An Inductive Logic Programming-Based Approach for Ontology Population from the Web

Developing linguistically data-compliant rules for entity extraction is usually an intensive and time-consuming process for any ontology engineer. Thus, an automated mechanism to convert textual data into ontology instances Ontology Population may be crucial. In this context, this paper presents an inductive logic programming-based method that induces rules for extracting instances of various entity classes. This method uses two sources of evidence: domain-independent linguistic patterns for identifying candidates of class instances, and a WordNet semantic similarity measure. These two evidences are integrated as background knowledge to automatically generate extractions rules by a generic inductive logic programming system. Some experiments were conducted on the class instance classification problem with encouraging results.