期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2018
卷号：6
期号：2
页码：1328
DOI：10.15680/IJIRCCE.2017.0602128
出版社：S&S Publications
摘要：This paper presents a Collaborative multi-domain sentiment classification approach at the same time totrain the sentiment classifiers for multiple domains. When the labelled data is scarce, the sentiment information indifferent domain is shared to train more accurate and robust sentiment classifiers for each domain. Global one andDomain specific one are the two components of the specific classifier of the each domain. When the model is capturethe general specific knowledge and is shared by various domain, and it is the global model. When the model is capturethe specific sentiment expression in each domain, and it is the Domain specific model, and also extract domain-specificsentiment knowledge from both labelled and unlabelled samples in each domain and it is used to enhance the learningof domain-specific. To estimate the similarities between domains into regularization over the domain-specificsentiment classifiers. This is used to encourage the sharing of sentiment information the similar domains. To measurethe domain similarities, explored the two kinds of domain similarities. One is based on textual content and other isbased on sentiment expressions. To solve this model we introduce two efficient algorithm, one is Experimental resultson benchmark datasets, this algorithm show that can effectively improve the performance of multi-domain sentimentclassification and another one is significantly outperform baseline methods.
关键词：Single domain classification; Query By Commitee; All Mixed Classification
