Abstract for woodland_lvcsr01

LVCSR Hub5 Workshop 2001, Linthicum Heights


P.C. Woodland, T. Hain, G. Evermann, D. Povey

May 2001

This paper describes the Cambridge University HTK (CU-HTK) system developed for the NIST March 2001 evaluation of English conversational telephone speech transcription (Hub5E). A number of new features were integrated into the system and described in detail. Also some experiments on data recorded over a cellular network are presented. The major new features are improved MMIE training and the use of lattice-based MLLR adaptation. Results are presented for the 2001 development and evaluation test sets. In the March 2001 Hub5 evaluation the CU-HTK system achieved an overall word error rate of 24.6%, which was the lowest word error rate by a statistically significant margin.

| (ftp:) woodland_lvcsr01.ps.gz | (http:) woodland_lvcsr01.ps.gz | (ftp:) woodland_lvcsr01.pdf | (http:) woodland_lvcsr01.pdf |

If you have difficulty viewing files that end '.gz', which are gzip compressed, then you may be able to find tools to uncompress them at the gzip web site.

If you have difficulty viewing files that are in PostScript, (ending '.ps' or '.ps.gz'), then you may be able to find tools to view them at the gsview web site.

We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.