Feature Based Opinion Mining System Research And Implementation In Chinese Online Reviews

With the popularity of the Internet and the rapid development of e-commerce platforms,online shopping has gradually become an important consumer habits.With the increase of trading volume,the number of user product reviews is increasing every day.It is difficult for users to obtain the fine-grained product features and associated opinions by browsing the reviews one by one.Review mining emerges as an automated review analysis method to solve this problem.The traditional document level or sentence level opinion mining method has been insufficient to satisfy the user demands of fine-grained product information.Thus,the feature based fine-grained product opinion mining research is more meaningful.In this paper,in view of the Chinese online product reviews,we have done some research on feature based opinion mining tasks:(1)Propose an explicit feature extraction method based on dependency parsing.A set of explicit feature opinion pair extraction rules and some pruning strategies have been constructed to extract explicit features directly.This method not only considers noun and noun phrase as features,but also adds verb and verb phrase.(2)Propose a feature clustering method based on similarity calculation.Feature clustering is realized by calculating the similarity of feature words,the similarity of associated opinions and combining the cluster constraints.(3)Propose an implicit feature mining method by combining the review content and two kind opinions.Based on the analysis of implicit feature appearing types in reviews,we use feature context information and two kind opinions(vague opinion and clear opinion)to extract implicit features according to the implicit indicator types.Besides,this paper verifies the positive impact of feature clustering on implicit feature extraction.(4)Finally,opinion summary merges the above feature opinion information,and then calculates the sentiment tendencies of each feature by constructing the triples<feature,opinion,negative words>with sentiment dictionaries.An opinion mining web system which provides variety views of opinion mining results is realized for users.