Research On Opinion Mining Based On Product Comments

With the advent of the information age,the amount of the comment text in the network is increasing,and people have put forward higher requirements on how to get valuable information from the mass comment text.More and more practical application demand makes opinion mining develop rapidly,and lots of scholars and experts have studied it and applied it to the real life.Opinion mining can be subdivided into many small tasks,which can be divided into opinion extraction,polarity analysis,and opinion summarization.Opinion target extraction belongs to the opinion extraction task,and the polarity judgment is one of polar analysis task s.These two subtasks are very important when it comes to opinion mining.Extracting opinion targets and judging the sentiment orientation of reviews is this paper's work.The majority of the existing evaluation methods are dependent on the rules,which means that it is better applied to a specific field.At the same time,in practice,the review text is updated and expa nded soon,and lots of artificial participation reduces the practicability of the method.In order to solve the above problems,this paper proposes a method for extracting opinion target based on CRFs and co-clustering,which regards opinion target extraction as one of the sequence labeling problems.The CRFs model is used to extract the candidate sets of opinion target.By utilizing the co-clustering algorithm,the information of opinion object and comment word is fully utilized.And opinion sets can be got at the same time.In this algorithm,adjustable parameters include the number of opinion target clusters and the distance between the opinion object and the comment word.In this paper,the emotional word dictionary is constructed based on the HowNet,which also concludes the synonyms and the Internet word,to provide the basis information for the polarity judgment.According to the characteristics of the comment words in the Internet,TF-IDF is used to calculate the intensity of the comment words.The polarity and intensity of the evaluation word according to the emotional word dictionary,the intensity of the comment word by TF-IDF,and the impact of the degree modification words are all taken into consideration for the completion of the polarity judgment.In this paper,hotel reviews and laptop reviews are used to verify the above method.Experimental results show that the proposed method has better result,which proves its practicability.