Integration of Speech and Natural Language
Abstract
This report describes the progress during both years of the project, from January 1, 1987 to December 30, 1988. During the course of the project, we have focused on two major activities: Developing syntactic and semantic components for natural language processing; and Integrating the developed syntax and semantics with speech for speech understanding. To measure the coverage of the syntactic and semantic components and the performance of the integrated system, we use the DARPA 1000-word Resource Management Database corpus. This corpus has been used for developing a standard speech database for evaluating the performance of speech recognition algorithms developed under the Strategic Computing Program. Development and evaluation of the natural language processing components-i.e. syntax and semantics-are done using a training corpus and a test corpus, both drawn from the Resource Management domain. The training corpus is examined by the system developers and is used to determine the phenomena that should be handled. The test set is never examined by the developers; its purpose is to allow us to estimate performance on an independent set that would be representative of the ultimate system's performance in the field.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1989
- Accession Number
- ADA206679
Entities
People
- Andrew Haas
- R. Ingria
- S. Boisen
- S. Roukos
- Y. Chow
Organizations
- BBN Technologies