CONCEPTUAL ANALYSIS IN PROLOG

Date
1985-12-01
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Artificial intelligence is maturing both as a powerful technology and as a science. Expert systems are acknowledged as useful applications of AI research; recent publications reflect upon AI methodology and re-evaluate significant past efforts in the field. This thesis adds to the growing body of evaluations and criticisms by describing MARVIN - a "rational" reconstruction of key elements underlying three important natural language analysis systems based on conceptual dependency theory. In addition, MARVIN aims to control system complexity by separating knowledge from control. An infrastructure based on PROLOG and an object oriented representation scheme is instrumental in achieving this separation. A detailed description of MARVIN's knowledge base and control mechanism is followed by an objective evaluation pointing out its many shortcomings. In addition, MARVIN is compared with ELI - one of the conceptual dependency systems which it aims to mimic. MARVIN fails to achieve the flexibility and power demonstrated by ELI, but it redeems itself on design. The thesis concludes on a positive note with respect to conceptual dependency theory: it is a viable alternative to traditional parsing techniques; but needs to be strengthened considerably to encompass a broader linguistic scope, and to substantiate its psychological claims.
Description
Keywords
Computer Science
Citation