Abstract

This paper describes a heuristics-based approach in the semi-automated generation of Entity-Relationship (ER) diagrams for database modelling from natural language specifications and describes the implementation of such a system called ER-Converter.Though this is a semi-automatic transformation process, ER-Converter aims to require minimal human intervention during the process. ER-Converter has been evaluated in blind trials against a set of database problems. ER-Converter has an average of 90% recall and 85% precision. In terms of user intervention, ER-Converter requires very little human assistance with only 1.6% in the test dataset. The evaluation results are discussed and demonstrate that ER-Converter could be used, for example, within the domain model of a multimedia intelligent tutoring system, designed to assist in the learning and teaching of databases.