Paper: Dual Training and Dual Prediction for Polarity Classification

ACL ID P13-2093
Title Dual Training and Dual Prediction for Polarity Classification
Venue Annual Meeting of the Association of Computational Linguistics
Session Short Paper
Year 2013
Authors

Bag-of-words (BOW) is now the most popular way to model text in machine learning based sentiment classification. However, the perfor- mance of such approach sometimes remains rather limited due to some fundamental defi- ciencies of the BOW model. In this paper, we focus on the polarity shift problem, and pro- pose a novel approach, called dual training and dual prediction (DTDP), to address it. The basic idea of DTDP is to first generate artifi- cial samples that are polarity-opposite to the original samples by polarity reversion, and then leverage both the original and opposite samples for (dual) training and (dual) predic- tion. Experimental results on four datasets demonstrate the effectiveness of the proposed approach for polarity classification.