Enriching ontologies with encyclopedic background knowledge for document indexing

Abstract

The rapidly increasing number of scientific documents available publicly on the Internet creates the challenge of efficiently organizing and indexing these documents. Due to the time consuming and tedious nature of manual classification and indexing, there is a need for better methods to automate this process. This thesis proposes an approach which leverages encyclopedic background knowledge for enriching domain-specific ontologies with textual and structural information about the semantic vicinity of the ontologies’ concepts. The proposed approach aims to exploit this information for improving both ontology-based methods for classifying and indexing documents and methods based on supervised machine learning.