Paper: A Speech Understanding Framework that Uses Multiple Language Models and Multiple Understanding Models

ACL ID N09-2034
Title A Speech Understanding Framework that Uses Multiple Language Models and Multiple Understanding Models
Venue Human Language Technologies
Session Short Paper
Year 2009
Authors

The optimal combination of language model (LM) and language understanding model (LUM) varies depending on available training data and utterances to be handled. Usually, a lot of effort and time are needed to find the op- timal combination. Instead, we have designed and developed a new framework that uses multiple LMs and LUMs to improve speech understanding accuracy under various situa- tions. As one implementation of the frame- work, we have developed a method for select- ing the most appropriate speech understand- ing result from several candidates. We use two LMs and three LUMs, and thus obtain six combinations of them. We empirically show that our method improves speech understand- ing accuracy. The performance of the oracle selection suggests further potential improve- ments in our s...