Detail publikace

Further Progress in Meeting Recognition: The ICSI-SRI Spring 2005 Speech-to-Text Evaluation System

STOLCKE, A., ANGUERA, X., BOAKYE, K., CETIN, Ö., GRÉZL, F., JANIN, A., MANDAL, A., PESKIN, B., WOOTERS, C., ZHENG, J.

Originální název

Further Progress in Meeting Recognition: The ICSI-SRI Spring 2005 Speech-to-Text Evaluation System

Anglický název

Further Progress in Meeting Recognition: The ICSI-SRI Spring 2005 Speech-to-Text Evaluation System

Jazyk

en

Originální abstrakt

We describe the development of our speech recognition system for
the NIST Spring 2005 Meeting Rich Transcription (RT-05S) evaluation, highlighting improvements made since last year. The system is based on the SRIICSI-UW RT-04F conversational telephone speech (CTS) recognition system, with meeting-adapted models and various audio preprocessing steps. This year's system features better delay-sum processing of distant microphone channels and energy-based crosstalk suppression for close-talking microphones. Acoustic modeling is improved by virtue of various enhancements to the background (CTS) models, including added training data, decision-tree based state tying, and the inclusion of discriminatively trained phone posterior features estimated by multilayer perceptrons. In particular, we make use of adaptation of both acoustic models and MLP features to the meeting domain. For distant microphone recognition we obtained considerable gains by combining and cross-adapting narrowband (telephone) acousticmodels with broadband (broadcast news) models. Language models (LMs) were improved with the inclusion of new meeting and web data. In spite of a lack of training data, we created effective LMs for the CHIL lecture domain. Results are reported on RT-04S and RT-05S meeting data. Measured on RT-04S conference data, we achieved an overall improvement of 17% relative in bothMDM and IHMconditions compared to last year's evaluation system. Results on lecture data are comparable to the best reported results for that task.

Anglický abstrakt

We describe the development of our speech recognition system for
the NIST Spring 2005 Meeting Rich Transcription (RT-05S) evaluation, highlighting improvements made since last year. The system is based on the SRIICSI-UW RT-04F conversational telephone speech (CTS) recognition system, with meeting-adapted models and various audio preprocessing steps. This year's system features better delay-sum processing of distant microphone channels and energy-based crosstalk suppression for close-talking microphones. Acoustic modeling is improved by virtue of various enhancements to the background (CTS) models, including added training data, decision-tree based state tying, and the inclusion of discriminatively trained phone posterior features estimated by multilayer perceptrons. In particular, we make use of adaptation of both acoustic models and MLP features to the meeting domain. For distant microphone recognition we obtained considerable gains by combining and cross-adapting narrowband (telephone) acousticmodels with broadband (broadcast news) models. Language models (LMs) were improved with the inclusion of new meeting and web data. In spite of a lack of training data, we created effective LMs for the CHIL lecture domain. Results are reported on RT-04S and RT-05S meeting data. Measured on RT-04S conference data, we achieved an overall improvement of 17% relative in bothMDM and IHMconditions compared to last year's evaluation system. Results on lecture data are comparable to the best reported results for that task.

Dokumenty

BibTex


@inproceedings{BUT18258,
  author="Andreas {Stolcke} and Xavier {Anguera} and Kofi {Boakye} and Özgür {Cetin} and František {Grézl} and Adam {Janin} and Arindam {Mandal} and Barbara {Peskin} and Chuck {Wooters} and Jing {Zheng}",
  title="Further Progress in Meeting Recognition: The ICSI-SRI Spring 2005 Speech-to-Text Evaluation System",
  annote="We describe the development of our speech recognition system for
the NIST Spring 2005 Meeting Rich Transcription (RT-05S) evaluation, highlighting improvements made since last year. The system is based on the SRIICSI-UW RT-04F conversational telephone speech (CTS) recognition system, with meeting-adapted models and various audio preprocessing steps. This year's system features better delay-sum processing of distant microphone channels and energy-based crosstalk suppression for close-talking microphones. Acoustic modeling is improved by virtue of various enhancements to the background (CTS) models, including added training data, decision-tree based state tying, and the inclusion of discriminatively trained phone posterior features estimated by multilayer perceptrons. In particular, we make use of adaptation of both acoustic models and MLP features to the meeting domain. For distant microphone recognition we obtained considerable gains by combining and cross-adapting narrowband (telephone) acousticmodels with broadband (broadcast news) models. Language models (LMs) were improved with the inclusion of new meeting and web data. In spite of a lack of training data, we created effective LMs for the CHIL lecture domain. Results are reported on RT-04S and RT-05S meeting data. Measured on RT-04S conference data, we achieved an overall improvement of 17% relative in bothMDM and IHMconditions compared to last year's evaluation system. Results on lecture data are comparable to the best reported results for that task.", address="University of Edinburgh", booktitle="Machine Learning for Multimodal Interaction, Second International Workshop, MLMI 2005, Edinburgh, UK, July 11-13, 2005, Revised Selected Papers", chapter="18258", edition="Lecture Notes in Computer Science 3869, Springer 2006", institution="University of Edinburgh", year="2005", month="july", pages="463--475", publisher="University of Edinburgh", type="conference paper" }