Multi-domain adaptation for sentiment classification: Using multiple classifier combining methods

Sentiment classification is very domain-specific and good domain adaptation methods, when the training and testing data are drawn from different domains, are sorely needed. In this paper, we address a new approach to domain adaptation for sentiment classification in which classifiers are adapted for a specific domain with training data from multiple source domains. We call this new approach dasiamulti-domain adaptationpsila and present a multiple classifier system (MCS) framework to describe and understand it. Under this framework, we propose a new combining method, called Multi-label Consensus Training (MCT), to combine the base classifiers for selecting dasiaautomatically-labeledpsila samples from unlabeled data in the target domain. The experimental results for sentiment classification show that multi-domain adaptation using this method improves adaptation performance.