A word graph algorithm for large vocabulary continuous speech recognition

Abstract

This paper describes a method for the construction of a word graph (or lattice) for large vocabulary, continuous speech recognition. The advantage of a word graph is that a fairly good degree of decoupling between acoustic recognition at the 10-ms level and the final search at the word level using a complicated language model can be achieved. The word graph algorithm is obtained as an extension of the one-pass beam search strategy using word dependent copies of the word models or lexical trees. The method has been tested successfully on the 20 000-word NAB'94 task (American English, continuous speech, 20 000 words, speaker independent) and compared with the integrated method. The experiments show that the word graph density can be reduced to an average number of about 10 word hypotheses, i.e. word edges in the graph, per spoken word with virtually no loss in recognition performance.