The CASPERS linguistic analysis system

CASPERS (Computer-Automated Speech Perception System) is a user-modifiable facility for translating strings of acoustic Symbols into sentences. Three distinctive aspects of the system's design will be discussed: the dynamic application of acoustic-phonological rules across word boundaries, an acoustic-unit splitting-and merging strategy for treating the dictionary matching problem, and an extensive capability for handling semantic routines within an augmented context-free grammar.