IR-894: (2012) Singh, S.,  Wick, M. and McCallum, A., "Monte Carlo MCMC: Efficient Inference by Sampling Factors," a workshop presented at the NAACL Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction (AKBC-WEKEX 2012), Montreal, Canada, June 7-8, 2012. [View bibtex]

Abstract

Discriminative graphical models such as con- ditional random fields and Markov logic net- works have achieved state of the art results in a variety of NLP and IE tasks including coreference and relation extraction. Increasingly, automated knowledge extraction is demanding models with more complex structure— higher tree-width, larger fan-out, more features, more data—rendering even approximate inference methods such as MCMC inefficient. In this paper we propose a new MCMC sampling scheme where transition probabilities are approximated. We demonstrate that our method converges more quickly than a traditional MCMC sampler for both marginal and MAP inference. For a task of author coref- erence over 5 million mentions, we achieve a speedup of 13 over regular MCMC inference.