# coding=utf-8 # Copyright 2020 HuggingFace Datasets Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """The lingual SemEval2014 Task5 Reviews Corpus""" import datasets _CITATION = """\ @article{2014SemEval, title={SemEval-2014 Task 4: Aspect Based Sentiment Analysis}, author={ Pontiki, M. and D Galanis and Pavlopoulos, J. and Papageorgiou, H. and Manandhar, S. }, journal={Proceedings of International Workshop on Semantic Evaluation at}, year={2014}, } """ _LICENSE = """\ Please click on the homepage URL for license details. """ _DESCRIPTION = """\ A collection of SemEval2014 specifically designed to aid research in Aspect Based Sentiment Analysis. """ _CONFIG = [ # restaurants domain "restaurants", # laptops domain "laptops", ] _VERSION = "0.0.1" _HOMEPAGE_URL = "https://alt.qcri.org/semeval2014/task4/index.php?id=data-and-tools" _DOWNLOAD_URL = "https://raw.githubusercontent.com/YaxinCui/ABSADataset/main/SemEval2014Task4/{split}/{domain}_{split}.xml" class SemEval2014Task4NLTKConfig(datasets.BuilderConfig): """BuilderConfig for SemEval2014Config.""" def __init__(self, _CONFIG, **kwargs): super(SemEval2014Task4NLTKConfig, self).__init__(version=datasets.Version(_VERSION, ""), **kwargs), self.configs = _CONFIG class SemEval2014Task4NLTK(datasets.GeneratorBasedBuilder): """The lingual Amazon Reviews Corpus""" BUILDER_CONFIGS = [ SemEval2014Task4NLTKConfig( name="All", _CONFIG=_CONFIG, description="A collection of SemEval2014 specifically designed to aid research in lingual Aspect Based Sentiment Analysis.", ) ] + [ SemEval2014Task4NLTKConfig( name=config, _CONFIG=[config], description=f"{config} of SemEval2014 specifically designed to aid research in Aspect Based Sentiment Analysis", ) for config in _CONFIG ] BUILDER_CONFIG_CLASS = SemEval2014Task4NLTKConfig DEFAULT_CONFIG_NAME = "All" def _info(self): return datasets.DatasetInfo( description=_DESCRIPTION, features=datasets.Features( {'text': datasets.Value(dtype='string'), 'aspectTerms': [ {'from': datasets.Value(dtype='string'), 'polarity': datasets.Value(dtype='string'), 'term': datasets.Value(dtype='string'), 'to': datasets.Value(dtype='string')} ], 'tokens': [datasets.Value(dtype='string')], 'ATESP_BIEOS_tags': [datasets.Value(dtype='string')], 'ATESP_BIO_tags': [datasets.Value(dtype='string')], 'ATE_BIEOS_tags': [datasets.Value(dtype='string')], 'ATE_BIO_tags': [datasets.Value(dtype='string')], 'aspectCategories': [ {'category': datasets.Value(dtype='string'), 'polarity': datasets.Value(dtype='string')} ], 'domain': datasets.Value(dtype='string'), 'sentenceId': datasets.Value(dtype='string') } ), supervised_keys=None, license=_LICENSE, homepage=_HOMEPAGE_URL, citation=_CITATION, ) def _split_generators(self, dl_manager): train_urls = [_DOWNLOAD_URL.format(split="train", domain=config) for config in self.config.configs] dev_urls = [_DOWNLOAD_URL.format(split="trial", domain=config) for config in self.config.configs] test_urls = [_DOWNLOAD_URL.format(split="test", domain=config) for config in self.config.configs] train_paths = dl_manager.download_and_extract(train_urls) dev_paths = dl_manager.download_and_extract(dev_urls) test_paths = dl_manager.download_and_extract(test_urls) return [ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"file_paths": train_paths, "domain_list": self.config.configs}), datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"file_paths": dev_paths, "domain_list": self.config.configs}), datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"file_paths": test_paths, "domain_list": self.config.configs}), ] def _generate_examples(self, file_paths, domain_list): row_count = 0 assert len(file_paths)==len(domain_list) for i in range(len(file_paths)): file_path, domain = file_paths[i], domain_list[i] semEvalDataset = SemEvalXMLDataset(file_path, domain) for example in semEvalDataset.SentenceWithOpinions: yield row_count, example row_count += 1 from xml.dom.minidom import parse class SemEvalXMLDataset(): def __init__(self, file_name, domain): # 获得SentenceWithOpinions，一个List包含(reviewId, sentenceId, text, Opinions) self.SentenceWithOpinions = [] self.xml_path = file_name self.sentenceXmlList = parse(open(self.xml_path)).getElementsByTagName('sentence') for sentenceXml in self.sentenceXmlList: sentenceId = sentenceXml.getAttribute("id") if len(sentenceXml.getElementsByTagName("text")[0].childNodes) < 1: # skip no reviews part continue text = sentenceXml.getElementsByTagName("text")[0].childNodes[0].nodeValue aspectTermsXLMList = sentenceXml.getElementsByTagName("aspectTerm") aspectTerms = [] for opinionXml in aspectTermsXLMList: # some text maybe have no opinion term = opinionXml.getAttribute("term") polarity = opinionXml.getAttribute("polarity") from_ = opinionXml.getAttribute("from") to = opinionXml.getAttribute("to") aspectTermDict = { "term": term, "polarity": polarity, "from": from_, "to": to } aspectTerms.append(aspectTermDict) # 从小到大排序 aspectTerms.sort(key=lambda x: int(x["from"])) aspectCategoriesXmlList = sentenceXml.getElementsByTagName("aspectCategory") aspectCategories = [] for aspectCategoryXml in aspectCategoriesXmlList: category = aspectCategoryXml.getAttribute("category") polarity = aspectCategoryXml.getAttribute("polarity") aspectCategoryDict = { "category": category, "polarity": polarity } aspectCategories.append(aspectCategoryDict) example = { "text": text, "aspectTerms": aspectTerms, "aspectCategories": aspectCategories, "domain": domain, "sentenceId": sentenceId } example = addTokenAndLabel(example) self.SentenceWithOpinions.append(example) import nltk def clearOpinion(example): opinions = example['aspectTerms'] skipNullOpinions = [] # 去掉NULL的opinion for opinion in opinions: targetKey = 'term' target = opinion[targetKey] from_ = opinion['from'] to = opinion['to'] # skill NULL if target.lower() == 'null' or target == '' or from_ == to: continue skipNullOpinions.append(opinion) # delete repeate Opinions skipNullOpinions.sort(key=lambda x: int(x['from'])) # 从小到大排序 UniOpinions = [] for opinion in skipNullOpinions: if len(UniOpinions) < 1: UniOpinions.append(opinion) else: if opinion['from'] != UniOpinions[-1]['from'] and opinion['to'] != UniOpinions[-1]['to']: UniOpinions.append(opinion) return UniOpinions def addTokenAndLabel(example): tokens = [] labels = [] text = example['text'] # UniOpinions = clearOpinion(example) UniOpinions = example['aspectTerms'] text_begin = 0 for aspect in UniOpinions: polarity = aspect['polarity'][:3].upper() pre_O_tokens = nltk.word_tokenize(text[text_begin: int(aspect['from'])]) tokens.extend(pre_O_tokens) labels.extend(['O']*len(pre_O_tokens)) BIES_tokens = nltk.word_tokenize(text[int(aspect['from']): int(aspect['to'])]) tokens.extend(BIES_tokens) assert len(BIES_tokens) > 0, print('error in BIES_tokens length') if len(BIES_tokens)==1: labels.append('S-'+polarity) elif len(BIES_tokens)==2: labels.append('B-'+polarity) labels.append('E-'+polarity) else: labels.append('B-'+polarity) labels.extend(['I-'+polarity]*(len(BIES_tokens)-2)) labels.append('E-'+polarity) text_begin = int(aspect['to']) pre_O_tokens = nltk.word_tokenize(text[text_begin: ]) labels.extend(['O']*len(pre_O_tokens)) tokens.extend(pre_O_tokens) example['tokens'] = tokens example['ATESP_BIEOS_tags'] = labels ATESP_BIO_labels = [] for label in labels: ATESP_BIO_labels.append(label.replace('E-', 'I-').replace('S-', 'B-')) example['ATESP_BIO_tags'] = ATESP_BIO_labels ATE_BIEOS_labels = [] for label in labels: ATE_BIEOS_labels.append(label[0]) example['ATE_BIEOS_tags'] = ATE_BIEOS_labels ATE_BIO_labels = [] for label in ATESP_BIO_labels: ATE_BIO_labels.append(label[0]) example['ATE_BIO_tags'] = ATE_BIO_labels return example