Structured Linear Classifier

Online

Discriminative training methods for hidden Markov models: Theory and experiments with perceptron algorithms, Collins. 2002, InProc. EMNLP. Weblink

Abstract

We describe new algorithms for training tagging models, as an alternative to maximum-entropy models or conditional random �fields (CRFs). The algorithms rely on Viterbi decoding of training examples, combined with simple additive updates. We describe theory justifying the algorithms through a modi�fication of the proof of convergence of the perceptron algorithm for classi�cation problems. We give experimental results on part-of-speech tagging and base noun phrase chunking, in both cases showing improvements over results for a maximum-entropy tagger.