Constructing candidate word lists using acoustically similar word groups

A method for identifying a set of candidate words that matches well with a given utterance is discussed. The method uses precomputed groups of acoustically similar words. Given a vocabulary, for each word a short list of words that are acoustically similar to it is computed. These groups are constructed using training utterances recorded by several speakers and used in an isolated utterance recognition system to construct candidate word lists during recognition. Experiments show that the correct word appears on the short list of candidate words constructed during recognition using this scheme over 98% of the time