Revisiting the Old Kitchen Sink: Do we Need Sentiment Domain Adaptation?

It is shown that across a wide set of domains, a simple “all-in-one” classifier that utilizes all available training data from all but the target domain tends to outperform published domain adaptation methods.

Abstract

In this paper we undertake a large crossdomain investigation of sentiment domain adaptation, challenging the practical necessity of sentiment domain adaptation algorithms. We first show that across a wide set of domains, a simple “all-in-one” classifier that utilizes all available training data from all but the target domain tends to outperform published domain adaptation methods. A very simple ensemble classifier also performs well in these scenarios. Combined with the fact that labeled data nowadays is inexpensive to come by, the “kitchen sink” approach, while technically nonglamorous, might be perfectly adequate in practice. We also show that the common anecdotal evidence for sentiment terms that “flip” polarity across domains is not borne out empirically.