An Event-Ontology-Based Approach to Constructing Episodic Knowledge from Unstructured Text Documents

Document summarization is an important function for knowledge management when a digital library of text documents grows. It allows documents to be presented in a concise manner for easy reading and understanding. Traditionally, document summarization adopts sentence-based mechanisms that identify and extract key sentences from long documents and assemble them together. Although that approach is useful in providing an abstract of documents, it cannot extract the relationship or sequence of a set of related events (also called episodes). This paper proposes an event-oriented ontology approach to constructing episodic knowledge to facilitate the understanding of documents. We also empirically evaluated the proposed approach by using instruments developed based on Bloom’s Taxonomy. The result reveals that the approach based on proposed event-oriented ontology outperformed the traditional text summarization approach in capturing conceptual and procedural knowledge, but the latter was still better in delivering factual knowledge .