Accession Number : ADA288515

Title :   MAC/FAC: A Model of Similarity-Based Retrieval.

Descriptive Note : Technical rept. 1 Sep 89-31 Oct 92,

Corporate Author : NORTHWESTERN UNIV EVANSTON IL INST FOR THE LEARNING SCIENCES

Personal Author(s) : Forbus, Kenneth D. ; Gentner, Dedre ; Law, Keith

PDF Url : ADA288515

Report Date : OCT 1994

Pagination or Media Count : 67

Abstract : We present a model of similarity-based retrieval that attempt to capture these seemingly contradictory psychological phenomena: (1) structural commonalities are weighed more heavily than surface commonalities in soundness or similarity judgments (when both members are present); (2) superficial similarity is more important in retrieval from long-term memory than is structural similarity; and yet (3) purely structural (analogical) remindings are sometimes experienced. Our model, called MAC/FAC (for "many are called but few are chosen") consists of two stages. The first stage (MAC) uses a computationally cheap, non-structured matcher to filter candidates from a pool of memory items. We redundantly encode structured representations as content vectors, whose dot product yields an estimate of how well the corresponding structural representations will match. The second stage (FAC) uses SME to compute a true structural match between the probe and output from the first stage. MAC/FAC has been fully implemented and tested on dozens of examples. We show that MAC/FAC is capable of modeling patterns of access found in psychological data, and illustrative via sensitivity analysis that these results exhibit the desire dependence on theoretically important factors. The relationship of MAC/FAC to other models of memory is discussed, along with implications and possible extensions.

Descriptors :   *MEMORY(PSYCHOLOGY), *INFORMATION RETRIEVAL, MODELS, STRUCTURAL PROPERTIES, SENSITIVITY, MEMORY DEVICES, YIELD, PATTERNS, FILTERS, PSYCHOLOGY, RETENTION(PSYCHOLOGY), JUDGEMENT(PSYCHOLOGY).

Subject Categories : Psychology

Distribution Statement : APPROVED FOR PUBLIC RELEASE