Datasets:

havens2

/

naacl2022

like 0

[ "(2018)", "designed", "a", "hierarchical", "aspect-specific", "attention", "model", "for", "aspect", "sentiment", "classification." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Wang", "et", "al." ]

[ 0, 0, 0 ]

[ "Ma", "et", "al.", "(2017)", "introduced", "an", "interactive", "attention", "mechanism", "to", "generate", "the", "representations", "for", "aspects", "and", "contexts", "separately." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Chen", "et", "al.", "(2017)", "proposed", "a", "multi-layer", "attention", "network", "to", "infer", "the", "sentiment", "polarity", "for", "the", "aspect." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Among", "them,", "(Wang", "et", "al.,", "2016)", "utilized", "attention", "mechanism", "to", "concentrate", "on", "different", "parts", "of", "a", "sentence", "to", "generate", "an", "attention", "vector", "for", "aspect", "sentiment", "classification." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Most", "recent", "researches", "solve", "aspect-based", "sentiment", "analysis", "by", "utilizing", "attention-based", "neural", "network", "to", "model", "semantic", "correlation", "between", "context", "and", "aspect", "term", "(Wang", "et", "al.,", "2016;", "Chen", "et", "al.,", "2017;", "Ma", "et", "al.,", "2017;", "Liu", "et", "al.,", "2018;", "Hu", "et", "al.,", "2018;", "Wang", "et", "al.,", "2018;", "Huang", "et", "al.,", "2018;", "Fan", "et", "al.,", "2018)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Earlier", "methods", "(Jiang", "et", "al.,", "2011;", "Kiritchenko", "et", "al.,", "2014)", "manually", "defined", "some", "syntactic", "rules", "to", "predict", "the", "sentiment", "polarity", "of", "aspect", "term." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Aspect-based", "sentiment", "analysis", "is", "a", "fine-grained", "sentiment", "analysis", "task", "and", "generally", "treated", "as", "a", "classification", "problem." ]

[ 11, 12, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "2", "Related", "Work" ]

[ 0, 0, 0 ]

[ "The", "code", "and", "datasets", "involved", "in", "this", "paper", "are", "provided", "on", "Github1." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "•", "Experimental", "results", "on", "three", "benchmark", "datasets", "show", "that", "the", "SSEGCN", "model", "achieves", "the", "stateof-the-art", "performance." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]

[ "Meanwhile,", "we", "construct", "syntactic", "mask", "matrices", "to", "complement", "with", "semantic", "information." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "propose", "an", "aspect-aware", "attention", "mechanism", "combined", "with", "self-attention", "to", "learn", "the", "aspect-related", "semantic", "correlations", "and", "the", "global", "semantic", "of", "the", "sentence." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "propose", "a", "SSEGCN", "model", "that", "effectively", "integrates", "syntactic", "structure", "and", "semantic", "correlation", "for", "ABSA", "task." ]

[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0 ]

[ "•" ]

[ 0 ]

[ "Our", "contributions", "are", "summarized", "as", "follows:" ]

[ 0, 0, 0, 0, 0, 0 ]

[ "The", "results", "show", "that", "our", "model", "is", "more", "effective", "than", "a", "range", "of", "baselines", "and", "achieves", "new", "state-of-the-art", "performance." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "evaluate", "our", "approach", "on", "three", "benchmark", "datasets." ]

[ 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Finally,", "a", "multilayer", "graph", "convolution", "operation", "is", "implemented", "to", "obtain", "aspect-specific", "features", "for", "aspect", "term", "sentiment", "classification." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "to", "enhance", "the", "conventional", "GCN." ]

[ 0, 0, 0, 0, 0 ]

[ "Then,", "we", "combine", "adjacency", "matrices", "with", "syntactic", "mask", "matrices" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Besides,", "to", "fully", "utilize", "syntactic", "structure", "to", "complement", "semantic", "information,", "rather", "than", "just", "syntactic", "first-order", "neighbor", "node", "information,", "we", "construct", "the", "syntactic", "mask", "matrices", "calculated", "by", "the", "different", "distances", "between", "words", "in", "the", "syntactic", "dependency", "structure", "of", "the", "sentence", "to", "learn", "structure", "information", "from", "local", "to", "global." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "take", "the", "obtained", "attention", "scores", "as", "the", "initial", "adjacency", "matrices", "for", "GCN." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "aspect-aware", "attention", "learns", "aspect-related", "semantic", "information,", "while", "selfattention", "learns", "global", "semantic", "of", "the", "sentence." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Secondly,", "to", "model", "particular", "semantic", "correlations", "for", "different", "aspect", "terms,", "we", "propose", "an", "aspect-aware", "attention", "mechanism", "to", "combine", "with", "self-attention." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Firstly,", "SSEGCN", "captures", "the", "contextualized", "word", "representations", "with", "sentence", "encoder." ]

[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "this", "paper,", "we", "propose", "a", "novel", "Syntactic", "and", "Semantic", "Enhanced", "Graph", "Convolutional", "Network", "(SSEGCN)", "model", "for", "integrating", "the", "syntactic", "and", "semantic", "information", "of", "a", "sentence", "to", "solve", "the", "above", "issues." ]

[ 0, 0, 0, 0, 0, 0, 0, 1, 2, 2, 2, 2, 2, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "It", "is", "necessary", "to", "harness", "the", "aspect-related", "semantic", "information", "for", "different", "aspect", "terms." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "As", "a", "result,", "simply", "considering", "the", "syntactic", "structure", "of", "dependency", "tree", "might", "be", "unsatisfactory." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "At", "the", "same", "time,", "noisy", "word", "“not”", "is", "also", "obtained", "for", "aspect", "term", "“service”." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "aspect", "term", "“food”,", "the", "representation", "of", "word", "“not”", "is", "obtained", "by", "utilizing", "two-layer", "GCN." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "the", "“good”", "refers", "to", "another", "aspect", "“service”." ]

[ 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "instance,", "consider", "the", "dependency", "tree", "as", "depicted", "in", "Figure", "1,", "a", "dependency", "connection", "exists", "between", "the", "aspect", "“food”", "and", "the", "opinion", "word", "“good”." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Recently,", "most", "methods", "apply", "multi-layer", "GCNs", "to", "derive", "the", "expression", "of", "opinion", "words", "which", "brings", "potential", "noise." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "How", "to", "capture", "the", "information", "of", "secondorder", "nodes,", "third-order", "nodes", "and", "even", "the", "global", "syntactic", "structure", "in", "one", "shot", "is", "still", "a", "challenge." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Moreover,", "some", "hard", "cases", "obscurely", "express", "sentiment", "of", "aspect", "term,", "there", "is", "no", "direct", "syntactic", "relationship", "between", "aspect", "term", "and", "opinion", "words." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Notablely,", "existing", "GCN-based", "approaches", "do", "not", "fully", "leverage", "syntactic", "structure,", "where", "only", "the", "information", "of", "the", "neighbor", "nodes", "is", "considered." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Following", "this", "line,", "Chen", "et", "al.", "(2020)", "and", "Li", "et", "al.", "(2021)", "combine", "a", "syntactic", "structure", "graph", "with", "a", "latent", "semantic", "graph", "but", "the", "two", "graphs", "are", "constructed", "independently." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "cal", "attention", "by", "only", "calculating", "the", "importance", "of", "neighboring", "nodes", "and", "neglects", "the", "global", "information", "of", "the", "sentence." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "attention", "mechanism", "is", "used", "to", "consider", "semantic", "correlation", "of", "each", "neighbor", "node,", "but", "it", "is", "a", "lo" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "To", "tackle", "this", "problem,", "Huang", "and", "Carley", "(2019)", "design", "a", "target-dependent", "graph", "attention", "network", "that", "updates", "each", "node", "representation", "by", "utilizing", "multi-head", "attention." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Accordingly,", "noisy", "nodes", "may", "cause", "the", "model", "to", "misjudge", "the", "sentiment", "polarity." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "these", "two", "studies", "treat", "all", "the", "neighbor", "nodes", "of", "the", "current", "node", "in", "the", "graph", "equally,", "and", "are", "limited", "in", "lacking", "of", "efficient", "mechanisms", "to", "distinguish", "the", "importance", "of", "neighbor", "nodes." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Sun", "et", "al.", "(2019)", "propose", "a", "GCN", "model", "to", "enhance", "the", "feature", "representations", "of", "aspects", "learned", "by", "a", "Bi-directional", "Long", "Short", "Term", "Memory", "(Bi-LSTM)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "(2019)", "employ", "Graph", "Convolutional", "Network", "(GCN)", "to", "integrate", "the", "syntactic", "information." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "these", "works,", "Zhang", "et", "al." ]

[ 0, 0, 0, 0, 0, 0 ]

[ "Recent", "works", "on", "ABSA", "leverage", "Graph", "Neural", "Networks", "(GNNs)", "over", "the", "dependency", "tree", "of", "a", "sentence", "to", "exploit", "syntactic", "structure", "(Sun", "et", "al.,", "2019;", "Zhang", "et", "al.,", "2019;", "Liang", "et", "al.,", "2020;", "Wang", "et", "al.,", "2020;", "Zhao", "et", "al.,", "2020;", "Li", "et", "al.,", "2021;", "Tian", "et", "al.,", "2021)." ]

[ 0, 0, 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "However,", "attention", "mechanism", "is", "vulnerable", "to", "noise", "in", "sentences,", "i.e.,", "the", "irrelevant", "words." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Prior", "studies", "exploit", "attention", "mechanism", "(Wang", "et", "al.,", "2016;", "Chen", "et", "al.,", "2017;", "Ma", "et", "al.,", "2017;", "Liu", "et", "al.,", "2018;", "Hu", "et", "al.,", "2018;", "Wang", "et", "al.,", "2018;", "Huang", "et", "al.,", "2018;", "Fan", "et", "al.,", "2018)", "to", "model", "the", "correlations", "between", "aspect", "term", "and", "the", "context." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "The", "main", "idea", "of", "most", "works", "is", "to", "model", "the", "dependency", "relation", "between", "aspects", "and", "their", "associated", "opinion", "words." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "different", "aspects." ]

[ 0, 0 ]

[ "That", "is,", "we", "need", "to", "discriminate", "sentiment", "polarities", "according", "to" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "aspect", "term", "“food”,", "the", "sentiment", "polarity", "is", "negative,", "but", "“service”", "is", "positive." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "example,", "in", "Figure", "1,", "ABSA", "determines", "the", "sentiment", "towards", "the", "aspects", "“food”", "and", "“service”." ]

[ 0, 0, 0, 0, 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Aspect-based", "Sentiment", "Analysis", "(ABSA)", "aims", "to", "determine", "the", "sentiment", "polarity", "of", "a", "given", "aspect", "term", "in", "a", "sentence,", "where", "sentiment", "polarity", "includes", "positive,", "negative", "and", "neutral." ]

[ 11, 12, 12, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Introduction" ]

[ 0 ]

[ "Experimental", "results", "on", "benchmark", "datasets", "illustrate", "that", "our", "proposed", "model", "outperforms", "state-of-the-art", "methods." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Finally,", "we", "enhance", "the", "node", "representations", "with", "graph", "convolutional", "network", "over", "attention", "score", "matrices", "for", "ABSA." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11 ]

[ "Furthermore,", "to", "combine", "syntactic", "structure", "and", "semantic", "information,", "we", "equip", "the", "attention", "score", "matrices", "by", "syntactic", "mask", "matrices." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "order", "to", "obtain", "comprehensive", "syntactic", "structure", "information,", "we", "construct", "syntactic", "mask", "matrices", "of", "the", "sentence", "according", "to", "the", "different", "syntactic", "distances", "between", "words." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Specifically,", "we", "propose", "an", "aspect-aware", "attention", "mechanism", "combined", "with", "self-attention", "to", "obtain", "attention", "score", "matrices", "of", "a", "sentence,", "which", "can", "not", "only", "learn", "the", "aspect-related", "semantic", "correlations,", "but", "also", "learn", "the", "global", "semantics", "of", "the", "sentence." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "In", "this", "paper,", "we", "propose", "a", "novel", "Syntactic", "and", "Semantic", "Enhanced", "Graph", "Convolutional", "Network", "(SSEGCN)", "model", "for", "ABSA", "task." ]

[ 0, 0, 0, 0, 0, 0, 0, 1, 2, 2, 2, 2, 2, 2, 1, 0, 0, 11, 0 ]

[ "However,", "how", "to", "effectively", "harness", "the", "semantic", "and", "syntactic", "structure", "information", "from", "the", "dependency", "tree", "remains", "a", "challenging", "research", "question." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Recently,", "graph", "neural", "networks", "based", "on", "dependency", "tree", "convey", "rich", "structural", "information", "which", "is", "proven", "to", "be", "utility", "for", "ABSA." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11 ]

[ "Aspect-based", "Sentiment", "Analysis", "(ABSA)", "aims", "to", "predict", "the", "sentiment", "polarity", "towards", "a", "particular", "aspect", "in", "a", "sentence." ]

[ 11, 12, 12, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Abstract" ]

[ 0 ]

[ "SSEGCN:", "Syntactic", "and", "Semantic", "Enhanced", "Graph", "Convolutional", "Network", "for", "Aspect-based", "Sentiment", "Analysis" ]

[ 1, 1, 2, 2, 2, 2, 2, 2, 0, 11, 12, 12 ]

[ "io/en/latest/." ]

[ 0 ]

[ "16https://pytorch-geometric.readthedocs." ]

[ 0 ]

[ "nese" ]

[ 0 ]

[ "15https://huggingface.co/bert-base-chi" ]

[ 0 ]

[ "ed" ]

[ 0 ]

[ "13https://github.com/huggingface" ]

[ 0 ]

[ "Our", "models", "are", "implemented", "in", "PyTorch", "using", "the", "HuggingFace", "library13", "and", "their", "pretrained" ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "A", "Implementation", "Details" ]

[ 0, 0, 0 ]

[ "All", "models", "use", "the", "Adam", "optimiser", "(Kingma", "and", "Ba,", "2015),", "and", "our", "experiments", "are", "run", "using", "4", "A100", "GPU", "with", "40GB", "Memory." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "For", "the", "user", "tree,", "we", "set", "the", "dimension", "of", "each", "node", "hidden", "features", "as", "256." ]

[ 0, 0, 0, 0, 0, 0, 3, 4, 4, 4, 4, 4, 4, 0, 5 ]

[ "[2e−", "of", "40." ]

[ 5, 0, 5 ]

[ "[1e−", "For", "the", "comment", "chain,", "the", "learning", "rate", "for", "twotier", "transformer", "(comment", "chain)", "is", "tuned", "in", "the", "5]", "with", "the", "maximum", "token", "length", "range" ]

[ 5, 0, 0, 0, 0, 0, 3, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0 ]

[ "Learning", "rate", "5]", "for", "BERT", "and", "5,", "5e−", "is", "tuned", "in", "the", "range", "[1e−", "4]", "for", "GAT", "based", "on", "the", "development", "set." ]

[ 3, 4, 5, 0, 0, 0, 5, 5, 0, 0, 0, 0, 0, 5, 6, 0, 0, 0, 0, 0, 0, 0 ]

[ "[0.5,", "0.6]", "for", "GAT", "and", "0.2", "for", "BERT", "embeddings." ]

[ 5, 5, 0, 0, 0, 5, 0, 0, 0 ]

[ "Graph", "neural", "networks", "are", "implemented", "with", "the", "Geometric16", "package." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "References" ]

[ 0 ]

[ "Lin", "Tian", "is", "supported", "by", "the", "RMIT", "University", "Vice-Chancellor", "PhD", "Scholarship", "(VCPS)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "This", "research", "is", "supported", "partially", "by", "the", "Australian", "Research", "Council", "Discovery", "Project", "DP200101441." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Acknowledgements" ]

[ 0 ]

[ "As", "such,", "we", "recommend", "these", "tools", "to", "be", "used", "as", "an", "aid,", "e.g.", "by", "filtering", "the", "enormous", "volume", "of", "data", "and", "help", "human", "analysts", "to", "narrow", "down", "and", "detect", "harmful", "stories", "on", "social", "media." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "If", "these", "systems", "are", "deployed", "on", "social", "media", "to", "monitor", "user", "posts", "without", "human", "oversight,", "there", "are", "implications", "when", "these", "systems", "misclassify", "(particularly", "in", "the", "false", "positive", "cases)", "and", "users", "are", "wrongfully", "accused", "for", "posting", "misinformation." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "contend", "that", "while", "automatic", "rumour", "detection", "systems", "can", "benefit", "combating", "the", "spread", "of", "misinformation,", "there", "are", "potential", "risks", "to", "them." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "Ethical", "Considerations" ]

[ 0, 0 ]

[ "DUCK", "substantially", "outperforms", "all", "benchmark", "methods", "consistently,", "creating", "a", "new", "state-of-the-art." ]

[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "found", "that", "the", "comment", "network", "contains", "the", "strongest", "signal", "for", "predicting", "rumours,", "and", "social", "relations", "are", "important", "for", "modelling", "the", "user", "network." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "conduct", "extensive", "experiments", "over", "four", "popular", "rumour", "benchmark", "datasets", "to", "evaluate", "DUCK." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1 ]

[ "Our", "approach", "is", "unique", "in", "how", "we", "model", "the", "comment", "as", "a", "graph", "(with", "BERT", "and", "GAT)", "and", "also", "as", "a", "stream", "(with", "transformers)", "and", "the", "user", "networks", "together", "with", "their", "peer", "relations", "(with", "GAT", "and", "GAE)." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "We", "presented", "DUCK,", "a", "social", "media", "rumour", "detection", "approach", "that", "models", "both", "the", "network", "of", "users", "who", "interact", "with", "a", "story", "as", "well", "as", "their", "comments/opinions." ]

[ 0, 0, 1, 0, 0, 0, 11, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "6", "Conclusion" ]

[ 0, 0 ]

[ "This", "finding", "is", "similar", "to", "what", "we", "saw", "earlier,", "where", "systems", "like", "stance-BERT", "and", "BiGCN", "that", "use", "comments", "tend", "to", "perform", "better." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]

[ "Results", "suggest", "that", "comment", "tree", "has", "the", "largest", "impact,", "followed", "by", "comment", "chain", "as", "they", "produce", "the", "largest", "performance", "drop", "when", "removed." ]

[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

[ "To", "understand", "the", "impact", "of", "each", "module", "in", "DUCK,", "we", "present", "variants", "where", "we", "remove", "one", "CT", "means", "comment", "tree", "remodule,", "e.g.", "DUCK", "moved." ]

[ 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 ]