评论对象抽取是情感分析的重要研究内容。基于语义词典,从评论对象的类别视角出发,运用语义相似度和相关度计算方法,该文提出用于评价对象抽取的七种新的语义特征。评价对象和评价词之间通常存在句法依存关系,并且评价词往往带有情感倾向,将句法依存分析和评价词识别结合,提出句法情感依存特征抽取方法,忽略无情感词和微情感词的句法依存关系,提高评价对象抽取的准确率。使用条件随机场模型,在SEMEVAL比赛的三个领域数据集上进行实验,新的语义特征和句法情感依存特征组合的F1分数比SEMEVAL比赛限制性系统最好成绩平均高3.78％,比非限制性系统最好成绩平均高2％,证明了所提特征的有效性。
Abstract
Opinion target extraction is an important task of sentiment analysis. Based on a semantic dictionary,this paper proposes seven semantic features of opinion targets in relation to their categories via the semantic similarity and relevance computation. Since there are exist syntactic dependency between the opinion targets and opinion words, this paper further presents the extraction method of sentiment syntactic dependency features,ignoring those objective words or micro sentiment words to improve the accuracy. In the experiments on three datasets of SEMEVAL,the combination of new semantic features and sentiment syntactic dependency features enable the CRFs a F1 score of 3.78 points higher than the SEMEVAL's best score for constrained systems,and 2 points higher for unconstrained systems.
关键词
评价对象抽取 /
条件随机场 /
语义特征 /
句法依存关系
{{custom_keyword}} /
Key words
opinion target extraction /
conditional random field /
semantic features /
syntactic dependency
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 王荣洋,鞠久朋,李寿山,等.基于CRFs的评价对象抽取特征研究[J].中文信息学报,2012,26(2):56-61.
[2] 戴敏,王荣洋,李寿山,等.基于句法特征的评价对象抽取方法研究[J].中文信息学报,2014,28(4):92-97.
[3] Bing Liu.Sentiment analysis and opinion mining[M].Morgan and Claypool Publishers,2012.
[4] Jakob N,Gurevych I.Extracting opinion targets in a single and cross-domain setting with conditional random Fields [C]//Proceeding of the EMNLP-2010,2010:1035-1045.
[5] Toh Z,Wang W.DLIREC:Aspect term extraction and term polarity classification system[C]//Proceeding of the International Workshop on Semantic Evaluation,2014:235-240.
[6] Hamdam Hussam,Bellot Patrice,Bechet Frederic.Lsislif:CRF and logistic regression for opinion target extraction and sentiment polarity analysis[C]//Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval-2015),2015:753-758.
[7] 徐冰,赵铁军,王山雨,等.基于浅层句法特征的评价对象抽取研究[J].自动化学报,2011,37(10):1241-1247.
[8] 杜丽萍,李晓戈.互信息改进方法在术语抽取中的应用[J].计算机应用,2015(4):996-1000.
[9] Vicente I S,Saralegi X,Agerri R.EliXa:A modular and flexible ABSA platform[C]//Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval-2015),2015:748-752.
[10] Toh Z,Su J.NLANGP:Improving aspect based sentiment analysis using neural network features[C]//Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016),2016:282-288
[11] 周红照,侯明午,颜彭莉,等.语义特征在评价对象抽取与极性判定中的作用[J].北京大学学报(自然科学版),2014,50(1):93-99.
[12] Lesk M.Automatic sense disambiguation using machine readable dictionaries:how to tell a pine cone from an ice cream cone[C]//Proceeding of the Acm Special Interest Group for Design of Communication,1986:24-26.
[13] Leacock C,Miller G A,Chodorow M.Using corpus statistics and WordNet relations for sense identification[J].Computational Linguistics,1998,24(1):147-165.
[14] Daniel M B,Imed Z.多语自然语言处理:从原理到实践[M].北京:机械工业出版社,2015.
[15] 戴敏,王荣洋,李寿山,等.基于句法特征的评价对象抽取方法研究[J].中文信息学报,2014,28(4):92-97.
[16] Hamdan H,Bellot P,Bechet F.Lsislif:CRF and logistic regression for opinion target extraction and sentiment polarity analysis[C]//Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval-2015),2015:753-758.
[17] Baccianella S,Esuli A,Sebastiani F.SentiWordNet 3.0:An enhanced lexical resource for sentiment analysis and opinion mining[C]//Proceedings of the International Conference on Language Resources and Evaluation,DBLP,2010:83-90.
[18] Guerini M,Gatti L,Turchi M.Sentiment Analysis:How to Derive Prior Polarities from SentiWordNet[J].Breast Cancer Immunodiagnosis and Immunotherapy,2013:3-11.
[19] Potts C.On the negativity of negation[C]//Proceedings of Semantics and Linguistic Theory 20.CLCPublications,2011,20:636-659.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
国家自然基金民航联合基金(U1633110);中央高校基本科研业务费专项基金(3122016D021)
{{custom_fund}}
PDF(1965 KB)
