Automatic verification of broadcast news transcriptions
Michael Pitz, Sirko Molau · 1999
In this paper we present a method for automatically detecting erroneous training scripts for speech corpora like Broadcast News and Switchboard.Based on the Hub-4 task we will report on the performance of error detection with the proposed method and investigate the effects of both manually and automatically cleaned training corpora on the performance of the RWTH speech recognition system.Our approach uses a forced Viterbi alignment on the training data and evaluates different transcription quality measures.The following three criteria proved to be useful to automatically detect most transcription errors: the difference between the final Viterbi alignment HMM state and the last state according to the transcriptions the normalized acoustic word scores the location of the boundary between adjacent segments obtained by forced alignment With manually corrected scripts we achieved a WER reduction on the 1996 HUB-4 eval.corpus.The recognizer's performance improved mainly on clean planned speech segments.Whereas the improvements were minor on these hand-transcribed training data, automatic training script verification will become more important for automatically transcribed new speech corpora.