Research Of Customer Evaluations Knowledge Acquisition Based On Sentiment Analysis

Along with the popularity of Internet applications, more and more people beganto use the web to express individual emotions and attitudes towards anything,including use evaluation from product users. These rich evaluation contents haveprovided a new approach for enterprises to acquire customer knowledge. However,the irregularly written evaluation contents on the web make it difficult to acquirecustomer evaluation knowledge on the web. Based on previous research foundation,this paper puts forward a method of customer evaluation knowledge acquisition basedon sentiment analysis, with specific research contents included as follows:(1) According to the characteristics of customer evaluation contents on the web,this paper in detail analyzes the processes of customer evaluation knowledgeacquisition from the web, as well as elaborating the functions which each processrealized and key techniques required, and then proposes the customer evaluationknowledge acquisition framework.(2) Product evaluation written by customers is represented by evaluation words,while sentiment lexicon can be used to determine evaluation contents. Thus, thispaper constructs sentimental words thesaurus and proposes the method of evaluationsentence identification based on sentimental lexicon.(3) Product feature extraction is the premise of analyzing customers’ sentimentalinclinations to specific product features. The actor of feature extraction is directlyrelated to the accuracy of customer evaluation knowledge acquisition. Thus, based ofprevious studies, this paper proposes the extraction method of product feature wordsbased on Apriori algorithm.(4) In product review sentences, it is an important aspect of customer evaluationknowledge acquisition to acquire the matching relationship between product featureand the corresponding evaluation word. Only if the corresponding relationshipbetween product feature word and evaluation word is determined can specificevaluation knowledge on specific product feature be accurately acquired. Therefore,this paper proposes the method of extracting the matching relationship betweenproduct feature and evaluation word based on maximum entropy model.Finally, combining with the research achievements and knowledge acquisition ofcar manufacturing enterprises, this paper designs and develops a system of knowledgeacquisition of customer evaluations in the field of automobile, which realizesautomatically acquiring customer evaluation knowledge about car products from webdata sources, therefore verifying the effectiveness of the methods proposed in thispaper.