Paper: Accelerated Estimation of Conditional Random Fields using a Pseudo-Likelihood-inspired Perceptron Variant

ACL ID E14-4015
Title Accelerated Estimation of Conditional Random Fields using a Pseudo-Likelihood-inspired Perceptron Variant
Venue Annual Meeting of The European Chapter of The Association of Computational Linguistics
Session Main Conference
Year 2014
Authors

We discuss a simple estimation approach for conditional random fields (CRFs). The approach is derived heuristically by defin- ing a variant of the classic perceptron al- gorithm in spirit of pseudo-likelihood for maximum likelihood estimation. The re- sulting approximative algorithm has a lin- ear time complexity in the size of the la- bel set and contains a minimal amount of tunable hyper-parameters. Consequently, the algorithm is suitable for learning CRF- based part-of-speech (POS) taggers in presence of large POS label sets. We present experiments on five languages. Despite its heuristic nature, the algorithm provides surprisingly competetive accura- cies and running times against reference methods.