CONF McGreevy04b/IDIAP Pseudo-Syntactic Language Modeling for Disfluent Speech Recognition McGreevy, Michael EXTERNAL http://publications.idiap.ch/attachments/papers/2004/mcgreevy-sst04.pdf PUBLIC http://publications.idiap.ch/index.php/publications/showcite/mcgreevy04a Related documents Proceedings of SST 2004 (10th Australian International Conference on Speech Science & Technology,',','), Sydney, Australia, 2004 2004 December 2004 IDIAP-RR 04-55 Language models for speech recognition are generally trained on text corpora. Since these corpora do not contain the disfluencies found in natural speech, there is a train/test mismatch when these models are applied to conversational speech. In this work we investigate a language model (LM) designed to model these disfluencies as a syntactic process. By modeling self-corrections we obtain an improvement over our baseline syntactic model. We also obtain a 30\% relative reduction in perplexity from the best performing standard {N-gram} model when we interpolate it with our syntactically derived models. REPORT McGreevy04a/IDIAP Pseudo-Syntactic Language Modeling for Disfluent Speech Recognition McGreevy, Michael EXTERNAL http://publications.idiap.ch/attachments/reports/2004/rr04-55.pdf PUBLIC Idiap-RR-55-2004 2004 IDIAP Published in Proceedings of SST, 2004 Language models for speech recognition are generally trained on text corpora. Since these corpora do not contain the disfluencies found in natural speech, there is a train/test mismatch when these models are applied to conversational speech. In this work we investigate a language model (LM) designed to model these disfluencies as a syntactic process. By modeling self-corrections we obtain an improvement over our baseline syntactic model. We also obtain a 30\% relative reduction in perplexity from the best performing standard {N-gram} model when we interpolate it with our syntactically derived models.