Effects of OOV rates on Keyphrase Rejection Schemes

摘要

Recognising directory listings for national telephone number inquiry is slowly getting within reach for modern ASR technology. Two key factors for a successful system design are (1) optimal extent of lexical modelling and (2) an effective utterance rejection method. In this paper we show how a choice for the first has consequences for the second. We have taken the approach of building a lexicon with multiword expressions for the most frequently requested telephone listings, stepwise extended with filler words and less frequently addressed listings. In doing so, we keep track of the consequences that different Out of Vocabulary (OOV) rates have on two diverging keyphrase rejection schemes. Experimental results on field data clearly show that tasks with high OOV rates benefit most from acoustic confidence measures, while tasks with low OOV rates are better off with N-best list-based rejection schemes.