Domain Adaptation for Sentiment Analysis Using Increased Intraclass Separation

Sentiment analysis is a costly yet necessary task for enterprises to study\nthe opinions of their customers to improve their products and to determine\noptimal marketing strategies. Due to the existence of a wide range of domains\nacross different products and services, cross-domain sentiment analysis methods\nhave received significant attention. These methods mitigate the domain gap\nbetween different applications by training cross-domain generalizable\nclassifiers which help to relax the need for data annotation for each domain.\nMost existing methods focus on learning domain-agnostic representations that\nare invariant with respect to both the source and the target domains. As a\nresult, a classifier that is trained using the source domain annotated data\nwould generalize well in a related target domain. We introduce a new domain\nadaptation method which induces large margins between different classes in an\nembedding space. This embedding space is trained to be domain-agnostic by\nmatching the data distributions across the domains. Large intraclass margins in\nthe source domain help to reduce the effect of "domain shift" on the classifier\nperformance in the target domain. Theoretical and empirical analysis are\nprovided to demonstrate that the proposed method is effective.\n