EUROSPEECH 2001 Scandinavia
7th European Conference on Speech Communication and Technology

Aalborg, Denmark
September 3-7, 2001


Lower WERs Do Not Guarantee Better Transcriptions

Judith M. Kessens, Helmer Strik

University of Nijmegen, The Netherlands

The goal of this paper is to investigate the effect of various properties of the CSR on automatic transcription. To this end, we used various versions of a continuous speech recognizer (CSR) to make automatic transcriptions. Our results show that changing certain properties of the CSR affects the resulting automatic transcriptions. The best results were obtained when 'short' hidden Markov models (HMMs), and context-independent HMMs were used. Furthermore, we found that minimizing the amount of contamination in the HMMs improves the quality of the automatic transcriptions. Another important result is that there does not appear to be a straightforward relation between word error rate (WER) and the transcription quality. In other words: A CSR with a lower WER does not always guarantee better transcriptions.

Full Paper

Bibliographic reference.  Kessens, Judith M. / Strik, Helmer (2001): "Lower WERs do not guarantee better transcriptions", In EUROSPEECH-2001, 1721-1724.