Paper: Part-of-speech tagging with antagonistic adversaries

ACL ID P13-2113
Title Part-of-speech tagging with antagonistic adversaries
Venue Annual Meeting of the Association of Computational Linguistics
Session Short Paper
Year 2013
Authors

Supervised NLP tools and on-line services are often used on data that is very dif- ferent from the manually annotated data used during development. The perfor- mance loss observed in such cross-domain applications is often attributed to covari- ate shifts, with out-of-vocabulary effects as an important subclass. Many discrim- inative learning algorithms are sensitive to such shifts because highly indicative fea- tures may swamp other indicative features. Regularized and adversarial learning algo- rithms have been proposed to be more ro- bust against covariate shifts. We present a new perceptron learning algorithm us- ing antagonistic adversaries and compare it to previous proposals on 12 multilin- gual cross-domain part-of-speech tagging datasets. While previous approaches do not improve o...