Product Review Emotion Classification Based On CRFs

As the number of Internet users dramatically increased and therapid development of electronic commerce, the network has emergedmany B2B and B2C website (such as360buy, TaoBao, DangDangnetwork, etc.).These sites have a common characteristic, which is inproduct sales, but also provides consumers with a published productreviews of the platform. Consumers can promptly to the evaluation ofcommodities show come out, these comments information can givefeedback to the merchants and potential consumers. Merchants canimprove goods or adjustment according to the feedback sales strategy,the potential consumers through the reference to the commodityreview of others, make the right shopping choice. But the Internetinformation a huge number, how in a lot of comments in theinformation found information of value, and reading these commentsall to make the right decision or choice, it is very difficult, so badlyneeded a kind of fast, accurate information of text mining method formass comments information analysis. In addition, with the changes ofthe network language, the mining method must have machine learningability, real-time update vocabulary, so as to get the right result.Emotional classification of text mining is an application and it isthe artificial intelligence research in the field of hot topic. Many scholarsdevoted to language and image of the emotional aspects of theresearch. The research and accurately familiar analyze Chineseemotional artificial intelligence machine learning method.This paper describes the concept of text mining technologyand related information, and then mainly introduced in this paper tocomment on the emotional mining which technology, including marktechnology, Conditional Random Field (CRFs) algorithm and evaluationindex. In this chapter3, we to the above involves to the comments text mining technology emotions are discussed, we borrow in the field oftext classification feature selection sure features a template, and thenthrough the word tagging system, will review text classification problemfor emotional transformation sequence mark task, then using CRFsmachine learning method to the classification. In addition, we putforward a kind of emotion of the strength of the score for the emotionalclassification results mechanism ranking. The experimental results showthat the emotional classification based on CRFs comments canachieve high precision, for classification results emotion also rankedcompared accord with the objective facts.The main contribution of this lies in the following aspects:1.Based on the traditional and two emotions tend to emotionaldictionary based on the " HowNet" words similarity calculation method,considering the part of speech of emotional words, adverbs of degreeof emotional inclinations, negative words, the influence of constructinga relatively mature emotion dictionary.2.The sequence mark technology, will CRFs machine learningmethod is introduced to comment text classification of emotion,achieving high accuracy classification results.3.Put forward based on the maximum entropy ranking algorithm tothe strength of the emotional inclinations rank, makes the classificationresults to be more objective.Finally,this paper constructed a web based product reviewsinformation classification system emotion, the system will be emotionalclassification results in the form of a list displayed, users can getvaluable information.