A parallel parser for spoken natural language

摘要: This paper describes SYNAPSIS, a parser for performing real-time understanding of spoken utterances in parallel computational environment. Understanding continuous speech allowing reasonably free syntax poses two main problems, namely the risk erroneous interpretations and largeness search space owing to high uncertainty input. The is characterized by an approach whose major novel features are 1) drastic reduction idle time, thanks asynchronous inter agent communications, 2) modularity, distribution homogeneous pieces knowledge rather than different parsing tasks. These make most apt implementation on Transputer-based distributed architectures.