Improving Lightly Supervised Training for Broadcast Transcriptions

Y. Long, M.J.F. Gales, P. Lanchantin, X. Liu, M.S. Seigel, P.C. Woodland

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper investigates improving lightly supervised acoustic model training for an archive of broadcast data. Standard lightly supervised training uses automatically derived decoding hypotheses using a biased language model. However, as the actual speech can deviate significantly from the original programme scripts that are supplied, the quality of standard lightly supervised hypotheses can be poor. To address this issue, word and segment level combination approaches are used between the lightly supervised transcripts and the original programme scripts which yield improved transcriptions. Experimental results show that systems trained using these improved transcriptions consistently outperform those trained using only the original lightly supervised decoding hypotheses. This is shown to be the case for both the maximum likelihood and minimum phone error trained systems.
Original languageEnglish
Title of host publicationProc Interspeech 2013
PublisherISCA
Publication statusPublished - 2013

Fingerprint

Dive into the research topics of 'Improving Lightly Supervised Training for Broadcast Transcriptions'. Together they form a unique fingerprint.

Cite this