching cold and losing one's wallet are negative. Understanding affective events is important to various natural language processing (NLP) applications such as dialogue systems BIBREF1, question-answering systems BIBREF2, and humor recognition BIBREF3. In this paper, we work on recognizing the polarity of an affective event that is represented by a score ranging from $-1$ (negative) to 1 (positive).", "Learning affective events is challenging because, as the examples above suggest, the polarity of an event is not necessarily predictable from its constituent words. Combined with the unbounded combinatorial nature of language, the non-compositionality of affective polarity entails the need for large amounts of world knowledge, which can hardly be learned from small annotated data.", "In this paper, we propose a simple and effective method for learning affective events that only requires a very small seed lexicon and a large raw corpus. As illustrated in Figure FIGREF1, our key idea is that we can exploit discourse relations BIBREF4 to efficiently propagate polarity from seed predicates that directly report one's emotions (e.g., “to be glad” is positive). Suppose that events $x_1$ are $x_2$ are in the discourse relation of Cause (i.e., $x_1$ causes $x_2$). If the seed lexicon suggests $x_2$ is positive, $x_1$ is also likely to be positive because it triggers the positive emotion. The fact that $x_2$ is known to be negative indicates the negative polarity of $x_1$. Similarly, if $x_1$ and $x_2$ are in the discourse relation of Concession (i.e., $x_2$ in spite of $x_1$), the reverse of $x_2$'s polarity can be propagated to $x_1$. Even if $x_2$'s polarity is not known in advance, we can exploit the tendency of $x_1$ and $x_2$ to be of the same polarity (for Cause) or of the reverse polarity (for Concession) although the heuristic is not exempt from counterexamples. We transform this idea into objective functions and train neural network models that predict the polarity of a given event.", "We trained the models using a Japanese web corpus. Given the minimum amount of supervision, they performed well. In addition, the combination of annotated and unannotated data yielded a gain over a purely supervised baseline when labeled data were small.", "Related Work", "Learning affective events is closely related to sentiment analysis. Whereas sentiment analysis usually focuses on the polarity of what are described (e.g., movies), we work on how people are typically affected by events. In sentiment analysis, much attention has been paid to compositionality. Word-level polarity BIBREF5, BIBREF6, BIBREF7 and the roles of negation and intensification BIBREF8, BIBREF6, BIBREF9 are among the most important topics. In contrast, we are more interested in recognizing the sentiment polarity of an event that pertains to commonsense knowledge (e.g., getting money and catching cold).", "Label propagation from seed instances is a common approach to inducing sentiment polarities. While BIBREF5 and BIBREF10 worked on word- and phrase-level polarities, BIBREF0 dealt with event-level polarities. BIBREF5 and BIBREF10 linked instances using co-occurrence information and/or phrase-level coordinations (e.g., “$A$ and $B$” and “$A$ but $B$”). We shift our scope to event pairs that are more complex than phrase pairs, and consequently exploit discourse connectives as event-level counterparts of phrase-level conjunctions.", "BIBREF0 constructed a network of events using word embedding-derived similarities. Compared with this method, our discourse relation-based linking of events is much simpler and more intuitive.", "Some previous studies made use of document structure to understand the sentiment. BIBREF11 proposed a sentiment-specific pre-training strategy using unlabeled dialog data (tweet-reply pairs). BIBREF12 proposed a method of building a polarity-tagged corpus (ACP Corpus). They automatically gathered sentences that had positive or negative opinions utilizing HTML layout structures in addition to linguistic patterns. Our method depends only on raw texts and thus has wider applicability.", "Proposed Method", "Proposed Method ::: Polarity Function", "Our goal is to learn the polarity function $p(x)$, which predicts the sentiment polarity score of an event $x$. We approximate $p(x)$ by a neural network with the following form:", "${\\rm Encoder}$ outputs a vector representation of the event $x$. ${\\rm Linear}$ is a fully-connected layer and transforms the representation into a scalar. ${\\rm tanh}$ is the hyperbolic tangent and transforms the scalar into a score ranging from $-1$ to 1. In Section SECREF21, we consider two specific implementations of ${\\rm Encoder}$.", "Proposed Method ::: Discourse Relation-Based Event Pairs", "Our method requires a very small seed lexicon and a large raw corpus. We assume that we can automatically extract discourse-tagged event pairs, $(x_{i1}, x_{i2})$ ($i=1, \\cdots $) from the raw corpus. We refer to $x_{i1}$ and $x_{i2}$ as former and latter events, respectively. As shown in Figure FIGREF1, we limit our scope to two discourse relations: Cause and Concession.", "The seed lexicon consists of positive and negative predicates. If the predicate of an extracted event is in the seed lexicon and does not involve complex phenomena like negation, we assign the corresponding polarity score ($+1$ for positive events and $-1$ for negative events) to the event. We expect the model to automatically learn complex phenomena through label propagation. Based on the availability of scores and the types of discourse relations, we classify the extracted event pairs into the following three types.", "Proposed Method ::: Discourse Relation-Based Event Pairs ::: AL (Automatically Labeled Pairs)", "The seed lexicon matches (1) the latter event but (2) not the former event, and (3) their discourse relation type is Cause or Concession. If the discourse relation type is Cause, the former event is given the same score as the latter. Likewise, if the discourse relation type is Concession, the former event is given the opposite of the latter's score. They are used as reference scores during training.", "Proposed Method ::: Discourse Relation-Based Event Pairs ::: CA (Cause Pairs)", "The seed lexicon matches neither the former nor the latter event, and their discourse relation type is Cause. We assume the two events have the same polarities.", "Proposed Method ::: Discourse Relation-Based Event Pairs ::: CO (Concession Pairs)", "The seed lexicon matches neither the former nor the latter event, and their discourse relation type is Concession. We assume the two events have the reversed polarities.", "Proposed Method ::: Loss Functions", "Using AL, CA, and CO data, we optimize the parameters of the polarity function $p(x)$. We define a loss function for each of the three types of event pairs and sum up the multiple loss functions.", "We use mean squared error to construct loss functions. For the AL data, the loss function is defined as:", "where $x_{i1}$ and $x_{i2}$ are the $i$-th pair of the AL data. $r_{i1}$ and $r_{i2}$ are the automatically-assigned scores of $x_{i1}$ and $x_{i2}$, respectively. $N_{\\rm AL}$ is the total number of AL pairs, and $\\lambda _{\\rm AL}$ is a hyperparameter.", "For the CA data, the loss function is defined as:", "$y_{i1}$ and $y_{i2}$ are the $i$-th pair of the CA pairs. $N_{\\rm CA}$ is the total number of CA pairs. $\\lambda _{\\rm CA}$ and $\\mu $ are hyperparameters. The first term makes the scores of the two events closer while the second term prevents the scores from shrinking to zero.", "The loss function for the CO data is defined analogously:", "The difference is that the first term makes the scores of the two events distant from each other.", "Experiments", "Experiments ::: Dataset", "Experiments ::: Dataset ::: AL, CA, and CO", "As a raw corpus, we used a Japanese web corpus that was compiled through the procedures proposed by BIBREF13. To extract event pairs tagged with discourse relations, we used the Japanese dependency parser KNP and in-house postprocessing scripts BIBREF14. KNP used hand-written rules to segment each sentence into what we conventionally called clauses (mostly consecutive text chunks), each of which contained one main predicate. KNP also identified the discourse relations of event pairs if explicit discourse connectives BIBREF4 such as “ので” (because) and “のに” (in spite of) were present. We treated Cause/Reason (原因・理由) and Condition (条件) in the original tagset BIBREF15 as Cause and Concession (逆接) as Concession, respectively. Here is an example of event pair extraction.", ". 重大な失敗を犯したので、仕事をクビになった。", "Because [I] made a serious mistake, [I] got fired.", "From this sentence, we extracted the event pair of “重大な失敗を犯す” ([I] make a serious mistake) and “仕事をクビになる” ([I] get fired), and tagged it with Cause.", "We constructed our seed lexicon consisting of 15 positive words and 15 negative words, as shown in Section SECREF27. From the corpus of about 100 million sentences, we obtained 1.4 millions event pairs for AL, 41 millions for CA, and 6 millions for CO. We randomly selected subsets of AL event pairs such that positive and negative latter events were equal in size. We also sampled event pairs for each of CA and CO such that it was five times larger than AL. The results are shown in Table TABREF16.", "Experiments ::: Dataset ::: ACP (ACP Corpus)", "We used the latest version of the ACP Corpus BIBREF12 for evaluation. It was used for (semi-)supervised training as well. Extracted from Japanese websites using HTML layouts and linguistic patterns, the dataset covered various genres. For example, the following two sentences were labeled positive and negative, respectively:", ". 作業が楽だ。", "The work is easy.", ". 駐車場がない。", "There is no parking lot.", "Although the ACP corpus was originally constructed in the context of sentiment analysis, we found that it could roughly be regarded as a collection of affective events. We parsed each sentence and extracted the last clause in it. The train/dev/test split of the data is shown in Table TABREF19.", "The objective function for supervised training is:", "where $v_i$ is the $i$-th event, $R_i$ is the reference score of $v_i$, and $N_{\\rm ACP}$ is the number of the events of the ACP Corpus.", "To optimize the hyperparameters, we used the dev set of the ACP Corpus. For the evaluation, we used the test set of the ACP Corpus. The model output was classified as positive if $p(x) > 0$ and negative if $p(x) \\le 0$.", "Experiments ::: Model Configurations", "As for ${\\rm Encoder}$, we compared two types of neural networks: BiGRU and BERT. GRU BIBREF16 is a recurrent neural network sequence encoder. BiGRU reads an input sequence forward and backward and the output is the concatenation of the final forward and backward hidden states.", "BERT BIBREF17 is a pre-trained multi-layer bidirectional Transformer BIBREF18 encoder. Its output is the final hidden state corresponding to the special classification tag ([CLS]). For the details of ${\\rm Encoder}$, see Sections SECREF30.", "We trained the model with the following four combinations of the datasets: AL, AL+CA+CO (two proposed models), ACP (supervised), and ACP+AL+CA+CO (semi-supervised). The corresponding objective functions were: $\\mathcal {L}_{\\rm AL}$, $\\mathcal {L}_{\\rm AL} + \\mathcal {L}_{\\rm CA} + \\mathcal {L}_{\\rm CO}$, $\\mathcal {L}_{\\rm ACP}$, and $\\mathcal {L}_{\\rm ACP} + \\mathcal {L}_{\\rm AL} + \\mathcal {L}_{\\rm CA} + \\mathcal {L}_{\\rm CO}$.", "Experiments ::: Results and Discussion", "Table TABREF23 shows accuracy. As the Random baseline suggests, positive and negative labels were distributed evenly. The Random+Seed baseline made use of the seed lexicon and output the corresponding label (or the reverse of it for negation) if the event's predicate is in the seed lexicon. We can see that the seed lexicon itself had practically no impact on prediction.", "The models in the top block performed considerably better than the random baselines. The performance gaps with their (semi-)supervised counterparts, shown in the middle block, were less than 7%. This demonstrates the effectiveness of discourse relation-based label propagation.", "Comparing the model variants, we obtained the highest score with the BiGRU encoder trained with the AL+CA+CO dataset. BERT was competitive but its performance went down if CA and CO were used in addition to AL. We conjecture that BERT was more sensitive to noises found more frequently in CA and CO.", "Contrary to our expectations, supervised models (ACP) outperformed semi-supervised models (ACP+AL+CA+CO). This suggests that the training set of 0.6 million events is sufficiently large for training the models. For comparison, we trained the models with a subset (6,000 events) of the ACP dataset. As the results shown in Table TABREF24 demonstrate, our method is effective when labeled data are small.", "The result of hyperparameter optimization for the BiGRU encoder was as follows:", "As the CA and CO pairs were equal in size (Table TABREF16), $\\lambda _{\\rm CA}$ and $\\lambda _{\\rm CO}$ were comparable values. $\\lambda _{\\rm CA}$ was about one-third of $\\lambda _{\\rm CO}$, and this indicated that the CA pairs were noisier than the CO pairs. A major type of CA pairs that violates our assumption was in the form of “$\\textit {problem}_{\\text{negative}}$ causes $\\textit {solution}_{\\text{positive}}$”:", ". (悪いところがある, よくなるように努力する)", "(there is a bad point, [I] try to improve [it])", "The polarities of the two events were reversed in spite of the Cause relation, and this lowered the value of $\\lambda _{\\rm CA}$.", "Some examples of model outputs are shown in Table TABREF26. The first two examples suggest that our model successfully learned negation without explicit supervision. Similarly, the next two examples differ only in voice but the model correctly recognized that they had opposite polarities. The last two examples share the predicate “落とす\" (drop) and only the objects are different. The second event “肩を落とす\" (lit. drop one's shoulders) is an idiom that expresses a disappointed feeling. The examples demonstrate that our model correctly learned non-compositional expressions.", "Conclusion", "In this paper, we proposed to use discourse relations to effectively propagate polarities of affective events from seeds. Experiments show that, even with a minimal amount of supervision, the proposed method performed well.", "Although event pairs linked by discourse analysis are shown to be useful, they nevertheless contain noises. Adding linguistically-motivated filtering rules would help improve the performance.", "Acknowledgments", "We thank Nobuhiro Kaji for providing the ACP Corpus and Hirokazu Kiyomaru and Yudai Kishimoto for their help in extracting event pairs. This work was partially supported by Yahoo! Japan Corporation.", "Appendices ::: Seed Lexicon ::: Positive Words", "喜ぶ (rejoice), 嬉しい (be glad), 楽しい (be pleasant), 幸せ (be happy), 感動 (be impressed), 興奮 (be excited), 懐かしい (feel nostalgic), 好き (like), 尊敬 (respect), 安心 (be relieved), 感心 (admire), 落ち着く (be calm), 満足 (be satisfied), 癒される (be healed), and スッキリ (be refreshed).", "Appendices ::: Seed Lexicon ::: Negative Words", "怒る (get angry), 悲しい (be sad), 寂しい (be lonely), 怖い (be scared), 不安 (feel anxious), 恥ずかしい (be embarrassed), 嫌 (hate), 落ち込む (feel down), 退屈 (be bored), 絶望 (feel hopeless), 辛い (have a hard time), 困る (have trouble), 憂鬱 (be depressed), 心配 (be worried), and 情けない (be sorry).", "Appendices ::: Settings of Encoder ::: BiGRU", "The dimension of the embedding layer was 256. The embedding layer was initialized with the word embeddings pretrained using the Web corpus. The input sentences were segmented into words by the morphological analyzer Juman++. The vocabulary size was 100,000. The number of hidden layers was 2. The dimension of hidden units was 256. The optimizer was Momentum SGD BIBREF21. The mini-batch size was 1024. We ran 100 epochs and selected the snapshot that achieved the highest score for the dev set.", "Appendices ::: Settings of Encoder ::: BERT", "We used a Japanese BERT model pretrained with Japanese Wikipedia. The input sentences were segmented into words by Juman++, and words were broken into subwords by applying BPE BIBREF20. The vocabulary size was 32,000. The maximum length of an input sequence was 128. The number of hidden layers was 12. The dimension of hidden units was 768. The number of self-attention heads was 12. The optimizer was Adam BIBREF19. The mini-batch size was 32. We ran 1 epoch." ], "question": "What are the results?", "question_id": "9d578ddccc27dd849244d632dd0f6bf27348ad81", "question_tokens": [ "What", "Ġare", "Ġthe", "Ġresults", "?" ] }
[ "Using", "Ġall", "Ġdata", "Ġto", "Ġtrain", ":", "ĠAL", "Ġ--", "ĠBi", "GR", "U", "Ġachieved", "Ġ0", ".", "8", "43", "Ġaccuracy", ",", "ĠAL", "Ġ--", "ĠB", "ERT", "Ġachieved", "Ġ0", ".", "86", "3", "Ġaccuracy", ",", "ĠAL", "+", "CA", "+", "CO", "Ġ--", "ĠBi", "GR", "U", "Ġachieved", "Ġ0", ".", "866", "Ġaccuracy", ",", "ĠAL", "+", "CA", "+", "CO", "Ġ--", "ĠB", "ERT", "Ġachieved", "Ġ0", ".", "8", "35", ",", "Ġaccuracy", ",", "ĠAC", "P", "Ġ--", "ĠBi", "GR", "U", "Ġachieved", "Ġ0", ".", "9", "19", "Ġaccuracy", ",", "ĠAC", "P", "Ġ--", "ĠB", "ERT", "Ġa", "ch", "ived", "Ġ0", ".", "9", "33", ",", "Ġaccuracy", ",", "ĠAC", "P", "+", "AL", "+", "CA", "+", "CO", "Ġ--", "ĠBi", "GR", "U", "Ġachieved", "Ġ0", ".", "9", "17", "Ġaccuracy", ",", "ĠAC", "P", "+", "AL", "+", "CA", "+", "CO", "Ġ--", "ĠB", "ERT", "Ġachieved", "Ġ0", ".", "9", "13", "Ġaccuracy", ".", "Ġ", "Ċ", "Using", "Ġa", "Ġsubset", "Ġto", "Ġtrain", ":", "ĠB", "ERT", "Ġachieved", "Ġ0", ".", "8", "76", "Ġaccuracy", "Ġusing", "ĠAC", "P", "Ġ(", "6", "K", "),", "ĠB", "ERT", "Ġachieved", "Ġ0", ".", "886", "Ġaccuracy", "Ġusing", "ĠAC", "P", "Ġ(", "6", "K", ")", "Ġ+", "ĠAL", ",", "ĠBi", "GR", "U", "Ġachieved", "Ġ0", ".", "830", "Ġaccuracy", "Ġusing", "ĠAC", "P", "Ġ(", "6", "K", "),", "ĠBi", "GR", "U", "Ġachieved", "Ġ0", ".", "8", "79", "Ġaccuracy", "Ġusing", "ĠAC", "P", "Ġ(", "6", "K", ")", "Ġ+", "ĠAL", "Ġ+", "ĠCA", "Ġ+", "ĠCO", "." ]
[ "<s>", "What", "Ġare", "Ġthe", "Ġresults", "?", "</s>", "Introduction", "</s>", "A", "ffect", "ive", "Ġevents", "ĠB", "IB", "REF", "0", "Ġare", "Ġevents", "Ġthat", "Ġtypically", "Ġaffect", "Ġpeople", "Ġin", "Ġpositive", "Ġor", "Ġnegative", "Ġways", ".", "ĠFor", "Ġexample", ",", "Ġgetting", "Ġmoney", "Ġand", "Ġplaying", "Ġsports", "Ġare", "Ġusually", "Ġpositive", "Ġto", "Ġthe", "Ġexperien", "cers", ";", "Ġcatching", "Ġcold", "Ġand", "Ġlosing", "Ġone", "'s", "Ġwallet", "Ġare", "Ġnegative", ".", "ĠUnderstanding", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġimportant", "Ġto", "Ġvarious", "Ġnatural", "Ġlanguage", "Ġprocessing", "Ġ(", "N", "LP", ")", "Ġapplications", "Ġsuch", "Ġas", "Ġdialogue", "Ġsystems", "ĠB", "IB", "REF", "1", ",", "Ġquestion", "-", "ans", "w", "ering", "Ġsystems", "ĠB", "IB", "REF", "2", ",", "Ġand", "Ġhumor", "Ġrecognition", "ĠB", "IB", "REF", "3", ".", "ĠIn", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġwork", "Ġon", "Ġrecognizing", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġaffect", "ive", "Ġevent", "Ġthat", "Ġis", "Ġrepresented", "Ġby", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġ(", "negative", ")", "Ġto", "Ġ1", "Ġ(", "positive", ").", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġchallenging", "Ġbecause", ",", "Ġas", "Ġthe", "Ġexamples", "Ġabove", "Ġsuggest", ",", "Ġthe", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġis", "Ġnot", "Ġnecessarily", "Ġpredictable", "Ġfrom", "Ġits", "Ġconstituent", "Ġwords", ".", "ĠCombined", "Ġwith", "Ġthe", "Ġunb", "ounded", "Ġcomb", "inator", "ial", "Ġnature", "Ġof", "Ġlanguage", ",", "Ġthe", "Ġnon", "-", "com", "position", "ality", "Ġof", "Ġaffect", "ive", "Ġpol", "arity", "Ġentails", "Ġthe", "Ġneed", "Ġfor", "Ġlarge", "Ġamounts", "Ġof", "Ġworld", "Ġknowledge", ",", "Ġwhich", "Ġcan", "Ġhardly", "Ġbe", "Ġlearned", "Ġfrom", "Ġsmall", "Ġannot", "ated", "Ġdata", ".", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġpropose", "Ġa", "Ġsimple", "Ġand", "Ġeffective", "Ġmethod", "Ġfor", "Ġlearning", "Ġaffect", "ive", "Ġevents", "Ġthat", "Ġonly", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠAs", "Ġillustrated", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġour", "Ġkey", "Ġidea", "Ġis", "Ġthat", "Ġwe", "Ġcan", "Ġexploit", "Ġdiscourse", "Ġrelations", "ĠB", "IB", "REF", "4", "Ġto", "Ġefficiently", "Ġpropagate", "Ġpol", "arity", "Ġfrom", "Ġseed", "Ġpred", "icates", "Ġthat", "Ġdirectly", "Ġreport", "Ġone", "'s", "Ġemotions", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "to", "Ġbe", "Ġglad", "âĢ", "Ŀ", "Ġis", "Ġpositive", ").", "ĠSuppose", "Ġthat", "Ġevents", "Ġ$", "x", "_", "1", "$", "Ġare", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCause", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "1", "$", "Ġcauses", "Ġ$", "x", "_", "2", "$", ").", "ĠIf", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġsuggests", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġpositive", ",", "Ġ$", "x", "_", "1", "$", "Ġis", "Ġalso", "Ġlikely", "Ġto", "Ġbe", "Ġpositive", "Ġbecause", "Ġit", "Ġtriggers", "Ġthe", "Ġpositive", "Ġemotion", ".", "ĠThe", "Ġfact", "Ġthat", "Ġ$", "x", "_", "2", "$", "Ġis", "Ġknown", "Ġto", "Ġbe", "Ġnegative", "Ġindicates", "Ġthe", "Ġnegative", "Ġpol", "arity", "Ġof", "Ġ$", "x", "_", "1", "$.", "ĠSimilarly", ",", "Ġif", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġare", "Ġin", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġof", "ĠCon", "cession", "Ġ(", "i", ".", "e", ".,", "Ġ$", "x", "_", "2", "$", "Ġin", "Ġspite", "Ġof", "Ġ$", "x", "_", "1", "$", "),", "Ġthe", "Ġreverse", "Ġof", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġcan", "Ġbe", "Ġpropag", "ated", "Ġto", "Ġ$", "x", "_", "1", "$.", "ĠEven", "Ġif", "Ġ$", "x", "_", "2", "$", "'", "s", "Ġpol", "arity", "Ġis", "Ġnot", "Ġknown", "Ġin", "Ġadvance", ",", "Ġwe", "Ġcan", "Ġexploit", "Ġthe", "Ġtendency", "Ġof", "Ġ$", "x", "_", "1", "$", "Ġand", "Ġ$", "x", "_", "2", "$", "Ġto", "Ġbe", "Ġof", "Ġthe", "Ġsame", "Ġpol", "arity", "Ġ(", "for", "ĠCause", ")", "Ġor", "Ġof", "Ġthe", "Ġreverse", "Ġpol", "arity", "Ġ(", "for", "ĠCon", "cession", ")", "Ġalthough", "Ġthe", "Ġhe", "uristic", "Ġis", "Ġnot", "Ġexempt", "Ġfrom", "Ġcount", "ere", "x", "amples", ".", "ĠWe", "Ġtransform", "Ġthis", "Ġidea", "Ġinto", "Ġobjective", "Ġfunctions", "Ġand", "Ġtrain", "Ġneural", "Ġnetwork", "Ġmodels", "Ġthat", "Ġpredict", "Ġthe", "Ġpol", "arity", "Ġof", "Ġa", "Ġgiven", "Ġevent", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodels", "Ġusing", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", ".", "ĠGiven", "Ġthe", "Ġminimum", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthey", "Ġperformed", "Ġwell", ".", "ĠIn", "Ġaddition", ",", "Ġthe", "Ġcombination", "Ġof", "Ġannot", "ated", "Ġand", "Ġun", "annot", "ated", "Ġdata", "Ġyielded", "Ġa", "Ġgain", "Ġover", "Ġa", "Ġpurely", "Ġsupervised", "Ġbaseline", "Ġwhen", "Ġlabeled", "Ġdata", "Ġwere", "Ġsmall", ".", "</s>", "Related", "ĠWork", "</s>", "Learning", "Ġaffect", "ive", "Ġevents", "Ġis", "Ġclosely", "Ġrelated", "Ġto", "Ġsentiment", "Ġanalysis", ".", "ĠWhereas", "Ġsentiment", "Ġanalysis", "Ġusually", "Ġfocuses", "Ġon", "Ġthe", "Ġpol", "arity", "Ġof", "Ġwhat", "Ġare", "Ġdescribed", "Ġ(", "e", ".", "g", ".,", "Ġmovies", "),", "Ġwe", "Ġwork", "Ġon", "Ġhow", "Ġpeople", "Ġare", "Ġtypically", "Ġaffected", "Ġby", "Ġevents", ".", "ĠIn", "Ġsentiment", "Ġanalysis", ",", "Ġmuch", "Ġattention", "Ġhas", "Ġbeen", "Ġpaid", "Ġto", "Ġcomposition", "ality", ".", "ĠWord", "-", "level", "Ġpol", "arity", "ĠB", "IB", "REF", "5", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "7", "Ġand", "Ġthe", "Ġroles", "Ġof", "Ġneg", "ation", "Ġand", "Ġintens", "ification", "ĠB", "IB", "REF", "8", ",", "ĠB", "IB", "REF", "6", ",", "ĠB", "IB", "REF", "9", "Ġare", "Ġamong", "Ġthe", "Ġmost", "Ġimportant", "Ġtopics", ".", "ĠIn", "Ġcontrast", ",", "Ġwe", "Ġare", "Ġmore", "Ġinterested", "Ġin", "Ġrecognizing", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġof", "Ġan", "Ġevent", "Ġthat", "Ġper", "tains", "Ġto", "Ġcommons", "ense", "Ġknowledge", "Ġ(", "e", ".", "g", ".,", "Ġgetting", "Ġmoney", "Ġand", "Ġcatching", "Ġcold", ").", "</s>", "Label", "Ġpropagation", "Ġfrom", "Ġseed", "Ġinstances", "Ġis", "Ġa", "Ġcommon", "Ġapproach", "Ġto", "Ġinducing", "Ġsentiment", "Ġpolar", "ities", ".", "ĠWhile", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġworked", "Ġon", "Ġword", "-", "Ġand", "Ġphrase", "-", "level", "Ġpolar", "ities", ",", "ĠB", "IB", "REF", "0", "Ġdealt", "Ġwith", "Ġevent", "-", "level", "Ġpolar", "ities", ".", "ĠB", "IB", "REF", "5", "Ġand", "ĠB", "IB", "REF", "10", "Ġlinked", "Ġinstances", "Ġusing", "Ġco", "-", "occ", "urrence", "Ġinformation", "Ġand", "/", "or", "Ġphrase", "-", "level", "Ġcoord", "inations", "Ġ(", "e", ".", "g", ".,", "ĠâĢ", "ľ", "$", "A", "$", "Ġand", "Ġ$", "B", "$", "âĢ", "Ŀ", "Ġand", "ĠâĢ", "ľ", "$", "A", "$", "Ġbut", "Ġ$", "B", "$", "âĢ", "Ŀ", ").", "ĠWe", "Ġshift", "Ġour", "Ġscope", "Ġto", "Ġevent", "Ġpairs", "Ġthat", "Ġare", "Ġmore", "Ġcomplex", "Ġthan", "Ġphrase", "Ġpairs", ",", "Ġand", "Ġconsequently", "Ġexploit", "Ġdiscourse", "Ġconnect", "ives", "Ġas", "Ġevent", "-", "level", "Ġcounterparts", "Ġof", "Ġphrase", "-", "level", "Ġconj", "unctions", ".", "</s>", "BI", "B", "REF", "0", "Ġconstructed", "Ġa", "Ġnetwork", "Ġof", "Ġevents", "Ġusing", "Ġword", "Ġembed", "ding", "-", "derived", "Ġsimilarities", ".", "ĠCompared", "Ġwith", "Ġthis", "Ġmethod", ",", "Ġour", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlinking", "Ġof", "Ġevents", "Ġis", "Ġmuch", "Ġsimpler", "Ġand", "Ġmore", "Ġintuitive", ".", "</s>", "Some", "Ġprevious", "Ġstudies", "Ġmade", "Ġuse", "Ġof", "Ġdocument", "Ġstructure", "Ġto", "Ġunderstand", "Ġthe", "Ġsentiment", ".", "ĠB", "IB", "REF", "11", "Ġproposed", "Ġa", "Ġsentiment", "-", "specific", "Ġpre", "-", "training", "Ġstrategy", "Ġusing", "Ġunl", "abel", "ed", "Ġdialog", "Ġdata", "Ġ(", "t", "weet", "-", "reply", "Ġpairs", ").", "ĠB", "IB", "REF", "12", "Ġproposed", "Ġa", "Ġmethod", "Ġof", "Ġbuilding", "Ġa", "Ġpol", "arity", "-", "tag", "ged", "Ġcorpus", "Ġ(", "ACP", "ĠCorpus", ").", "ĠThey", "Ġautomatically", "Ġgathered", "Ġsentences", "Ġthat", "Ġhad", "Ġpositive", "Ġor", "Ġnegative", "Ġopinions", "Ġutilizing", "ĠHTML", "Ġlayout", "Ġstructures", "Ġin", "Ġaddition", "Ġto", "Ġlinguistic", "Ġpatterns", ".", "ĠOur", "Ġmethod", "Ġdepends", "Ġonly", "Ġon", "Ġraw", "Ġtexts", "Ġand", "Ġthus", "Ġhas", "Ġwider", "Ġapplic", "ability", ".", "</s>", "Prop", "osed", "ĠMethod", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠPol", "arity", "ĠFunction", "</s>", "Our", "Ġgoal", "Ġis", "Ġto", "Ġlearn", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$,", "Ġwhich", "Ġpredicts", "Ġthe", "Ġsentiment", "Ġpol", "arity", "Ġscore", "Ġof", "Ġan", "Ġevent", "Ġ$", "x", "$.", "ĠWe", "Ġapproximate", "Ġ$", "p", "(", "x", ")", "$", "Ġby", "Ġa", "Ġneural", "Ġnetwork", "Ġwith", "Ġthe", "Ġfollowing", "Ġform", ":", "</s>", "$", "{\\", "rm", "ĠEnc", "oder", "}", "$", "Ġoutputs", "Ġa", "Ġvector", "Ġrepresentation", "Ġof", "Ġthe", "Ġevent", "Ġ$", "x", "$.", "Ġ${", "\\", "rm", "ĠLinear", "}", "$", "Ġis", "Ġa", "Ġfully", "-", "connected", "Ġlayer", "Ġand", "Ġtransforms", "Ġthe", "Ġrepresentation", "Ġinto", "Ġa", "Ġscal", "ar", ".", "Ġ${", "\\", "rm", "Ġtan", "h", "}", "$", "Ġis", "Ġthe", "Ġhyper", "b", "olic", "Ġtang", "ent", "Ġand", "Ġtransforms", "Ġthe", "Ġscal", "ar", "Ġinto", "Ġa", "Ġscore", "Ġranging", "Ġfrom", "Ġ$", "-", "1", "$", "Ġto", "Ġ1", ".", "ĠIn", "ĠSection", "ĠSEC", "REF", "21", ",", "Ġwe", "Ġconsider", "Ġtwo", "Ġspecific", "Ġimplementations", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$.", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "</s>", "Our", "Ġmethod", "Ġrequires", "Ġa", "Ġvery", "Ġsmall", "Ġseed", "Ġlex", "icon", "Ġand", "Ġa", "Ġlarge", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġassume", "Ġthat", "Ġwe", "Ġcan", "Ġautomatically", "Ġextract", "Ġdiscourse", "-", "tag", "ged", "Ġevent", "Ġpairs", ",", "Ġ$(", "x", "_{", "i", "1", "},", "Ġx", "_{", "i", "2", "})", "$", "Ġ($", "i", "=", "1", ",", "Ġ\\", "cd", "ots", "Ġ$", ")", "Ġfrom", "Ġthe", "Ġraw", "Ġcorpus", ".", "ĠWe", "Ġrefer", "Ġto", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġas", "Ġformer", "Ġand", "Ġlatter", "Ġevents", ",", "Ġrespectively", ".", "ĠAs", "Ġshown", "Ġin", "ĠFigure", "ĠFIG", "REF", "1", ",", "Ġwe", "Ġlimit", "Ġour", "Ġscope", "Ġto", "Ġtwo", "Ġdiscourse", "Ġrelations", ":", "ĠCause", "Ġand", "ĠCon", "cession", ".", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġconsists", "Ġof", "Ġpositive", "Ġand", "Ġnegative", "Ġpred", "icates", ".", "ĠIf", "Ġthe", "Ġpredicate", "Ġof", "Ġan", "Ġextracted", "Ġevent", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġdoes", "Ġnot", "Ġinvolve", "Ġcomplex", "Ġphenomena", "Ġlike", "Ġneg", "ation", ",", "Ġwe", "Ġassign", "Ġthe", "Ġcorresponding", "Ġpol", "arity", "Ġscore", "Ġ($", "+", "1", "$", "Ġfor", "Ġpositive", "Ġevents", "Ġand", "Ġ$", "-", "1", "$", "Ġfor", "Ġnegative", "Ġevents", ")", "Ġto", "Ġthe", "Ġevent", ".", "ĠWe", "Ġexpect", "Ġthe", "Ġmodel", "Ġto", "Ġautomatically", "Ġlearn", "Ġcomplex", "Ġphenomena", "Ġthrough", "Ġlabel", "Ġpropagation", ".", "ĠBased", "Ġon", "Ġthe", "Ġavailability", "Ġof", "Ġscores", "Ġand", "Ġthe", "Ġtypes", "Ġof", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġclassify", "Ġthe", "Ġextracted", "Ġevent", "Ġpairs", "Ġinto", "Ġthe", "Ġfollowing", "Ġthree", "Ġtypes", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠAL", "Ġ(", "Autom", "atically", "ĠLab", "eled", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġ(", "1", ")", "Ġthe", "Ġlatter", "Ġevent", "Ġbut", "Ġ(", "2", ")", "Ġnot", "Ġthe", "Ġformer", "Ġevent", ",", "Ġand", "Ġ(", "3", ")", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", "Ġor", "ĠCon", "cession", ".", "ĠIf", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġsame", "Ġscore", "Ġas", "Ġthe", "Ġlatter", ".", "ĠLikewise", ",", "Ġif", "Ġthe", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ",", "Ġthe", "Ġformer", "Ġevent", "Ġis", "Ġgiven", "Ġthe", "Ġopposite", "Ġof", "Ġthe", "Ġlatter", "'s", "Ġscore", ".", "ĠThey", "Ġare", "Ġused", "Ġas", "Ġreference", "Ġscores", "Ġduring", "Ġtraining", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCA", "Ġ(", "Cause", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCause", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġsame", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠDisc", "ourse", "ĠRel", "ation", "-", "Based", "ĠEvent", "ĠP", "airs", "Ġ:", "::", "ĠCO", "Ġ(", "Con", "cession", "ĠP", "airs", ")", "</s>", "The", "Ġseed", "Ġlex", "icon", "Ġmatches", "Ġneither", "Ġthe", "Ġformer", "Ġnor", "Ġthe", "Ġlatter", "Ġevent", ",", "Ġand", "Ġtheir", "Ġdiscourse", "Ġrelation", "Ġtype", "Ġis", "ĠCon", "cession", ".", "ĠWe", "Ġassume", "Ġthe", "Ġtwo", "Ġevents", "Ġhave", "Ġthe", "Ġreversed", "Ġpolar", "ities", ".", "</s>", "Prop", "osed", "ĠMethod", "Ġ:", "::", "ĠLoss", "ĠFunctions", "</s>", "Using", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "Ġdata", ",", "Ġwe", "Ġoptimize", "Ġthe", "Ġparameters", "Ġof", "Ġthe", "Ġpol", "arity", "Ġfunction", "Ġ$", "p", "(", "x", ")", "$.", "ĠWe", "Ġdefine", "Ġa", "Ġloss", "Ġfunction", "Ġfor", "Ġeach", "Ġof", "Ġthe", "Ġthree", "Ġtypes", "Ġof", "Ġevent", "Ġpairs", "Ġand", "Ġsum", "Ġup", "Ġthe", "Ġmultiple", "Ġloss", "Ġfunctions", ".", "</s>", "We", "Ġuse", "Ġmean", "Ġsquared", "Ġerror", "Ġto", "Ġconstruct", "Ġloss", "Ġfunctions", ".", "ĠFor", "Ġthe", "ĠAL", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "where", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠAL", "Ġdata", ".", "Ġ$", "r", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "r", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġautomatically", "-", "ass", "igned", "Ġscores", "Ġof", "Ġ$", "x", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "x", "_{", "i", "2", "}", "$,", "Ġrespectively", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠAL", "Ġpairs", ",", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠAL", "}", "$", "Ġis", "Ġa", "Ġhyper", "param", "eter", ".", "</s>", "For", "Ġthe", "ĠCA", "Ġdata", ",", "Ġthe", "Ġloss", "Ġfunction", "Ġis", "Ġdefined", "Ġas", ":", "</s>", "$", "y", "_{", "i", "1", "}", "$", "Ġand", "Ġ$", "y", "_{", "i", "2", "}", "$", "Ġare", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġpair", "Ġof", "Ġthe", "ĠCA", "Ġpairs", ".", "Ġ$", "N", "_{", "\\", "rm", "ĠCA", "}", "$", "Ġis", "Ġthe", "Ġtotal", "Ġnumber", "Ġof", "ĠCA", "Ġpairs", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "mu", "Ġ$", "Ġare", "Ġhyper", "param", "eters", ".", "ĠThe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġcloser", "Ġwhile", "Ġthe", "Ġsecond", "Ġterm", "Ġprevents", "Ġthe", "Ġscores", "Ġfrom", "Ġshrinking", "Ġto", "Ġzero", ".", "</s>", "The", "Ġloss", "Ġfunction", "Ġfor", "Ġthe", "ĠCO", "Ġdata", "Ġis", "Ġdefined", "Ġanalog", "ously", ":", "</s>", "The", "Ġdifference", "Ġis", "Ġthat", "Ġthe", "Ġfirst", "Ġterm", "Ġmakes", "Ġthe", "Ġscores", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġdistant", "Ġfrom", "Ġeach", "Ġother", ".", "</s>", "Exper", "iments", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAL", ",", "ĠCA", ",", "Ġand", "ĠCO", "</s>", "As", "Ġa", "Ġraw", "Ġcorpus", ",", "Ġwe", "Ġused", "Ġa", "ĠJapanese", "Ġweb", "Ġcorpus", "Ġthat", "Ġwas", "Ġcompiled", "Ġthrough", "Ġthe", "Ġprocedures", "Ġproposed", "Ġby", "ĠB", "IB", "REF", "13", ".", "ĠTo", "Ġextract", "Ġevent", "Ġpairs", "Ġtagged", "Ġwith", "Ġdiscourse", "Ġrelations", ",", "Ġwe", "Ġused", "Ġthe", "ĠJapanese", "Ġdependency", "Ġparser", "ĠK", "NP", "Ġand", "Ġin", "-", "house", "Ġpost", "processing", "Ġscripts", "ĠB", "IB", "REF", "14", ".", "ĠK", "NP", "Ġused", "Ġhand", "-", "written", "Ġrules", "Ġto", "Ġsegment", "Ġeach", "Ġsentence", "Ġinto", "Ġwhat", "Ġwe", "Ġconvention", "ally", "Ġcalled", "Ġclauses", "Ġ(", "mostly", "Ġconsecutive", "Ġtext", "Ġchunks", "),", "Ġeach", "Ġof", "Ġwhich", "Ġcontained", "Ġone", "Ġmain", "Ġpredicate", ".", "ĠK", "NP", "Ġalso", "Ġidentified", "Ġthe", "Ġdiscourse", "Ġrelations", "Ġof", "Ġevent", "Ġpairs", "Ġif", "Ġexplicit", "Ġdiscourse", "Ġconnect", "ives", "ĠB", "IB", "REF", "4", "Ġsuch", "Ġas", "ĠâĢ", "ľ", "ãģ®", "ãģ§", "âĢ", "Ŀ", "Ġ(", "because", ")", "Ġand", "ĠâĢ", "ľ", "ãģ®", "ãģ«", "âĢ", "Ŀ", "Ġ(", "in", "Ġspite", "Ġof", ")", "Ġwere", "Ġpresent", ".", "ĠWe", "Ġtreated", "ĠCause", "/", "Reason", "Ġ(", "åİ", "Ł", "åĽ", "ł", "ãĥ»", "çIJ", "Ĩ", "çĶ", "±", ")", "Ġand", "ĠCondition", "Ġ(", "æĿ", "¡", "ä»", "¶", ")", "Ġin", "Ġthe", "Ġoriginal", "Ġtag", "set", "ĠB", "IB", "REF", "15", "Ġas", "ĠCause", "Ġand", "ĠCon", "cession", "Ġ(", "éĢ", "Ĩ", "æ", "İ", "¥", ")", "Ġas", "ĠCon", "cession", ",", "Ġrespectively", ".", "ĠHere", "Ġis", "Ġan", "Ġexample", "Ġof", "Ġevent", "Ġpair", "Ġextraction", ".", "</s>", ".", "Ġé", "ĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĹ", "ãģŁ", "ãģ®", "ãģ§", "ãĢģ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãģ£", "ãģŁ", "ãĢĤ", "</s>", "Because", "Ġ[", "I", "]", "Ġmade", "Ġa", "Ġserious", "Ġmistake", ",", "Ġ[", "I", "]", "Ġgot", "Ġfired", ".", "</s>", "From", "Ġthis", "Ġsentence", ",", "Ġwe", "Ġextracted", "Ġthe", "Ġevent", "Ġpair", "Ġof", "ĠâĢ", "ľ", "éĩ", "į", "大", "ãģª", "å¤", "±", "æķ", "Ĺ", "ãĤĴ", "ç", "Ĭ", "¯", "ãģĻ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġmake", "Ġa", "Ġserious", "Ġmistake", ")", "Ġand", "ĠâĢ", "ľ", "ä»", "ķ", "äº", "ĭ", "ãĤĴ", "ãĤ¯", "ãĥĵ", "ãģ«", "ãģª", "ãĤĭ", "âĢ", "Ŀ", "Ġ([", "I", "]", "Ġget", "Ġfired", "),", "Ġand", "Ġtagged", "Ġit", "Ġwith", "ĠCause", ".", "</s>", "We", "Ġconstructed", "Ġour", "Ġseed", "Ġlex", "icon", "Ġconsisting", "Ġof", "Ġ15", "Ġpositive", "Ġwords", "Ġand", "Ġ15", "Ġnegative", "Ġwords", ",", "Ġas", "Ġshown", "Ġin", "ĠSection", "ĠSEC", "REF", "27", ".", "ĠFrom", "Ġthe", "Ġcorpus", "Ġof", "Ġabout", "Ġ100", "Ġmillion", "Ġsentences", ",", "Ġwe", "Ġobtained", "Ġ1", ".", "4", "Ġmillions", "Ġevent", "Ġpairs", "Ġfor", "ĠAL", ",", "Ġ41", "Ġmillions", "Ġfor", "ĠCA", ",", "Ġand", "Ġ6", "Ġmillions", "Ġfor", "ĠCO", ".", "ĠWe", "Ġrandomly", "Ġselected", "Ġsubs", "ets", "Ġof", "ĠAL", "Ġevent", "Ġpairs", "Ġsuch", "Ġthat", "Ġpositive", "Ġand", "Ġnegative", "Ġlatter", "Ġevents", "Ġwere", "Ġequal", "Ġin", "Ġsize", ".", "ĠWe", "Ġalso", "Ġsampled", "Ġevent", "Ġpairs", "Ġfor", "Ġeach", "Ġof", "ĠCA", "Ġand", "ĠCO", "Ġsuch", "Ġthat", "Ġit", "Ġwas", "Ġfive", "Ġtimes", "Ġlarger", "Ġthan", "ĠAL", ".", "ĠThe", "Ġresults", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "16", ".", "</s>", "Exper", "iments", "Ġ:", "::", "ĠDat", "as", "et", "Ġ:", "::", "ĠAC", "P", "Ġ(", "ACP", "ĠCorpus", ")", "</s>", "We", "Ġused", "Ġthe", "Ġlatest", "Ġversion", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", "ĠB", "IB", "REF", "12", "Ġfor", "Ġevaluation", ".", "ĠIt", "Ġwas", "Ġused", "Ġfor", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġtraining", "Ġas", "Ġwell", ".", "ĠExt", "racted", "Ġfrom", "ĠJapanese", "Ġwebsites", "Ġusing", "ĠHTML", "Ġlayouts", "Ġand", "Ġlinguistic", "Ġpatterns", ",", "Ġthe", "Ġdataset", "Ġcovered", "Ġvarious", "Ġgenres", ".", "ĠFor", "Ġexample", ",", "Ġthe", "Ġfollowing", "Ġtwo", "Ġsentences", "Ġwere", "Ġlabeled", "Ġpositive", "Ġand", "Ġnegative", ",", "Ġrespectively", ":", "</s>", ".", "Ġ", "ä½ľ", "æ", "¥", "Ń", "ãģĮ", "æ", "¥", "½", "ãģł", "ãĢĤ", "</s>", "The", "Ġwork", "Ġis", "Ġeasy", ".", "</s>", ".", "Ġé", "§", "IJ", "è»", "Ĭ", "å", "ł", "´", "ãģĮ", "ãģª", "ãģĦ", "ãĢĤ", "</s>", "There", "Ġis", "Ġno", "Ġparking", "Ġlot", ".", "</s>", "Although", "Ġthe", "ĠAC", "P", "Ġcorpus", "Ġwas", "Ġoriginally", "Ġconstructed", "Ġin", "Ġthe", "Ġcontext", "Ġof", "Ġsentiment", "Ġanalysis", ",", "Ġwe", "Ġfound", "Ġthat", "Ġit", "Ġcould", "Ġroughly", "Ġbe", "Ġregarded", "Ġas", "Ġa", "Ġcollection", "Ġof", "Ġaffect", "ive", "Ġevents", ".", "ĠWe", "Ġparsed", "Ġeach", "Ġsentence", "Ġand", "Ġextracted", "Ġthe", "Ġlast", "Ġclause", "Ġin", "Ġit", ".", "ĠThe", "Ġtrain", "/", "dev", "/", "test", "Ġsplit", "Ġof", "Ġthe", "Ġdata", "Ġis", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "19", ".", "</s>", "The", "Ġobjective", "Ġfunction", "Ġfor", "Ġsupervised", "Ġtraining", "Ġis", ":", "</s>", "where", "Ġ$", "v", "_", "i", "$", "Ġis", "Ġthe", "Ġ$", "i", "$", "-", "th", "Ġevent", ",", "Ġ$", "R", "_", "i", "$", "Ġis", "Ġthe", "Ġreference", "Ġscore", "Ġof", "Ġ$", "v", "_", "i", "$,", "Ġand", "Ġ$", "N", "_{", "\\", "rm", "ĠAC", "P", "}", "$", "Ġis", "Ġthe", "Ġnumber", "Ġof", "Ġthe", "Ġevents", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "</s>", "To", "Ġoptimize", "Ġthe", "Ġhyper", "param", "eters", ",", "Ġwe", "Ġused", "Ġthe", "Ġdev", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠFor", "Ġthe", "Ġevaluation", ",", "Ġwe", "Ġused", "Ġthe", "Ġtest", "Ġset", "Ġof", "Ġthe", "ĠAC", "P", "ĠCorpus", ".", "ĠThe", "Ġmodel", "Ġoutput", "Ġwas", "Ġclassified", "Ġas", "Ġpositive", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ>", "Ġ0", "$", "Ġand", "Ġnegative", "Ġif", "Ġ$", "p", "(", "x", ")", "Ġ\\", "le", "Ġ0", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠModel", "ĠConfig", "urations", "</s>", "As", "Ġfor", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġwe", "Ġcompared", "Ġtwo", "Ġtypes", "Ġof", "Ġneural", "Ġnetworks", ":", "ĠBi", "GR", "U", "Ġand", "ĠB", "ERT", ".", "ĠGR", "U", "ĠB", "IB", "REF", "16", "Ġis", "Ġa", "Ġrecurrent", "Ġneural", "Ġnetwork", "Ġsequence", "Ġenc", "oder", ".", "ĠBi", "GR", "U", "Ġreads", "Ġan", "Ġinput", "Ġsequence", "Ġforward", "Ġand", "Ġbackward", "Ġand", "Ġthe", "Ġoutput", "Ġis", "Ġthe", "Ġconc", "aten", "ation", "Ġof", "Ġthe", "Ġfinal", "Ġforward", "Ġand", "Ġbackward", "Ġhidden", "Ġstates", ".", "</s>", "BER", "T", "ĠB", "IB", "REF", "17", "Ġis", "Ġa", "Ġpre", "-", "trained", "Ġmulti", "-", "layer", "Ġbid", "irection", "al", "ĠTrans", "former", "ĠB", "IB", "REF", "18", "Ġenc", "oder", ".", "ĠIts", "Ġoutput", "Ġis", "Ġthe", "Ġfinal", "Ġhidden", "Ġstate", "Ġcorresponding", "Ġto", "Ġthe", "Ġspecial", "Ġclassification", "Ġtag", "Ġ([", "CL", "S", "]).", "ĠFor", "Ġthe", "Ġdetails", "Ġof", "Ġ${", "\\", "rm", "ĠEnc", "oder", "}", "$,", "Ġsee", "ĠSections", "ĠSEC", "REF", "30", ".", "</s>", "We", "Ġtrained", "Ġthe", "Ġmodel", "Ġwith", "Ġthe", "Ġfollowing", "Ġfour", "Ġcombinations", "Ġof", "Ġthe", "Ġdatasets", ":", "ĠAL", ",", "ĠAL", "+", "CA", "+", "CO", "Ġ(", "two", "Ġproposed", "Ġmodels", "),", "ĠAC", "P", "Ġ(", "super", "vised", "),", "Ġand", "ĠAC", "P", "+", "AL", "+", "CA", "+", "CO", "Ġ(", "se", "mi", "-", "super", "vised", ").", "ĠThe", "Ġcorresponding", "Ġobjective", "Ġfunctions", "Ġwere", ":", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$,", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "$,", "Ġand", "Ġ$\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAC", "P", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠAL", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCA", "}", "Ġ+", "Ġ\\", "math", "cal", "Ġ{", "L", "}", "_{", "\\", "rm", "ĠCO", "}", "$.", "</s>", "Exper", "iments", "Ġ:", "::", "ĠResults", "Ġand", "ĠDiscussion", "</s>", "Table", "ĠT", "AB", "REF", "23", "Ġshows", "Ġaccuracy", ".", "ĠAs", "Ġthe", "ĠRandom", "Ġbaseline", "Ġsuggests", ",", "Ġpositive", "Ġand", "Ġnegative", "Ġlabels", "Ġwere", "Ġdistributed", "Ġevenly", ".", "ĠThe", "ĠRandom", "+", "S", "eed", "Ġbaseline", "Ġmade", "Ġuse", "Ġof", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġand", "Ġoutput", "Ġthe", "Ġcorresponding", "Ġlabel", "Ġ(", "or", "Ġthe", "Ġreverse", "Ġof", "Ġit", "Ġfor", "Ġneg", "ation", ")", "Ġif", "Ġthe", "Ġevent", "'s", "Ġpredicate", "Ġis", "Ġin", "Ġthe", "Ġseed", "Ġlex", "icon", ".", "ĠWe", "Ġcan", "Ġsee", "Ġthat", "Ġthe", "Ġseed", "Ġlex", "icon", "Ġitself", "Ġhad", "Ġpractically", "Ġno", "Ġimpact", "Ġon", "Ġprediction", ".", "</s>", "The", "Ġmodels", "Ġin", "Ġthe", "Ġtop", "Ġblock", "Ġperformed", "Ġconsiderably", "Ġbetter", "Ġthan", "Ġthe", "Ġrandom", "Ġbas", "elines", ".", "ĠThe", "Ġperformance", "Ġgaps", "Ġwith", "Ġtheir", "Ġ(", "se", "mi", "-)", "super", "vised", "Ġcounterparts", ",", "Ġshown", "Ġin", "Ġthe", "Ġmiddle", "Ġblock", ",", "Ġwere", "Ġless", "Ġthan", "Ġ7", "%.", "ĠThis", "Ġdemonstrates", "Ġthe", "Ġeffectiveness", "Ġof", "Ġdiscourse", "Ġrelation", "-", "based", "Ġlabel", "Ġpropagation", ".", "</s>", "Comp", "aring", "Ġthe", "Ġmodel", "Ġvariants", ",", "Ġwe", "Ġobtained", "Ġthe", "Ġhighest", "Ġscore", "Ġwith", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġtrained", "Ġwith", "Ġthe", "ĠAL", "+", "CA", "+", "CO", "Ġdataset", ".", "ĠB", "ERT", "Ġwas", "Ġcompetitive", "Ġbut", "Ġits", "Ġperformance", "Ġwent", "Ġdown", "Ġif", "ĠCA", "Ġand", "ĠCO", "Ġwere", "Ġused", "Ġin", "Ġaddition", "Ġto", "ĠAL", ".", "ĠWe", "Ġconjecture", "Ġthat", "ĠB", "ERT", "Ġwas", "Ġmore", "Ġsensitive", "Ġto", "Ġnoises", "Ġfound", "Ġmore", "Ġfrequently", "Ġin", "ĠCA", "Ġand", "ĠCO", ".", "</s>", "Cont", "rary", "Ġto", "Ġour", "Ġexpectations", ",", "Ġsupervised", "Ġmodels", "Ġ(", "ACP", ")", "Ġoutper", "formed", "Ġsemi", "-", "super", "vised", "Ġmodels", "Ġ(", "ACP", "+", "AL", "+", "CA", "+", "CO", ").", "ĠThis", "Ġsuggests", "Ġthat", "Ġthe", "Ġtraining", "Ġset", "Ġof", "Ġ0", ".", "6", "Ġmillion", "Ġevents", "Ġis", "Ġsufficiently", "Ġlarge", "Ġfor", "Ġtraining", "Ġthe", "Ġmodels", ".", "ĠFor", "Ġcomparison", ",", "Ġwe", "Ġtrained", "Ġthe", "Ġmodels", "Ġwith", "Ġa", "Ġsubset", "Ġ(", "6", ",", "000", "Ġevents", ")", "Ġof", "Ġthe", "ĠAC", "P", "Ġdataset", ".", "ĠAs", "Ġthe", "Ġresults", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "24", "Ġdemonstrate", ",", "Ġour", "Ġmethod", "Ġis", "Ġeffective", "Ġwhen", "Ġlabeled", "Ġdata", "Ġare", "Ġsmall", ".", "</s>", "The", "Ġresult", "Ġof", "Ġhyper", "param", "eter", "Ġoptimization", "Ġfor", "Ġthe", "ĠBi", "GR", "U", "Ġenc", "oder", "Ġwas", "Ġas", "Ġfollows", ":", "</s>", "As", "Ġthe", "ĠCA", "Ġand", "ĠCO", "Ġpairs", "Ġwere", "Ġequal", "Ġin", "Ġsize", "Ġ(", "Table", "ĠT", "AB", "REF", "16", "),", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġand", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$", "Ġwere", "Ġcomparable", "Ġvalues", ".", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$", "Ġwas", "Ġabout", "Ġone", "-", "third", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCO", "}", "$,", "Ġand", "Ġthis", "Ġindicated", "Ġthat", "Ġthe", "ĠCA", "Ġpairs", "Ġwere", "Ġno", "is", "ier", "Ġthan", "Ġthe", "ĠCO", "Ġpairs", ".", "ĠA", "Ġmajor", "Ġtype", "Ġof", "ĠCA", "Ġpairs", "Ġthat", "Ġviolates", "Ġour", "Ġassumption", "Ġwas", "Ġin", "Ġthe", "Ġform", "Ġof", "ĠâĢ", "ľ", "$", "\\", "text", "it", "Ġ{", "problem", "}", "_{", "\\", "text", "{", "negative", "}}", "$", "Ġcauses", "Ġ$\\", "text", "it", "Ġ{", "s", "olution", "}", "_{", "\\", "text", "{", "positive", "}}", "$", "âĢ", "Ŀ", ":", "</s>", ".", "Ġ(", "æ", "Ĥª", "ãģĦ", "ãģ¨", "ãģĵ", "ãĤ", "į", "ãģĮ", "ãģĤ", "ãĤĭ", ",", "ĠãĤ", "Ī", "ãģı", "ãģª", "ãĤĭ", "ãĤ", "Ī", "ãģĨ", "ãģ«", "åĬ", "ª", "åĬ", "Ľ", "ãģĻ", "ãĤĭ", ")", "</s>", "(", "there", "Ġis", "Ġa", "Ġbad", "Ġpoint", ",", "Ġ[", "I", "]", "Ġtry", "Ġto", "Ġimprove", "Ġ[", "it", "])", "</s>", "The", "Ġpolar", "ities", "Ġof", "Ġthe", "Ġtwo", "Ġevents", "Ġwere", "Ġreversed", "Ġin", "Ġspite", "Ġof", "Ġthe", "ĠCause", "Ġrelation", ",", "Ġand", "Ġthis", "Ġlowered", "Ġthe", "Ġvalue", "Ġof", "Ġ$\\", "lambda", "Ġ_", "{\\", "rm", "ĠCA", "}", "$.", "</s>", "Some", "Ġexamples", "Ġof", "Ġmodel", "Ġoutputs", "Ġare", "Ġshown", "Ġin", "ĠTable", "ĠT", "AB", "REF", "26", ".", "ĠThe", "Ġfirst", "Ġtwo", "Ġexamples", "Ġsuggest", "Ġthat", "Ġour", "Ġmodel", "Ġsuccessfully", "Ġlearned", "Ġneg", "ation", "Ġwithout", "Ġexplicit", "Ġsupervision", ".", "ĠSimilarly", ",", "Ġthe", "Ġnext", "Ġtwo", "Ġexamples", "Ġdiffer", "Ġonly", "Ġin", "Ġvoice", "Ġbut", "Ġthe", "Ġmodel", "Ġcorrectly", "Ġrecognized", "Ġthat", "Ġthey", "Ġhad", "Ġopposite", "Ġpolar", "ities", ".", "ĠThe", "Ġlast", "Ġtwo", "Ġexamples", "Ġshare", "Ġthe", "Ġpredicate", "ĠâĢ", "ľ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "drop", ")", "Ġand", "Ġonly", "Ġthe", "Ġobjects", "Ġare", "Ġdifferent", ".", "ĠThe", "Ġsecond", "Ġevent", "ĠâĢ", "ľ", "è", "Ĥ", "©", "ãĤĴ", "è", "IJ", "½", "ãģ¨", "ãģĻ", "\"", "Ġ(", "lit", ".", "Ġdrop", "Ġone", "'s", "Ġshoulders", ")", "Ġis", "Ġan", "Ġid", "iom", "Ġthat", "Ġexpresses", "Ġa", "Ġdisappointed", "Ġfeeling", ".", "ĠThe", "Ġexamples", "Ġdemonstrate", "Ġthat", "Ġour", "Ġmodel", "Ġcorrectly", "Ġlearned", "Ġnon", "-", "com", "pos", "itional", "Ġexpressions", ".", "</s>", "Conclusion", "</s>", "In", "Ġthis", "Ġpaper", ",", "Ġwe", "Ġproposed", "Ġto", "Ġuse", "Ġdiscourse", "Ġrelations", "Ġto", "Ġeffectively", "Ġpropagate", "Ġpolar", "ities", "Ġof", "Ġaffect", "ive", "Ġevents", "Ġfrom", "Ġseeds", ".", "ĠExper", "iments", "Ġshow", "Ġthat", ",", "Ġeven", "Ġwith", "Ġa", "Ġminimal", "Ġamount", "Ġof", "Ġsupervision", ",", "Ġthe", "Ġproposed", "Ġmethod", "Ġperformed", "Ġwell", ".", "</s>", "Although", "Ġevent", "Ġpairs", "Ġlinked", "Ġby", "Ġdiscourse", "Ġanalysis", "Ġare", "Ġshown", "Ġto", "Ġbe", "Ġuseful", ",", "Ġthey", "Ġnevertheless", "Ġcontain", "Ġnoises", ".", "ĠAdding", "Ġlingu", "istically", "-", "mot", "ivated", "Ġfiltering", "Ġrules", "Ġwould", "Ġhelp", "Ġimprove", "Ġthe", "Ġperformance", ".", "</s>", "Acknowled", "gments", "</s>", "We", "Ġthank", "ĠNob", "u", "hiro", "ĠK", "aji", "Ġfor", "Ġproviding", "Ġthe", "ĠAC", "P", "ĠCorpus", "Ġand", "ĠHiro", "k", "az", "u", "ĠK", "iy", "om", "aru", "Ġand", "ĠY", "ud", "ai", "ĠK", "ish", "imoto", "Ġfor", "Ġtheir", "Ġhelp", "Ġin", "Ġextracting", "Ġevent", "Ġpairs", ".", "ĠThis", "Ġwork", "Ġwas", "Ġpartially", "Ġsupported", "Ġby", "ĠYahoo", "!", "ĠJapan", "ĠCorporation", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠPositive", "ĠWords", "</s>", "å", "ĸ", "ľ", "ãģ", "¶", "Ġ(", "re", "joice", "),", "Ġå", "¬", "ī", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġglad", "),", "Ġæ", "¥", "½", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġpleasant", "),", "Ġå", "¹", "¸", "ãģ", "Ľ", "Ġ(", "be", "Ġhappy", "),", "Ġæ", "Ħ", "Ł", "åĭ", "ķ", "Ġ(", "be", "Ġimpressed", "),", "Ġè", "Ī", "Ī", "å¥", "®", "Ġ(", "be", "Ġexcited", "),", "Ġæ", "ĩ", "IJ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "feel", "Ġnostalgic", "),", "Ġå", "¥", "½", "ãģį", "Ġ(", "like", "),", "Ġå", "°", "Ĭ", "æķ", "¬", "Ġ(", "respect", "),", "Ġå", "®", "ī", "å¿", "ĥ", "Ġ(", "be", "Ġrelieved", "),", "Ġæ", "Ħ", "Ł", "å¿", "ĥ", "Ġ(", "ad", "mire", "),", "Ġè", "IJ", "½", "ãģ", "¡", "ç", "Ŀ", "Ģ", "ãģı", "Ġ(", "be", "Ġcalm", "),", "Ġæ", "º", "Ģ", "è", "¶", "³", "Ġ(", "be", "Ġsatisfied", "),", "Ġç", "Ļ", "Ĵ", "ãģķ", "ãĤĮ", "ãĤĭ", "Ġ(", "be", "Ġhealed", "),", "Ġand", "Ġ", "ãĤ¹", "ãĥĥ", "ãĤŃ", "ãĥª", "Ġ(", "be", "Ġrefreshed", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSeed", "ĠLex", "icon", "Ġ:", "::", "ĠNegative", "ĠWords", "</s>", "æĢ", "Ĵ", "ãĤĭ", "Ġ(", "get", "Ġangry", "),", "Ġæ", "Ĥ", "²", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġsad", "),", "Ġå", "¯", "Ĥ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġlonely", "),", "Ġæ", "Ģ", "ĸ", "ãģĦ", "Ġ(", "be", "Ġscared", "),", "Ġ", "ä¸į", "å®", "ī", "Ġ(", "feel", "Ġanxious", "),", "Ġæ", "ģ", "¥", "ãģ", "ļ", "ãģĭ", "ãģĹ", "ãģĦ", "Ġ(", "be", "Ġembarrassed", "),", "Ġå", "«", "Į", "Ġ(", "hate", "),", "Ġè", "IJ", "½", "ãģ", "¡", "è", "¾", "¼", "ãĤ", "Ģ", "Ġ(", "feel", "Ġdown", "),", "Ġé", "Ģ", "Ģ", "å", "±", "Ī", "Ġ(", "be", "Ġbored", "),", "Ġç", "µ", "¶", "æľ", "Ľ", "Ġ(", "feel", "Ġhopeless", "),", "Ġè", "¾", "Ľ", "ãģĦ", "Ġ(", "have", "Ġa", "Ġhard", "Ġtime", "),", "Ġå", "Ľ", "°", "ãĤĭ", "Ġ(", "have", "Ġtrouble", "),", "Ġæ", "Ĩ", "Ĥ", "é", "¬", "±", "Ġ(", "be", "Ġdepressed", "),", "Ġå", "¿", "ĥ", "é", "ħ", "į", "Ġ(", "be", "Ġworried", "),", "Ġand", "Ġæ", "ĥ", "ħ", "ãģ", "ij", "ãģª", "ãģĦ", "Ġ(", "be", "Ġsorry", ").", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠBi", "GR", "U", "</s>", "The", "Ġdimension", "Ġof", "Ġthe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġ256", ".", "ĠThe", "Ġembed", "ding", "Ġlayer", "Ġwas", "Ġinitialized", "Ġwith", "Ġthe", "Ġword", "Ġembed", "d", "ings", "Ġpret", "rained", "Ġusing", "Ġthe", "ĠWeb", "Ġcorpus", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "Ġthe", "Ġmorph", "ological", "Ġanaly", "zer", "ĠJ", "uman", "++", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ100", ",", "000", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ2", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ256", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠMoment", "um", "ĠSG", "D", "ĠB", "IB", "REF", "21", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ1024", ".", "ĠWe", "Ġran", "Ġ100", "Ġepoch", "s", "Ġand", "Ġselected", "Ġthe", "Ġsnapshot", "Ġthat", "Ġachieved", "Ġthe", "Ġhighest", "Ġscore", "Ġfor", "Ġthe", "Ġdev", "Ġset", ".", "</s>", "App", "end", "ices", "Ġ:", "::", "ĠSettings", "Ġof", "ĠEnc", "oder", "Ġ:", "::", "ĠB", "ERT", "</s>", "We", "Ġused", "Ġa", "ĠJapanese", "ĠB", "ERT", "Ġmodel", "Ġpret", "rained", "Ġwith", "ĠJapanese", "ĠWikipedia", ".", "ĠThe", "Ġinput", "Ġsentences", "Ġwere", "Ġsegment", "ed", "Ġinto", "Ġwords", "Ġby", "ĠJ", "uman", "++", ",", "Ġand", "Ġwords", "Ġwere", "Ġbroken", "Ġinto", "Ġsub", "words", "Ġby", "Ġapplying", "ĠB", "PE", "ĠB", "IB", "REF", "20", ".", "ĠThe", "Ġvocabulary", "Ġsize", "Ġwas", "Ġ32", ",", "000", ".", "ĠThe", "Ġmaximum", "Ġlength", "Ġof", "Ġan", "Ġinput", "Ġsequence", "Ġwas", "Ġ128", ".", "ĠThe", "Ġnumber", "Ġof", "Ġhidden", "Ġlayers", "Ġwas", "Ġ12", ".", "ĠThe", "Ġdimension", "Ġof", "Ġhidden", "Ġunits", "Ġwas", "Ġ768", ".", "ĠThe", "Ġnumber", "Ġof", "Ġself", "-", "att", "ention", "Ġheads", "Ġwas", "Ġ12", ".", "ĠThe", "Ġoptim", "izer", "Ġwas", "ĠAdam", "ĠB", "IB", "REF", "19", ".", "ĠThe", "Ġmini", "-", "batch", "Ġsize", "Ġwas", "Ġ32", ".", "ĠWe", "Ġran", "Ġ1", "Ġepoch", "." ]
[ 6, 8, 135, 209, 541, 590, 593, 732, 868, 906, 1000, 1004, 1013, 1059, 1148, 1163, 1268, 1369, 1395, 1483, 1506, 1539, 1563, 1597, 1605, 1652, 1675, 1776, 1789, 1874, 1887, 1907, 1910, 1918, 1934, 2125, 2157, 2173, 2236, 2345, 2361, 2426, 2439, 2445, 2459, 2466, 2529, 2538, 2591, 2653, 2661, 2728, 2789, 2954, 2962, 3041, 3093, 3160, 3252, 3271, 3397, 3427, 3444, 3475, 3601, 3603, 3645, 3678, 3681, 3730, 3743, 3884, 3897, 4040, 4055, 4167, 4181 ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
{"all_answers":[{"text":"based on the relation between events, the suggested polarity of one event c(...TRUNCATED)
["based","Ġon","Ġthe","Ġrelation","Ġbetween","Ġevents",",","Ġthe","Ġsuggested","Ġpol","arity(...TRUNCATED)
["<s>","How","Ġare","Ġrelations","Ġused","Ġto","Ġpropagate","Ġpol","arity","?","</s>","Introdu(...TRUNCATED)
[10,12,139,213,545,594,597,736,872,910,1004,1008,1017,1063,1152,1167,1272,1373,1399,1487,1510,1543,1(...TRUNCATED)
[0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"7000000 pairs of events were extracted from the Japanese Web corpus, 529850(...TRUNCATED)
["7","000000","Ġpairs","Ġof","Ġevents","Ġwere","Ġextracted","Ġfrom","Ġthe","ĠJapanese","ĠWe(...TRUNCATED)
["<s>","How","Ġbig","Ġis","Ġthe","ĠJapanese","Ġdata","?","</s>","Introduction","</s>","A","ffec(...TRUNCATED)
[8,10,137,211,543,592,595,734,870,908,1002,1006,1015,1061,1150,1165,1270,1371,1397,1485,1508,1541,15(...TRUNCATED)
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,1,0,1,0,0,0,0,1,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"negative, positive","type":"EXTRACTIVE"}],"all_evidence":[["Affective event(...TRUNCATED)
[ "negative", ",", "Ġpositive" ]
["<s>","What","Ġare","Ġlabels","Ġavailable","Ġin","Ġdataset","Ġfor","Ġsupervision","?","</s>"(...TRUNCATED)
[10,12,139,213,545,594,597,736,872,910,1004,1008,1017,1063,1152,1167,1272,1373,1399,1487,1510,1543,1(...TRUNCATED)
[0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"3%","type":"ABSTRACTIVE"}],"all_evidence":[["FLOAT SELECTED: Table 4: Resul(...TRUNCATED)
[ "3", "%" ]
["<s>","How","Ġbig","Ġare","Ġimprovements","Ġof","Ġsuper","vs","zed","Ġlearning","Ġresults","(...TRUNCATED)
[27,29,156,230,562,611,614,753,889,927,1021,1025,1034,1080,1169,1184,1289,1390,1416,1504,1527,1560,1(...TRUNCATED)
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"by exploiting discourse relations to propagate polarity from seed predicate(...TRUNCATED)
["by","Ġexploiting","Ġdiscourse","Ġrelations","Ġto","Ġpropagate","Ġpol","arity","Ġfrom","Ġse(...TRUNCATED)
["<s>","How","Ġdoes","Ġtheir","Ġmodel","Ġlearn","Ġusing","Ġmostly","Ġraw","Ġdata","?","</s>"(...TRUNCATED)
[11,13,140,214,546,595,598,737,873,911,1005,1009,1018,1064,1153,1168,1273,1374,1400,1488,1511,1544,1(...TRUNCATED)
[0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"30 words","type":"ABSTRACTIVE"}],"all_evidence":[["We constructed our seed (...TRUNCATED)
[ "30", "Ġwords" ]
["<s>","How","Ġbig","Ġis","Ġseed","Ġlex","icon","Ġused","Ġfor","Ġtraining","?","</s>","Introd(...TRUNCATED)
[11,13,140,214,546,595,598,737,873,911,1005,1009,1018,1064,1153,1168,1273,1374,1400,1488,1511,1544,1(...TRUNCATED)
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"100 million sentences","type":"EXTRACTIVE"}],"all_evidence":[["As a raw cor(...TRUNCATED)
[ "100", "Ġmillion", "Ġsentences" ]
["<s>","How","Ġlarge","Ġis","Ġraw","Ġcorpus","Ġused","Ġfor","Ġtraining","?","</s>","Introduct(...TRUNCATED)
[10,12,139,213,545,594,597,736,872,910,1004,1008,1017,1063,1152,1167,1272,1373,1399,1487,1510,1543,1(...TRUNCATED)
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)
{"all_answers":[{"text":"Yes","type":"BOOLEAN"},{"text":"Yes","type":"BOOLEAN"}],"all_evidence":[["F(...TRUNCATED)
[ "Yes" ]
["<s>","Does","Ġthe","Ġpaper","Ġreport","Ġmacro","ĠF","1","?","</s>","</s>","1",".","1","em","<(...TRUNCATED)
[9,10,15,19,26,33,42,85,108,131,157,191,215,448,476,478,668,873,952,1180,1404,1602,1639,1650,1821,20(...TRUNCATED)
[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED)

Preprocessed QASPER dataset

Working doc: https://docs.google.com/document/d/1gYPhPNJ5LGttgjix1dwai8pdNcqS6PbqhsM7W0rhKNQ/edit?usp=sharing

Original:

Differences of our implementation over the original implementation:

  1. We use the dataset provided at https://huggingface.co/datasets/allenai/qasper since it doesn't require manually downloading files.
  2. We remove usage of allennlp since the Python package cannot be installed anymore.
  3. We add baselines to qasper/models. Currently, we have
    • QASPER (Longformer Encoder Decoder)
    • GPT-3.5-Turbo
    • TODO: RAG (with R=TF-IDF or Contriever) implemented in LangChain?
  4. We replace allennlp special tokens with the special tokens of the HF transformer tokenizer:
    • paragraph separator: '' -> tokenizer.sep_token
    • sequence pair start tokens: _tokenizer.sequence_pair_start_tokens -> tokenizer.bos_token

Usage

from datasets import load_dataset

dataset = load_dataset("ag2435/qasper")
Downloads last month
58
Edit dataset card