The Rich Transcription 2006 Spring Meeting Recognition Evaluation

Published

May 4, 2006

Author(s)

Jonathan G. Fiscus, Jerome G. Ajot, Martial Michel, John S. Garofolo

Abstract

This paper presents the design and results of the Rich Transcription Spring 2006 (RT-06S) Meeting Recognition Evaluation. This evaluation is the fourth in a series of community-wide evaluations of language technologies in the meeting domain. For 2006, three evaluation tasks were supported for two sub domains of meetings. The tasks were the speech-to-text (STT) transcription task and two diarization tasks: ¿Who Spoke When¿ and ¿Speech Activity Detection¿. The meeting sub domains were the Conference Meetings (confmtg) and Lecture Meeting (lectmtg) sub domains. For the STT task, the lowest word error rate for the up to 4 simultaneous speakers for the multiple distant microphone condition was 46.3% for the confmtg sub domain and 53.4% for the lectmtg sub domain. For the ¿Who Spoke When¿ task, the lowest diarization error rate for all speech was 35.8% for the confmtg sub domain and 24.0% for the lectmtg sub domain. For the ¿Speech Activity Detection¿ task, the lowest diarization error rate was 4.3% for the confmtg sub domain and 8.0% for the lectmtg sub domain.

Conference Title

Rich Transcription Spring Meeting Recognition Evaluation 2006

Pub Type

Conferences

Citation

Fiscus, J. , Ajot, J. , Michel, M. and Garofolo, J. (2006), The Rich Transcription 2006 Spring Meeting Recognition Evaluation, Rich Transcription Spring Meeting Recognition Evaluation 2006 (Accessed April 23, 2024)