A Comparison of Methods for Transductive Transfer Learning

In this paper we examine the problem of domain adaptation for protein name extraction. First we define the general problem of transfer learning and the particular subproblem of domain adaptation. We then describe some current state of the art supervised and transductive approaches involving support vector machines and maximum entropy models. Using these as inspiration, we turn to the unsupervised version of the problem and introduce a novel maximum entropy based technique, pseudo-label based rescaling (PLR), that achieves comparable performance with no labeled target data. We present the results of experimental comparisons between all the methods described and conclude with a discussion of trends observed and promising routes for future work.