Integrating Speech and Natural Language

Salim Roukos

Abstract: The overall goal of this project is to integrate speech and natural language knowledge sources to build a speech understanding system for human-machine communication using spoken English. The speech knowledge sources use acoustic models based on hidden Markov modeling techniques. The natural language knowledge sources use a Unification grammar formalism for describing the syntax of English, a higher-order intensional logic language for representing the meaning of an utterance, and a 'Montague Grammar style' framework for interfacing syntax and semantics.