Projection in Conceptual Graphs Using Neural Networks

Published In

Northcon - Conference Record

Document Type

Citation

Publication Date

12-1-1995

Abstract

A technique, Transitive Supertype Graph, used to train a neural network as part of an overall system for processing of Conceptual Graph queries is outlined. Neural network techniques for CG applications rely on certain features of the problem context, particularly that the number of distinct concepts is reasonable size. Because the technique work on a much reduced form of the CG, topological distinctions between graphs containing same concepts are not used in the query process, so further processing must weed out erroneous matches.

Locate the Document

https://doi.org/10.1109/NORTHC.1995.485025

DOI

10.1109/NORTHC.1995.485025

Persistent Identifier

https://archives.pdx.edu/ds/psu/37319

Share

COinS