Word graph rescoring using confidence measures
- 24 December 2002
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1, 10-13
- https://doi.org/10.1109/icslp.1996.606917
Abstract
This paper presents a novel approach to using confidence scores for word graph rescoring. For each word in the system's vocabulary we computed the probability that the observation is correct given its acoustic score. Afterwards, we used these probabilities for rescoring word graphs outputted by the recognizer. We present some implementation detail as well as accuracy improvements obtained using this method.Keywords
This publication has 5 references indexed in Scilit:
- DP-based wordgraph pruningPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- A confidence measure for acoustic likelihood scoresPublished by International Speech Communication Association ,1995
- Large vocabulary word scoring as a basis for transcription generationPublished by International Speech Communication Association ,1995
- Recognition Confidence Measures: Detection of Misrecognitions and Out- Of-Vocabulary WordsPublished by Defense Technical Information Center (DTIC) ,1994
- Optimization of an HMM - based continuous speech recognizerPublished by International Speech Communication Association ,1993