Speech segmentation and word discovery: a computational perspective

The segmentation and word discovery problem arises because speech does not contain any reliable acoustic analog of the blank spaces between words of p rinted English. As a result, children must segment the utterances they hear in order to discover the sound ms of individual words in their language. A number of computational models have been proposed to explain how children segment speech and discover words, including ten new models in the last fiv e years. This paper reviews all proposed models and organizes them accordin g to their fundamental segmentation strategies, their processing characteri stics, and the ways in which they use memory. All proposed models are found to use one of three fundamental strategies: the utterance-boundary strateg y, the predictability strategy, or the word-recognition strategy. Selected predictions of the models are explained, their performance in computer simu lations is summarized, and behavioral evidence bearing on them is discussed . Finally, ideas about how these diverse models might be synthesized into o ne comprehensive model are offered.