Frustratingly Easy Domain Adaptation

TL;DR

This paper introduces a simple and effective domain adaptation method that leverages target data to improve performance over source-only models, outperforming existing approaches and easily extending to multiple domains.

Contribution

The authors present an extremely simple, easy-to-implement domain adaptation technique that outperforms state-of-the-art methods and can be extended to multi-domain scenarios.

Findings

Outperforms state-of-the-art domain adaptation methods

Easy to implement as a short preprocessing step

Effective in multi-domain adaptation

Abstract

We describe an approach to domain adaptation that is appropriate exactly in the case when one has enough ``target'' data to do slightly better than just using only ``source'' data. Our approach is incredibly simple, easy to implement as a preprocessing step (10 lines of Perl!) and outperforms state-of-the-art approaches on a range of datasets. Moreover, it is trivially extended to a multi-domain adaptation problem, where one has data from a variety of different domains.