A multi-processing approach to natural language

  • Authors:
  • Ronald M. Kaplan

  • Affiliations:
  • Harvard University, Cambridge, Massachusetts

  • Venue:
  • AFIPS '73 Proceedings of the June 4-8, 1973, national computer conference and exposition
  • Year:
  • 1973

Quantified Score

Hi-index 0.00

Visualization

Abstract

Natural languages such as English are exceedingly complicated media for the communication of information, attitudes, beliefs, and feelings. Computer systems that attempt to process natural languages in more than the most trivial ways are correspondingly complex. Not only must they be capable of dealing with elaborate descriptions of how the language is put together (in the form of large dictionaries, grammars, sets of inference strategies, etc.), but they must also be able to coordinate the activities and interactions of the many different components that use these descriptions. For example, speech understanding systems of the sort that are currently being developed under ARPA sponsorship must have procedures for the reception of speech input, phonological segmentation and word recognition, dictionary consultation, and morphological, syntactic, semantic, and pragmatic analyses. The problems of coordination and control are reduced only slightly in less ambitious projects such as question answering, automatic programming, content analysis, and information retrieval. Of course, large-scale software systems in other domains might rival natural language programs in terms of the number and complexity of individual components. The central theme of the present paper, however, is that natural language control problems have a fundamentally different character from those of most other systems and require a somewhat unusual solution: the many natural language procedures should be conceptualized and implemented as a collection of asynchronous communicating parallel processes.