Analysis of Languages for Man-Machine Voice Communication
Abstract
This dissertation describes a general model for the analysis of languages for man-machine communication. It is the first known study of ambiguity at all levels of recognition and represents the best analytical tool, to date, for the design of languages. The model unifies the concepts of ambiguity and restriction done by expressing each as a branching factor--a notion which is easily understood and visualized. Ambiguity increases the branching factor while restriction reduces it. Using branching factor has the advantage that an effective search space size may be computed for any language. Further, since ambiguity and syntactic restriction are expressed in a uniform way, the effect of one with respect to the other may be evaluated by considering search space reduction ratios. The model is useful for comparing the relative complexities faced by speech understanding systems. Effective vocabulary size provides a way of measuring the complexity in isolated word recognition while effective search space size measures language complexity. Thus, the performance of 2 systems may be contrasted by using these measures; previously, this could be done only if the 2 systems had been tested using the same data; a situation which occured rarely.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1976
- Accession Number
- ADA035564
Entities
People
- Robert G. Goodman
Organizations
- Carnegie Mellon University