Files
Abstract
This thesis presents an ontology-based approach to automatic extractive summarization of text. Most of the extractive summarization systems so far have used statistical importance measures to determine importance of sentences. We use a knowledge-based approach which makes use of ontological knowledge to determine sentence importance. The Wikipedia ontology is the source of this knowledge. A sub-graph of the ontology is extracted after mapping the input document onto the ontology. The sub-graph, called the Thematic Graph, contains ontology concepts which match the terms in the document and edges from the ontology which represent relationships between the concepts. Hence, the thematic graph represents the theme of the input document. The thematic graph thus obtained is then analyzed using various graph-based importance measures to determine the relative importance of nodes. These values are used ultimately to decide which sentences are included in the summary for the document.0