Opinion Analysis Using Polarity Shift Model

Rui Xia,Feng Xu,Jianfei Yu,Yong Qi

2017-01-01

Abstract:The user-generated text volume on the Web in the form of reviews, blogs, and social networks has grown dramatically in recent years. This was mirrored by an increasing interest, from both the academic and the business world, in the field of opinion analysis. The polarity shift problem is a major factor that affects classification performance of machine-learning-based sentiment analysis systems .Here we proposing a three-stage cascade model to address the polarity shift problem in the context of document-level sentiment classification. Firstly splitting of each document into a set of sub sentences and build a hybrid model that employs rules and statistical methods to detect explicit and implicit polarity shifts, respectively is done. Secondly, proposing a polarity shift elimination method, to remove polarity shift in negations. Afterwards, training the base classifiers on training subsets divided by different types of polarity shifts, and uses a weighted combination of the component classifiers for sentiment classification. On this basis, also proposing a dual training algorithm to make use of original and reversed training reviews in duality for learning a sentiment classifier, and a dual prediction algorithm to classify the test results by keeping in mind both of two phases of one result. An extended framework from polarity (positive-negative) classification to 3-class (positivenegative-neutral) classification has to be done, by taking the neutral reviews into consideration. Finally, a corpus-based approach is constructed for pseudo-antonym dictionary, which elimination of Dual Sentential approach's dependency on an external antonym dictionary for