Active, semi-supervised learning to utilize human oracles

Abstract

We present an approach to interactive machine learning, in which unlabeled data is employed in conjunction with active learning to better utilize the valuable resources that the human oracles provide. We empirically evaluate the approach in two very different applications, smartphone interruptibility prediction and semantic parsing. In both applications, we show that the use of active, semi-supervised training results in an improvement compared to a traditionally trained classifier relying only on full-supervision with random sampling.

Extracted Key Phrases

17 Figures and Tables

Cite this paper

@inproceedings{Fisher2012ActiveSL, title={Active, semi-supervised learning to utilize human oracles}, author={Robert Fisher and Reid Simmons}, year={2012} }