Browse Prior Art Database

MAPPING INPUT ONTO SCHEMAS

IP.com Disclosure Number: IPCOM000128614D
Original Publication Date: 1978-Dec-31
Included in the Prior Art Database: 2005-Sep-16
Document File: 19 page(s) / 68K

Publishing Venue

Software Patent Institute

Related People

Philip J. Hayes: AUTHOR [+3]

Abstract

Much recent work in artificial intelligence has been based on the assumption that it is useful to represent and use knowledge in coherent chunks. Such chunks of knowledge group together a number of more basic pieces of information in a way that is natural for the domain concerned. This trend was first identified by Minsky , who introduced the term frame to describe such chunks of knowledge. While Minsky's paper performed an important service in identifying the fundamental idea of pre-computed chunks of knowledge, it did not provide much analysis of the general issues that are raised by using large chunks of knowledge. It is the purpose of this paper to identify and explore some of those issues.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 5% of the total text.

Page 1 of 19

THIS DOCUMENT IS AN APPROXIMATE REPRESENTATION OF THE ORIGINAL.

MAPPING INPUT ONTO SCHEMAS

Philip J. Hayes

Department of Computer Science University of Rochester June 1978 TR 29 Much recent work in Artificial Intelligence has depended on the use of pre-stored descriptions of objects or events called frames or schemas to interpret "real-world" input. The interpretation of such input, whether of a visual, linguistic, or acoustic nature, is facilitated by mapping it onto some element of an appropriate schema, and assuming that the pre-stored information about that schema element applies also to the input. Two questions naturally arise: how to perform the mapping, and how to select the appropriate schema to match against. The first of these questions is considered in some detail; the second in as far as it interacts with the first.

Techniques for mapping single words are considered. Of the two primary techniques, one is characterised as associative and the other as predictive; it is claimed that a comprehensive system must use both types of technique. The techniques are then extended for mapping event descriptions; it is shown that the use of canonical representations of events is important for the extension of the associative technique. Next, the mapping problem is addressed for visual input; comparisons are made with the linguistic case; and it is suggested that lingu could benefit from some of the techniques that ar in the vision domain. Finally, schemas describing events or patterns of behaviour are considered. conclusion is that no one technique can solve the problem; rather, several quite different techniq levels of directness and cost must be used. istic mapping ise naturally very general The overall entire mapping ues at varying The preparation of this paper was supported in part by the Alfred P, Sloan Foundation under grant 74-12-5.

I Introduction

Much recent work in artificial intelligence has been based on the assumption that it is useful to represent and use knowledge in coherent chunks. Such chunks of knowledge group together a number of more basic pieces of information in a way that is natural for the domain concerned. This trend was first identified by Minsky , who introduced the term frame to describe such chunks of knowledge. While Minsky's paper performed an important service in identifying the fundamental idea of pre-computed chunks of knowledge, it did not provide much analysis of the general issues that are raised by using large chunks of knowledge. It is the purpose of this paper to identify and explore some of those issues.

Before tackling specific issues, we will try to make the scope of the paper a little more precise. Minsky explained knowledge clumping largely in terms of static descriptions of objects or situations; example frames included a visual description of a room from a certain point of view, or a description of a child's birthday party in terms of participants, decor, required accessories, etc. Cha...