Study on the Extraction of "Opinion Feature-Opinion Word" Based on Conditional Random Fields Model
Zhiyi Li,Mian Wang,Pengwu Zhao
DOI: https://doi.org/10.3772/j.issn.1000-0135.2017.04.010
2017-01-01
Abstract:In environment of "we-media",the sentiment analysis and monitoring of massive evaluation information has become more and more important,it can not only promote the in-depth study of opinion mining,but also can help the enterprise to study the needs of users to generate huge commercial value.Based on conditional random fields (CRFs) model,combined with the characteristics of syntax,syntactic properties of suitable combinations are selected by experiments,and this paper finds a path evaluation feature extraction.Secondly,a review corpus of mobile phone is constructed.Then we use a parser to review sentences into the corresponding syntax tree while observing the relationshipbetweenthe opinion features and opinion words of the existence of dependency,as well as the structural integrity of the information point of view.Using algorithm the pairs of "Opinion features and opinion words" are extracted from review corpus.A method based on dependency grammar is proposed,which is based on the evaluation of the evaluation feature,and the Arbor.js is used to carry out a visual presentation.From the experimental results,this method can improve the accuracy of the evaluation of feature extraction,and correctly access of opinion information,and also make the foundation for the polarity judgment and complete opinion mining research.