en
Proceedings chapter
Open access
English

Training Statistical Language Models from Grammar-Generated Data: a Comparative Case-Study

Presented at Gothenburg (Sweden)
Publication date2008
Abstract

Statistical language models (SLMs) for speech recognition have the advantage of robustness, and grammar-based models (GLMs) the advantage that they can be built even when little corpus data is available. A known way to attempt to combine these two methodologies is first to create a GLM, and then use that GLM to generate training data for an SLM. It has however been difficult to evaluate the true utility of the idea, since the corpus data used to create the GLM has not in general been explicitly available. We exploit the Open Source Regulus platform, which supports corpus-based construction of linguistically motivated GLMs, to perform a methodologically sound comparison: the same data is used both to create an SLM directly, and also to create a GLM, which is then used to generate data to train an SLM. An evaluation on a medium-vocabulary task showed that the indirect method of constructing the SLM is in fact only marginally better than the direct one. The method used to create the training data is critical, with PCFG generation heavily outscoring CFG generation.

Citation (ISO format)
HOCKEY, Beth Ann, RAYNER, Emmanuel, CHRISTIAN, Gwen. Training Statistical Language Models from Grammar-Generated Data: a Comparative Case-Study. In: Proceedings of the GoTAL Conference. Gothenburg (Sweden). [s.l.] : [s.n.], 2008.
Main files (1)
Proceedings chapter
accessLevelPublic
Identifiers
  • PID : unige:3475
558views
247downloads

Technical informations

Creation10/02/2009 9:29:05 AM
First validation10/02/2009 9:29:05 AM
Update time03/14/2023 3:14:57 PM
Status update03/14/2023 3:14:57 PM
Last indexation02/12/2024 6:13:41 PM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack