Sentiment Lexicon Construction Based On Multi-source Information Fusion

With the continuous innovation of network technology,more and more Internet products have become a must for people's daily life,such as Twitter,Weibo,Amazon Mall and so on.These products not only serve users,but also a good platform for data generation.Emotional analysis of the textual data generated by these platforms is conducive to understanding popular ideas,helping businesses optimize products,controlling public opinion,etc.Emotional analysis techniques have been applied to various fields of academia and industry.Despite this,review sentiment classification is still a challenge,because there is no length limit for reviews,the comments contain a lot of noise data,the length is not limited,and there is no emotional label.When using supervised learning methods for text sentiment classification,it is often necessary to manually mark large amounts of text data.In order to reduce the artificial participation,this paper proposes a tag data expansion method based on similar user rating comments,which can reduce the number of manual tag comments required for supervised learning methods to a certain extent.In order to further reduce the demand for labeled data in sentiment analysis tasks,we proposed an emotion dictionary construction method of multi-source information fusion for commenting emotion classification in the basic idea of tag data expansion method.The method can integrate the existing sentiment dictionary,a small amount of labeled data,a large amount of unlabeled comment data,and four kinds of sentiment information extracted from four sources of scoring data to generate a sentiment dictionary with emotional polarity values.Our research also uses an optimization method based on ADMM algorithm to solve the proposed model.Finally,more experiments were conducted on five actual Amazon merchandise sales datasets.The results show that the sentiment dictionary generated by the proposed method has a significant improvement in the accuracy of text sentiment classification compared with the current mainstream methods.The above data proves that the emotional dictionary constructed by the proposed method is better than the current mainstream emotional dictionary construction method,which provides some theoretical support and practical reference for researchers in this field.