We attemped to improve recognition accuracy by reducing the inadequacies of the lexicon and language model. Specifically we address the following three problems: (1) the best size...
Richard M. Schwartz, Long Nguyen, Francis Kubala, ...
We desert'be our latest attempt at adaptive language modeling. At the heart of our approachis a Maximum Entropy(ME) modelwhich inc.orlxnatesmanyknowledgesources in a consiste...
We have been developing a spoken language system to recognize and understand spontaneous speech. It is difficult for such systems to achieve good coverage of the lexicon and gramm...
In this paper, we describe a reversibleletter-to-sound/soundto-letter generation system based on an approach which combines a rule-based formalism with data-driven techniques. We ...
The BYBLOS continuous speech recognition system is applied to on-line cursive handwriting recognition. By exploiting similarities between on-line cursive handwriting and continuou...
John Makhoul, Thad Starner, Richard M. Schwartz, G...
This session had five papers related to different topics in CSR Search. The topics ranged from integration of many knowledge sources within a practical system, to different search...
Because of the complexity of documents and the variety of applications which must be supported, document understanding requires the integration of image understanding with text un...
Suzanne Liebowitz Taylor, Deborah A. Dahl, Mark Li...
The goal of this study is to evaluate the potential for using large vocabulary continuous speech recognition as an engine for automatically classifying utterances according to the...
Steve Lowe, Anne Demedts, Larry Gillick, Mark Mand...
Development has begun on a semantic evaluation (SemEval) methodology and infrastructure for the ARPA Spoken Language Program. SemEval is an attempt to define a task-independent te...
Thispaperintroduces a simple mixtare languagemodelthat attempts to capture long distance conslraints in a sentence orparagraph. The model is an m-component mixture of Irigram mode...