Research On The Construction Algorithm Of Sentiment Dictionary Based On Depth Representation

With the rapid development of Web 2.0,the Internet ushers in the social media era.People are getting used to sharing their views and positions on social media websites.The Internet has become an important platform for people to acquire and exchange information.Content posted by users on the Internet has important social and commercial value for public opinion analysis,e-commerce,information retrieval,etc.How to effectively mine and utilize this information from the mass of text has become an urgent problem to be solved.This thesis focuses on the sentiment analysis of social media tasks to study the distributed representation and sentiment lexicon construction.The first two chapters introduce the research background,research status and basic techniques,and analyze the deficiencies from different perspectives.In the third and fourth chapters,two kinds of sentiment lexicon construction methods with representation learning based on hierarchical sentiment supervision are proposed.The main contributions of this thesis are as follows:(1)To the best of our knowledge,this is the first work that learns the sentiment-aware word representation under supervision at both document and word levels.By leveraging the sentiment supervision at both document and word level,our approach can avoid the sentiment learning flaws caused by coarse-grained document-level supervision by incorporating fine-grained word-level supervision,and improve the quality of sentiment-aware word embedding.(2)The approach proposed in this thesis supports several kinds of word level sentiment annotations such as 1)predefined sentiment lexicon;2)PMI-SO lexicon with hard sentiment annotation;3)PMI-SO lexicon with soft sentiment annotation.By using PMI-SO dictionary as word-level sentiment annotation,our approach is totally corpus-based,without any external resource.(3)In addition,two kinds of sentiment lexicon construction methods are proposed,step construction method and integrated construction method.For the first method,it trains sentiment distributed representation and then predict the sentiment orientation in a supervised learning way to construct sentiment lexicon.To simplify the sentiment lexicon construction process,an integrated method is proposed.Our approaches obtain the state-of-the-art performance in comparison with several strong sentiment lexicon construction methods,on the benchmark SemEval 2013-2016 datasets for twitter sentiment classification.