Electronic Business Review Text Opinion Mining System Research And Realization

With the advent of the Web2.0era, the comments information on the Internet isincreasing dramatically. These text messages include the users’ views and judgmentabout something. Identifying and using these comment text view informationaccurately and efficiently will have great practical value and economic value.The paper first introduced the background sense of perspective mining, and thengenerally cited some domestic and foreign representative scholars’ studies. For thispaper, it is mainly to solve three major issues: Affective Computing, Feature Miningand Perspectives Subtotals. Through the study of relevant research results at homeand abroad, some of the traditional algorithms will be improved, and some innovativealgorithm will be proposed.(1) In affective computing, this paper not only considered the effecting of clauseitself on sentence sentiment orientation, but also considered the influence of thecontext on clause emotion tendency calculation. And proposed a kind of clauseemotion inference algorithm based on the context.(2) In characteristic attributes mining, this paper used bi-iterative method toexpand attributes words and emotional words which based on the modifiedrelationship between attributes words and emotional words. According to the Chineselanguage features, using neighboring principles methods to find the modifiedrelationship between attributes words and emotional words. And using the modifiedrelationship between attributes words and emotional words to discover implicitattribute words. (3) In subtotal of opinions, this paper used the modified relationship betweenattributes words and emotional words to strengthen cluster computing of attributewords, and gained the attribute word clusters in various granularity. And then we triedto gain the classifications of opinion according to the attribute word clusters and someother information. That information will include a variety of topics in different levelsthat users will be interested in.The experimental results show that: Compared with traditional methods, theresults of precision and recall rate are greatly improved using these methods proposedby this paper, And the acquired information can intuitively specific services for userssimultaneously.