Discriminative training methods for hidden Markov models: theory and experiments with perceptron algorithms
M. Collins. Proceedings of the ACL-02 conference on Empirical methods in natural language processing - Volume 10, page 1--8. Stroudsburg, PA, USA, Association for Computational Linguistics, (2002)
DOI: 10.3115/1118693.1118694
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 modification of the proof of convergence of the perceptron algorithm for classification 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.
Description
Discriminative training methods for hidden Markov models
%0 Conference Paper
%1 collins2002perceptron
%A Collins, Michael
%B Proceedings of the ACL-02 conference on Empirical methods in natural language processing - Volume 10
%C Stroudsburg, PA, USA
%D 2002
%I Association for Computational Linguistics
%K averaged docrped perceptron prediction sequential structured viterbi
%P 1--8
%R 10.3115/1118693.1118694
%T Discriminative training methods for hidden Markov models: theory and experiments with perceptron algorithms
%U http://dx.doi.org/10.3115/1118693.1118694
%X 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 modification of the proof of convergence of the perceptron algorithm for classification 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.
@inproceedings{collins2002perceptron,
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 modification of the proof of convergence of the perceptron algorithm for classification 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.},
acmid = {1118694},
added-at = {2012-10-25T19:28:25.000+0200},
address = {Stroudsburg, PA, USA},
author = {Collins, Michael},
biburl = {https://www.bibsonomy.org/bibtex/2710aacdc4a3b3281cf8cb6cf8beaa231/jil},
booktitle = {Proceedings of the ACL-02 conference on Empirical methods in natural language processing - Volume 10},
description = {Discriminative training methods for hidden Markov models},
doi = {10.3115/1118693.1118694},
interhash = {3612e78d171711e74bc0d1ed72ebee14},
intrahash = {710aacdc4a3b3281cf8cb6cf8beaa231},
keywords = {averaged docrped perceptron prediction sequential structured viterbi},
numpages = {8},
pages = {1--8},
publisher = {Association for Computational Linguistics},
series = {EMNLP '02},
timestamp = {2015-01-31T19:32:36.000+0100},
title = {Discriminative training methods for hidden Markov models: theory and experiments with perceptron algorithms},
url = {http://dx.doi.org/10.3115/1118693.1118694},
year = 2002
}