Language Modelling for Automatic Speech Recognition: A Review

摘要

This memorandum reviews recent studies and developments in methods of language modelling which are specifically relevant to automatic speech recognition (ASR). An introduction is given to the general area of language models and the ways of formalising linguistic knowledge. Various techniques for applying phonological, syntactic and semantic constraints to ASR are discussed. The review covers papers written as early as the 1970's but the emphasis is on the more recent developments and techniques which are now being used in speech research. The formal methods of applying linguistic constraints are discussed and criticised according to their suitability for the speech research work carried out at RSRE.