Third International Conference on Spoken Language Processing (ICSLP 94)
Proposed in this paper is a model of inferring the meanings of unknown words and rinding plausible words that fit to phoneme sequence containing inaudible/unsure articulation in dialog speech. The model is divided into two cases which correspond to particular mechanisms for 1) recognizing unknown words, and then inferring their meanings by looking for a coherent interpretation that satisfies semantic, syntactic and pragmatic constraints, and 2) finding the most plausible words to be assigned to some part of phoneme sequence containing inaudible sounds or unsure phoneme recognition.
Case 1) appears when one clearly hears words out of his/her vocabulary, while case 2) usually occurs to aged persons having hearing loss. The latter sometimes happens even to persons of ordinary hearing particularly in case he/she holds preconceptions or preoccupations.
Bibliographic reference. Yanagida, Masuzo (1994): "A cognitive model of inferring unknown words and uncertain sound sequence", In ICSLP-1994, 1423-1426.