from flair.data import Sentence from flair.embeddings import TransformerWordEmbeddings from pathlib import Path from typing import List def prepare_ajmc_corpus( file_in: Path, file_out: Path, eos_marker: str, document_separator: str, add_document_separator: bool ): with open(file_in, "rt") as f_p: lines = f_p.readlines() with open(file_out, "wt") as f_out: # Add missing newline after header f_out.write(lines[0] + "\n") for line in lines[1:]: if line.startswith(" \t"): # Workaround for empty tokens continue line = line.strip() # HIPE-2022 late pre-submission fix: # Our hmBERT model has never seen Fraktur, so we replace long s line = line.replace("ſ", "s") # Add "real" document marker if add_document_separator and line.startswith(document_separator): f_out.write("-DOCSTART- O\n\n") f_out.write(line + "\n") if eos_marker in line: f_out.write("\n") print("Special preprocessing for AJMC has finished!") def prepare_clef_2020_corpus( file_in: Path, file_out: Path, eos_marker: str, document_separator: str, add_document_separator: bool ): with open(file_in, "rt") as f_p: original_lines = f_p.readlines() lines = [] # Add missing newline after header lines.append(original_lines[0]) for line in original_lines[1:]: if line.startswith(" \t"): # Workaround for empty tokens continue line = line.strip() # Add "real" document marker if add_document_separator and line.startswith(document_separator): lines.append("-DOCSTART- O") lines.append("") lines.append(line) if eos_marker in line: lines.append("") # Now here comes the de-hyphenation part ;) word_seperator = "¬" for index, line in enumerate(lines): if line.startswith("#"): continue if line.startswith(word_seperator): continue if not line: continue prev_line = lines[index - 1] prev_prev_line = lines[index - 2] if not prev_line.startswith(word_seperator): continue # Example: # Po <- prev_prev_line # ¬ <- prev_line # len <- current_line # # will be de-hyphenated to: # # Polen Dehyphenated-3 # # ¬ # # len suffix = line.split("\t")[0] prev_prev_line_splitted = lines[index - 2].split("\t") prev_prev_line_splitted[0] += suffix prev_line_splitted = lines[index - 1].split("\t") prev_line_splitted[0] = "#" + prev_line_splitted[0] prev_line_splitted[-1] += "|Commented" current_line_splitted = line.split("\t") current_line_splitted[0] = "#" + current_line_splitted[0] current_line_splitted[-1] += "|Commented" # Add some meta information about suffix length # Later, it is possible to re-construct original token and suffix prev_prev_line_splitted[9] += f"|Dehyphenated-{len(suffix)}" lines[index - 2] = "\t".join(prev_prev_line_splitted) lines[index - 1] = "\t".join(prev_line_splitted) lines[index] = "\t".join(current_line_splitted) # Post-Processing I for index, line in enumerate(lines): if not line: continue if not line.startswith(word_seperator): continue # oh noooo current_line_splitted = line.split("\t") current_line_splitted[0] = "#" + current_line_splitted[0] current_line_splitted[-1] += "|Commented" lines[index] = "\t".join(current_line_splitted) # Post-Processing II # Beautify: _|Commented –> Commented for index, line in enumerate(lines): if not line: continue if not line.startswith("#"): continue current_line_splitted = line.split("\t") if current_line_splitted[-1] == "_|Commented": current_line_splitted[-1] = "Commented" lines[index] = "\t".join(current_line_splitted) # Finally, save it! with open(file_out, "wt") as f_out: for line in lines: f_out.write(line + "\n") def prepare_newseye_fi_sv_corpus( file_in: Path, file_out: Path, eos_marker: str, document_separator: str, add_document_separator: bool ): with open(file_in, "rt") as f_p: original_lines = f_p.readlines() lines = [] # Add missing newline after header lines.append(original_lines[0]) for line in original_lines[1:]: if line.startswith(" \t"): # Workaround for empty tokens continue line = line.strip() # Add "real" document marker if add_document_separator and line.startswith(document_separator): lines.append("-DOCSTART- O") lines.append("") lines.append(line) if eos_marker in line: lines.append("") # Now here comes the de-hyphenation part # And we want to avoid matching "-DOCSTART-" lines here, so append a tab word_seperator = "-\t" for index, line in enumerate(lines): if line.startswith("#"): continue if line.startswith(word_seperator): continue if not line: continue prev_line = lines[index - 1] prev_prev_line = lines[index - 2] if not prev_line.startswith(word_seperator): continue # Example: # Po NoSpaceAfter <- prev_prev_line # - <- prev_line # len <- current_line # # will be de-hyphenated to: # # Polen Dehyphenated-3 # # - # # len # # It is really important, that "NoSpaceAfter" in the previous # line before hyphenation character! Otherwise, it is no real # hyphenation! if not "NoSpaceAfter" in prev_line: continue if not prev_prev_line: continue suffix = line.split("\t")[0] prev_prev_line_splitted = lines[index - 2].split("\t") prev_prev_line_splitted[0] += suffix prev_line_splitted = lines[index - 1].split("\t") prev_line_splitted[0] = "# " + prev_line_splitted[0] prev_line_splitted[-1] += "|Commented" current_line_splitted = line.split("\t") current_line_splitted[0] = "# " + current_line_splitted[0] current_line_splitted[-1] += "|Commented" # Add some meta information about suffix length # Later, it is possible to re-construct original token and suffix prev_prev_line_splitted[9] += f"|Dehyphenated-{len(suffix)}" lines[index - 2] = "\t".join(prev_prev_line_splitted) lines[index - 1] = "\t".join(prev_line_splitted) lines[index] = "\t".join(current_line_splitted) # Post-Processing I for index, line in enumerate(lines): if not line: continue if not line.startswith(word_seperator): continue # oh noooo current_line_splitted = line.split("\t") current_line_splitted[0] = "# " + current_line_splitted[0] current_line_splitted[-1] += "|Commented" lines[index] = "\t".join(current_line_splitted) # Post-Processing II # Beautify: _|Commented –> Commented for index, line in enumerate(lines): if not line: continue if not line.startswith("#"): continue current_line_splitted = line.split("\t") if current_line_splitted[-1] == "_|Commented": current_line_splitted[-1] = "Commented" lines[index] = "\t".join(current_line_splitted) # Finally, save it! with open(file_out, "wt") as f_out: for line in lines: f_out.write(line + "\n") def prepare_newseye_de_fr_corpus( file_in: Path, file_out: Path, eos_marker: str, document_separator: str, add_document_separator: bool ): with open(file_in, "rt") as f_p: original_lines = f_p.readlines() lines = [] # Add missing newline after header lines.append(original_lines[0]) for line in original_lines[1:]: if line.startswith(" \t"): # Workaround for empty tokens continue line = line.strip() # Add "real" document marker if add_document_separator and line.startswith(document_separator): lines.append("-DOCSTART- O") lines.append("") lines.append(line) if eos_marker in line: lines.append("") # Now here comes the de-hyphenation part ;) word_seperator = "¬" for index, line in enumerate(lines): if line.startswith("#"): continue if not line: continue last_line = lines[index - 1] last_line_splitted = last_line.split("\t") if not last_line_splitted[0].endswith(word_seperator): continue # The following example # # den O O O null null SpaceAfter # Ver¬ B-LOC O O null n <- last_line # einigten I-LOC O O null n SpaceAfter <- current_line # Staaten I-LOC O O null n # . O O O null null # # will be transformed to: # # den O O O null null SpaceAfter # Vereinigten B-LOC O O null n |Normalized-8 # #einigten I-LOC O O null n SpaceAfter|Commented # Staaten I-LOC O O null n # . O O O null null suffix = last_line.split("\t")[0].replace(word_seperator, "") # Will be "Ver" prefix_length = len(line.split("\t")[0]) # Override last_line: # Ver¬ will be transformed to Vereinigten with normalized information at the end last_line_splitted[0] = suffix + line.split("\t")[0] last_line_splitted[9] += f"|Dehyphenated-{prefix_length}" current_line_splitted = line.split("\t") current_line_splitted[0] = "# " + current_line_splitted[0] current_line_splitted[-1] += "|Commented" lines[index - 1] = "\t".join(last_line_splitted) lines[index] = "\t".join(current_line_splitted) # Post-Processing I # Beautify: _|Commented –> Commented for index, line in enumerate(lines): if not line: continue if not line.startswith("#"): continue current_line_splitted = line.split("\t") if current_line_splitted[-1] == "_|Commented": current_line_splitted[-1] = "Commented" lines[index] = "\t".join(current_line_splitted) # Finally, save it! with open(file_out, "wt") as f_out: for line in lines: f_out.write(line + "\n") print("Special preprocessing for German/French NewsEye dataset has finished!")