Font Size: a A A

Comments Mining Semantic Information Extraction,

Posted on:2009-12-02Degree:MasterType:Thesis
Country:ChinaCandidate:J F ZhangFull Text:PDF
GTID:2208360272459706Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Opinion mining is the mining and analysis of review text, including the determination of subjectiveness or objectiveness,the sentiment classification and the extraction of opinion elements which contain product name,product feature,opinion-bearing word and semantic orientation.The opinion mining is a hot topic in current natural language processing field.It has been widely applied in many fields such as user feedback analysis of commercial products,public opinion survey,junk mail filtering,information security,text summarization,etc.The opinion mining is involved in wide aspects which include natural language processing,machine learning,statistical analysis and so on. This article combined with practical system development performs a deep research on emphasizing on some key techniques of opinion mining.First,this paper gives an introduction of the task of opinion mining and related work.Then this article combined with practical system development performs a deep research on emphasizing on some key techniques of it.What is the most important is the semantic information extraction. This is to extract the structured opinion unit from unstructured opinion text.We divide it into four phases:1st,extract the opinion targets and opinion-bearing words,2nd,extract the relationship between opinion targets and opinion-beating terms,3rd,extend the opinion pairs to opinion triple,4th,determine the semantic orientation.We apply different policies to recognize the opinion targets and opinion-bearing terms.The experiments show that the segmentation features can improve the performance of the reorganization of opinion targets.And we employed the conditional random field to combine lexical,part of speech,semantic and positional features derived from text.Our method solves the problem of coreference and omitting of opinion targets to some extent.The experiments showed that the F value of our method is 15%higher than that of baseline which takes the nearest opinion target as the real target.Besides,the experiments found that the intensifiers can improve the performance of relation extraction.At last,the research achievements are concluded and an affirmative outlook is presented for the research of opinion mining.
Keywords/Search Tags:Opinion mining, Conditional random field, Relation extraction, Name entity recognition
PDF Full Text Request
Related items