An economical program for limited parsing of English

Abstract

Automatic syntactic analysis has often been proposed as a component of mechanized indexing systems.1,2 However, up to this time, frequency counting and statistical association techniques have been more favored, since these involve operations which can be performed with great speed on present day computers. Syntactic analysis programs,3 especially the few which have relatively complete grammars, have suffered from the disadvantage of slow and expensive operation and consequently have seldom been applied beyond the field of mechanical translation. In this paper, we report the design and testing of a limited syntactic recognition program for English which shows promise of becoming accurate enough to aid in mechanized indexing, yet sufficiently inexpensive to make large-scale use practicable.