Research On The Method Of Extracting Opinion Words And Product Features From Chinese Online Reviews

After shopping in the e-commerce website,people usually will comment on the quality of the product.Previously consumers can only rely on word-of-mouth advertising effect to purchase products.Now the e-commerce platform provides an area to comment,consumers can acquire the product overview through browse these reviews,and then make a decision whether to buy it.In addition,the e-commerce platform can present some accurate advertisements based on these reviews,manufactures can also make reasonable analysis and decision making to improve product quality,so they can produce many product that consumers need.At present,the online reviews is very large,if these reviews are handled manually,it is very time consuming and inefficient.Therefore,product reviews mining technology emerge as the times require.In the online reviews,as fine-grained aspects of product,product feature shows the focus of products which consumers pay attention to,opinion word is a reflection of consumer's attitude to product,it reflects the user 's feelings after experience product,product feature and opinion word are foundation of further sentiment tendentiousness recognition.The main task of review mining is to dig opinion words and product features from product reviews.The main contents of this paper are as follows:By analyzing the online reviews,this paper puts forward the opinion word extraction method based on dependency relations.In the process of opinion word automatic recognition,first,preprocess these reviews,the steps of preprocessing are divided into word segmentation,part-of-speech tagging and syntactic analysis.Then design POS template which contains opinion word,and combined with syntactic dependency relations which extract from reviews to get the opinion word set.This paper proposes a product features extraction model based on opinion word co-occurrence which realize product features extraction of online reviews.After the preprocessing of the reviews,we use language rules and based on association rule mining algorithm respectively to produce the product candidate feature set,then pruning and filtering it.Last,get product feature set with the method of opinion word co-occurrence in considering the context of feature.Finally,this paper gives experiments of extraction of opinion words and product features,three kinds of product reviews which grab from Internet are tested by those above method,the experimental results show that those presented methods achieved better precision and F-measure.