Pronunciation Adaption at the Lexical Level

There are various kinds o f adaptation which can be used to enhance the perform ance o f automatic speech recognizers. This paper is about pronunciation adaptation at the lexical level, i.e. about m odeling pronunciation variation at the lexical level. In the early years o f automatic speech recognition (ASR) research, the amount o f pronunciation variation was lim ited by using isolated w ords. Since the focus gradually shifted from isolated w ords to conversational speech, the amount o f pronunciation variation present in the speech signals has increased, as has the need to m odel it. This is reflected by the growing attention for this topic. In this paper, an overview o f the studies on lexicon adaptation is presented. Furtherm ore, m any examples are m entioned o f situations in which lexicon adaptation is likely to improve the perform ance o f speech recognizers. Finally, it is argued that some assumptions made in current standard A SR systems are not in line with the properties o f the speech signals. Consequently, the problem o f pronunciation variation at the lexical level probably cannot be solved by simply adding new transcriptions to the lexicon, as it is generally done at the moment.