import datasets import hashlib import functools import glob import os import json import openpyxl _DESCRIPTION = """ """ _CITATION = """ """ _VERSION = datasets.Version('1.0.0', "") _DATASET_ROOT = { # folder 'common_squad': 'AIHub/common', 'paper_summary': 'AIHub/논문자료 요약', 'paper_patent_section': 'AIHub/논문자료 요약', 'paper_patent_total': 'AIHub/논문자료 요약', 'document_summary_law': 'AIHub/문서요약 텍스트', 'document_summary_editorial': 'AIHub/문서요약 텍스트', 'emotional_talk': 'AIHub/감성대화', } _PARAGRAPHS_SEQUENCE = datasets.Sequence({ # common_squad 'qas': datasets.Sequence({ 'question': datasets.Value("string"), 'answers': datasets.Sequence({ 'answer_start': datasets.Value("int32"), 'text': datasets.Value("string"), }), 'id': datasets.Value("string"), }), 'context': datasets.Value("string"), }) _COMMON_SQUAD_FEATURE = datasets.Features({ # common_squad 'id': datasets.Value("int32"), 'paragraphs': _PARAGRAPHS_SEQUENCE, 'title': datasets.Value("string"), }) _SUMMARY_FEATRUE = datasets.Sequence({ # paper_summary 'orginal_text': datasets.Value("string"), 'summary_text': datasets.Value("string"), }) _PAPER_SUMMARY_FEATURE = datasets.Features({ # paper_summary 'id': datasets.Value("int32"), 'doc_type': datasets.Value("string"), 'doc_id': datasets.Value("string"), 'title': datasets.Value("string"), 'date': datasets.Value("string"), 'reg_no': datasets.Value("string"), 'ipc': datasets.Value("string"), 'issued_by': datasets.Value("string"), 'author': datasets.Value("string"), 'summary_entire': _SUMMARY_FEATRUE, 'summary_section': _SUMMARY_FEATRUE, }) _PAPER_PATENT_SECTION_FEATURE = datasets.Features({ # paper_patent_section 'id': datasets.Value("int32"), 'doc_type': datasets.Value("string"), 'doc_id': datasets.Value("string"), 'title': datasets.Value("string"), 'date': datasets.Value("string"), 'reg_no': datasets.Value("string"), 'ipc': datasets.Value("string"), 'author': datasets.Value("string"), 'summary_section': _SUMMARY_FEATRUE, }) _PAPER_PATENT_TOTAL_FEATURE = datasets.Features({ # paper_patent_total 'id': datasets.Value("int32"), 'doc_type': datasets.Value("string"), 'doc_id': datasets.Value("string"), 'title': datasets.Value("string"), 'date': datasets.Value("string"), 'reg_no': datasets.Value("string"), 'ipc': datasets.Value("string"), 'author': datasets.Value("string"), 'summary_entire': _SUMMARY_FEATRUE, 'summary_section': _SUMMARY_FEATRUE, }) # document_summary_law, document_summary_editorial, document_summary_newspaper _TEXT_FEATURE = datasets.Sequence({ 'index': datasets.Value("int32"), 'sentence': datasets.Value("string"), 'highlight_indices': datasets.Value("string"), }) # document_summary_law, document_summary_editorial, document_summary_newspaper _DOCUMENT_QUALITY_SCORES = datasets.Features({ 'readable': datasets.Value("int32"), 'accurate': datasets.Value("int32"), 'informative': datasets.Value("int32"), 'trustworthy': datasets.Value("int32"), }) _DOCUMENT_SUMMARY_LAW_FEATURE = datasets.Features({ # document_summary_law 'id': datasets.Value("string"), 'category': datasets.Value("string"), 'size': datasets.Value("string"), 'char_count': datasets.Value("int32"), 'publish_date': datasets.Value("string"), 'title': datasets.Value("string"), 'text': _TEXT_FEATURE, 'annotator_id': datasets.Value("int32"), 'document_quality_scores': _DOCUMENT_QUALITY_SCORES, 'extractive': datasets.Sequence(datasets.Value("int32")), 'abstractive': datasets.Sequence(datasets.Value("string")), }) # document_summary_editorial, document_summary_newspaper _DOCUMENT_SUMMARY_FEATURE = datasets.Features({ 'id': datasets.Value("string"), 'category': datasets.Value("string"), 'media_type': datasets.Value("string"), 'media_sub_type': datasets.Value("string"), 'media_name': datasets.Value("string"), 'size': datasets.Value("string"), 'char_count': datasets.Value("string"), 'publish_date': datasets.Value("string"), 'title': datasets.Value("string"), 'text': _TEXT_FEATURE, 'annotator_id': datasets.Value("int32"), 'document_quality_scores': _DOCUMENT_QUALITY_SCORES, 'extractive': datasets.Sequence(datasets.Value("int32")), 'abstractive': datasets.Sequence(datasets.Value("string")), }) _PERSONA_FEATURE = datasets.Features({ # emotional_talk 'persona-id': datasets.Value("string"), 'human': datasets.Sequence( datasets.Value("string"), ), 'computer': datasets.Sequence( datasets.Value("string"), ), }) _EMOTION_FEATURE = datasets.Features({ # emotional_talk 'emotion-id': datasets.Value("string"), 'type': datasets.Value("string"), 'situation': datasets.Sequence( datasets.Value("string"), ), }) _PROFILE_FEATURE = datasets.Features({ # emotional_talk 'persona-id': datasets.Value("string"), 'persona': _PERSONA_FEATURE, 'emotion': _EMOTION_FEATURE, }) _CONTENT_FEATURE = datasets.Features({ # emotional_talk 'HS01': datasets.Value("string"), 'SS01': datasets.Value("string"), 'HS02': datasets.Value("string"), 'SS02': datasets.Value("string"), 'HS03': datasets.Value("string"), 'SS03': datasets.Value("string"), }) _TALK_FEATURE = datasets.Features({ # emotional_talk 'id': datasets.Features({ 'profile-id': datasets.Value("string"), 'talk-id': datasets.Value("string"), }), 'content': _CONTENT_FEATURE, }) _EMOTIONAL_TALK_FEATURE = datasets.Features({ # emotional_talk 'id': datasets.Value("int32"), 'profile': _PROFILE_FEATURE, 'talk': _TALK_FEATURE, }) def _parsing_common_squad(file_path): # common_squad with open(file_path, mode='r') as f: obj = json.loads(f.read()) for id, sample in enumerate(obj['data']): _id = id _paragraphs = sample['paragraphs'] _title = sample['title'] yield _id, { 'id': _id, 'paragraphs': _paragraphs, 'title': _title, } def _parsing_paper_summary(file_path): # paper_summary with open(file_path, mode='r') as f: obj = json.loads(f.read()) for id, sample in enumerate(obj['data']): _id = id _doc_type = sample['doc_type'] _doc_id = sample['doc_id'] _title = sample['title'] _date = sample['date'] _reg_no = sample['reg_no'] _ipc = sample['reg_no'] _issued_by = sample['issued_by'] _author = sample['author'] _summary_entire = sample['summary_entire'] _summary_section = sample['summary_section'] yield _id, { 'id': _id, 'doc_type': _doc_type, 'doc_id': _doc_id, 'title': _title, 'date': _date, 'reg_no': _reg_no, 'ipc': _ipc, 'issued_by': _issued_by, 'author': _author, 'summary_entire': _summary_entire, 'summary_section': _summary_section, } def _parsing_paper_patent_section(file_path): # paper_patent_section with open(file_path, mode='r') as f: obj = json.loads(f.read()) for id, sample in enumerate(obj['data']): _id = id _doc_type = sample['doc_type'] _doc_id = sample['doc_id'] _title = sample['title'] _date = sample['date'] _reg_no = sample['reg_no'] _ipc = sample['reg_no'] _author = sample['author'] _summary_section = sample['summary_section'] yield _id, { 'id': _id, 'doc_type': _doc_type, 'doc_id': _doc_id, 'title': _title, 'date': _date, 'reg_no': _reg_no, 'ipc': _ipc, 'author': _author, 'summary_section': _summary_section, } def _parsing_paper_patent_total(file_path): # paper_patent_total with open(file_path, mode='r') as f: obj = json.loads(f.read()) for id, sample in enumerate(obj['data']): _id = id _doc_type = sample['doc_type'] _doc_id = sample['doc_id'] _title = sample['title'] _date = sample['date'] _reg_no = sample['reg_no'] _ipc = sample['reg_no'] _author = sample['author'] _summary_section = sample['summary_section'] yield _id, { 'id': _id, 'doc_type': _doc_type, 'doc_id': _doc_id, 'title': _title, 'date': _date, 'reg_no': _reg_no, 'ipc': _ipc, 'author': _author, 'summary_entire': _summary_section, 'summary_section': _summary_section, } def _parsing_document_summary_law(file_path): # document_summary_law with open(file_path, mode='r') as f: obj = json.loads(f.read()) for sample in obj: _id = sample['id'] _category = sample['category'] _size = sample['size'] _char_count = sample['char_count'] _publish_date = sample['publish_date'] _title = sample['title'] _text = sample['text'] _annotator_id = sample['annotator_id'] _document_quality_scores = sample['document_quality_scores'] _extractive = sample['extractive'] _abstractive = sample['abstractive'] yield _id, { 'id': _id, 'category': _category, 'size': _size, 'char_count': _char_count, 'publish_date': _publish_date, 'title': _title, 'text': _text, 'annotator_id': _annotator_id, 'document_quality_scores': _document_quality_scores, 'extractive': _extractive, 'abstractive': _abstractive, } # document_summary_editorial, document_summary_newspaper def _parsing_document_summary(file_path): with open(file_path, mode='r') as f: obj = json.loads(f.read()) for sample in obj: _id = sample['id'] _category = sample['category'] _media_type = sample['media_type'] _media_sub_type = sample['media_sub_type'] _media_name = sample['media_name'] _size = sample['size'] _char_count = str(sample['char_count']) _publish_date = sample['publish_date'] _title = sample['title'] _text = sample['text'] _annotator_id = sample['annotator_id'] _document_quality_scores = sample['document_quality_scores'] _extractive = sample['extractive'] _abstractive = sample['abstractive'] yield _id, { 'id': _id, 'category': _category, 'media_type': _media_type, 'media_sub_type': _media_sub_type, 'media_name': _media_name, 'size': _size, 'char_count': _char_count, 'publish_date': _publish_date, 'title': _title, 'text': _text, 'annotator_id': _annotator_id, 'document_quality_scores': _document_quality_scores, 'extractive': _extractive, 'abstractive': _abstractive, } def _parsing_emotional_talk(file_path): # emotional talk with open(file_path, mode='r') as f: obj = json.loads(f.read()) for id, sample in enumerate(obj): _id = id _profile = sample['profile'] _talk = sample['talk'] yield _id, { 'id':_id, 'profile': _profile, 'talk': _talk, } def _hash_text(text): return hashlib.md5(text.encode("utf-8")).hexdigest() def _filter_fn_hash_id(uid, split_fn): hash_id = _hash_text(str(uid)) val = int(hash_id, 16) return split_fn(val) _DEFAULT_RAW_CORPUS_SPLIT = { 'source': [datasets.Split.TRAIN], 'split': { datasets.Split.TRAIN: lambda x: x % 1000 > 0, datasets.Split.VALIDATION: lambda x: x % 1000 == 0, }} _DEFAULT_DOWNSTREAMTASK_CORPUS_SPLIT = { 'source': [datasets.Split.TRAIN], 'split': { datasets.Split.TRAIN: lambda x: x % 10 > 1, datasets.Split.VALIDATION: lambda x: x % 10 == 0, datasets.Split.TEST: lambda x: x % 10 == 1, }} class AIHubConfig(datasets.BuilderConfig): def __init__(self, name, data_root, feature, data_sp_path, reading_fn, parsing_fn, additional_data_root=None, homepage='https://aihub.or.kr/', split_fn=None, metadata=None, **kwargs): super(AIHubConfig, self).__init__( name=name, # error...? version=_VERSION, **kwargs ) self.data_root = data_root self.feature = feature self.data_sp_path = data_sp_path self.reading_fn = reading_fn self.parsing_fn = parsing_fn self.additional_data_root = additional_data_root self.homepage = homepage self.split_fn = split_fn self.metadata = metadata class AIHub(datasets.GeneratorBasedBuilder): """DatasetBuilder for AIHub dataset.""" RELEASE_NOTES = { '1.0.0': 'Initial release.', } BUILDER_CONFIGS = [ AIHubConfig( name='common.squad.v1.0', data_root=_DATASET_ROOT['common_squad'], feature=_COMMON_SQUAD_FEATURE, data_sp_path={datasets.Split.TRAIN: ['nia_common_02_squad_질문, 답변, 제시문 말뭉치/ko_wiki_v1_squad.json']}, reading_fn=_parsing_common_squad, parsing_fn=lambda x:x, ), AIHubConfig( name='common.squad.v1.0.split', data_root=_DATASET_ROOT['common_squad'], feature=_COMMON_SQUAD_FEATURE, data_sp_path={datasets.Split.TRAIN: ['nia_common_02_squad_질문, 답변, 제시문 말뭉치/ko_wiki_v1_squad.json']}, reading_fn=_parsing_common_squad, parsing_fn=lambda x:x, split_fn=_DEFAULT_RAW_CORPUS_SPLIT, ), AIHubConfig( name='paper.summary.v1.0.split', data_root=_DATASET_ROOT['paper_summary'], feature=_PAPER_SUMMARY_FEATURE, data_sp_path={datasets.Split.TRAIN: ['Training/training_논문/*.json'], datasets.Split.VALIDATION: ['Validation/validation_논문/*.json']}, reading_fn=_parsing_paper_summary, parsing_fn=lambda x:x, ), AIHubConfig( name='paper.patent.section.v1.0.split', data_root=_DATASET_ROOT['paper_patent_section'], feature=_PAPER_PATENT_SECTION_FEATURE, data_sp_path={datasets.Split.TRAIN: ['Training/training_특허섹션만/*.json'], datasets.Split.VALIDATION: ['Validation/validation_특허섹션만/*.json']}, reading_fn=_parsing_paper_patent_section, parsing_fn=lambda x:x, ), AIHubConfig( name='paper.patent.total.v1.0.split', data_root=_DATASET_ROOT['paper_patent_total'], feature=_PAPER_PATENT_TOTAL_FEATURE, data_sp_path={datasets.Split.TRAIN: ['Training/training_특허전체/*.json'], datasets.Split.VALIDATION: ['Validation/validation_특허전체/*.json']}, reading_fn=_parsing_paper_patent_total, parsing_fn=lambda x:x, ), AIHubConfig( name='document.summary.law.v1.0.split', data_root=_DATASET_ROOT['document_summary_law'], feature=_DOCUMENT_SUMMARY_LAW_FEATURE, data_sp_path={datasets.Split.TRAIN: ['Training/train_법률_data/법률문서/train_original.json'], datasets.Split.VALIDATION: ['Validation/valid_법률_data/법률문서/dev_original.json']}, reading_fn=_parsing_document_summary_law, parsing_fn=lambda x:x, ), AIHubConfig( name='document.summary.editorial.v1.0.split', data_root=_DATASET_ROOT['document_summary_editorial'], feature=_DOCUMENT_SUMMARY_FEATURE, data_sp_path={datasets.Split.TRAIN: ['Training/train_사설잡지_data/train_original.json'], datasets.Split.VALIDATION: ['Validation/valid_사설잡지_data/dev_original.json']}, reading_fn=_parsing_document_summary, parsing_fn=lambda x:x, ), AIHubConfig( name='emotional.talk.v1.0.split', data_root=_DATASET_ROOT['emotional_talk'], feature=_EMOTIONAL_TALK_FEATURE, data_sp_path={datasets.Split.TRAIN: ['Training/감성대화말뭉치(최종데이터)_Training/감성대화말뭉치(최종데이터)_Training.json'], datasets.Split.VALIDATION: ['Validation/감성대화말뭉치(최종데이터)_Validation/감성대화말뭉치(최종데이터)_Validation.json']}, reading_fn=_parsing_emotional_talk, parsing_fn=lambda x:x, ), ] MANUAL_DOWNLOAD_INSTRUCTIONS = """ For the NIKL, you must manually download NIKL data from https://aihub.or.kr/ and extract it under the proper location. all the data have to located under manual_dir/AIHub. This is dataset and path pairs. (all the paths are case-sensitive!) ============================================ COMMON_SQUAD(v1.0): manual_dir/AIHub/common/nia_common_02_squad_질문, 답변, 제시문 말뭉치/ko_wiki_v1_squad.json PAPER_SUMMARY(v1.0): manual_dir/AIHub/논문자료 요약/Training/training_논문/*.json manual_dir/AIHub/논문자료 요약/Validation/validation_논문/*.json PAPER_PATENT_SECTION(v1.0): manual_dir/AIHub/논문자료 요약/Training/training_특허섹션만/*.json manual_dir/AIHub/논문자료 요약/Validation/validation_특허섹션만/*.json PAPER_PATENT_TOTAL(v1.0): manual_dir/AIHub/논문자료 요약/Training/training_특허전체/*.json manual_dir/AIHub/논문자료 요약/Validation/validation_특허전체/*.json DOCUMENT_SUMMARY_LAW(v1.0): manual_dir/AIHub/문서요약 텍스트/1.Training/train_법률_data/법률문서/train_original.json manual_dir/AIHub/문서요약 텍스트/2.Validation/valid_법률_data/법률문서/dev_original.json DOCUMENT_SUMMARY_EDITORIAL(v1.0): manual_dir/AIHub/문서요약 텍스트/Training/train_사설잡지_data/train_original.json manual_dir/AIHub/문서요약 텍스트/Validation/valid_사설잡지_data/dev_original.json EMOTIONAL_TALK(v1.0): manual_dir/AIHub/감성대화/Training/감성대화말뭉치(최종데이터)_Training/감성대화말뭉치(최종데이터)_Training.json manual_dir/AIHub/감성대화/Validation/감성대화말뭉치(최종데이터)_Validation/감성대화말뭉치(최종데이터)_Validation.json ============================================ """ def _info(self) -> datasets.DatasetInfo: """Returns the dataset metadata.""" return datasets.DatasetInfo( description=_DESCRIPTION, features=self.config.feature, homepage=self.config.homepage, citation=_CITATION, ) def _split_generators(self, dl_manager: datasets.DownloadManager): """Returns SplitGenerators.""" path_kv = {} for k, v in self.config.data_sp_path.items(): path_list = [] for vv in v: path_list.extend(glob.glob(os.path.join( dl_manager.manual_dir, self.config.data_root, vv))) path_kv[k] = path_list path_list = [] for _, v in path_kv.items(): if len(v) == 0: raise AssertionError("For the AIHub dataset, you must manually download and extract dataset under {0}/{1}.".format( dl_manager.manual_dir, self.config.data_root )) if self.config.split_fn is not None: in_files = [] for sp_s_key in self.config.split_fn['source']: in_files.extend(path_kv[sp_s_key]) split_fn_kv = self.config.split_fn['split'] return [ datasets.SplitGenerator(name=k, gen_kwargs={'path_list': in_files, 'split_fn': v}) for k, v in split_fn_kv.items() ] return [ datasets.SplitGenerator(name=k, gen_kwargs={'path_list': v}) for k, v in path_kv.items() ] def _generate_examples(self, path_list, split_fn=None): """Yields examples.""" if split_fn is not None: split_filter = functools.partial(_filter_fn_hash_id, split_fn=split_fn) _hash_set = set() for file_path in path_list: try: for example in iter(self.config.reading_fn(file_path)): uid, ex = self.config.parsing_fn(example) if split_fn is not None: if not split_filter(str(uid)): continue hash_id = _hash_text(str(uid)) if hash_id not in _hash_set: _hash_set.add(hash_id) yield uid, ex except Exception as e: print(e)