Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

A Semantic Interpretation Based on Detecting Concepts for Spontaneous Speech Understanding

Akito Nagai, Yasushi Ishikawa, Kunio Nakajima

Mitsubishi Electric Corp. Computer and Information Systems Lab., Kanagawa, Japan

This paper describes a two-stage semantic interpretation method where an intention is associated with an input sentence and it integrates concept hypotheses detected from the sentence. In this approach, a concept represented by several phrases is an unit of semantic interpretation, and a spontaneously spoken sentence is regarded as a sequence of concepts. Syntactic constraint is limited within a phrase to robustly understand various sentential expressions. For disambiguation of concept hypotheses, they are evaluated with scoring criteria based on linguistic knowledge. The method has been evaluated on test sets collected as text data including various type of sentences which subjects made with no limitation. The experimental results of semantic understanding using 22 concept frames with approximately 1000-word vocabulary shows that the scoring method contributes to disambiguation of meaning candidates, and that 75 % of test sentences are correctly understood.

Full Paper

Bibliographic reference.  Nagai, Akito / Ishikawa, Yasushi / Nakajima, Kunio (1994): "A semantic interpretation based on detecting concepts for spontaneous speech understanding", In ICSLP-1994, 95-98.