Entity-Relationship Modelling for Semantic Networks

摘要

Entity-Relationship modelling was used to knowledge engineer a structure for data extracted from biological domain. This analysis created a Semantic Network which used nodes to represent concepts and links to represent relationships between them in an integrated structure. A small set of domain dependent relationships were identified, as were a number of different concept categorise, and these are shown to be capable of representing a wide variety of domain knowledge. The generated structure, reflecting the structure of data from this domain, should provide a better support for complex reasoning tasks, serving as a knowledge base for such tasks as natural language understanding and machine translation. Only semantically correct combinations of network components are permitted within the network, valid and invalid combinations being identified by the modelling process. E-R modelling also aided the identification of inference rules which infer new semantic structures and are based on network structure as opposed to domain knowledge. Finally, we show that the network structure identified is of the type developed by Thomason and Touretzky.