Semi-Automatic Rule Learning Method Enabling Information Extraction for Ontology Population

Abstract

Business firms around the world have been generating enormous amounts of domain-related documents. Most of these firms are adapting semantic Web-based techniques into their software systems. Hence, they want to semantically enrich their documents to enable more meaningful querying or processing of the information in the documents. To impart semantics into these documents, ontologies relevant to the business domain should be used. In this context, to populate the domain ontology with the information from the source documents, a method for semi-automatic learning of extraction rules for populating the ontology is presented and implemented in the rule learning system. In addition to the rule learning system, a framework for separating the business logic from application logic and storing the business rules and extraction rules in external user-friendly format is presented in brief. The rule learning system is mainly developed to be a part of the presented framework, but it can be used as a standalone system to learn any decision or association rules too. The framework uses the rule learning system for learning extraction rules. The main idea behind the work presented is to learn extraction rules to be used by an information extraction system (part of the framework) to populate the domain ontology. The extraction rules learned by the rule learning system can be used with any business rules management system (BRMS) with appropriate wrappers to populate the domain ontology.