Investigation on Applying Modular Ontology to Statistical Language Model for Information Retrieval

Abstract

The objective of this research is to provide a novel approach to improving retrieval performance by exploiting Ontology with the statistical language model (SLM). The proposed methods consist of two major processes, namely ontology-based query expansion (OQE) and ontology-based document classification (ODC). Research experiments have required development of an independent search tool that can combine the OQE and ODC in a traditional SLM-based information retrieval (IR) process using a Web document collection.
This research considers the ongoing challenges of modular ontology enhanced SLM-based search and addresses three contribution aspects. The first concerns how to apply modular ontology to query expansion, in a bespoke language model search tool (LMST). The second considers how to incorporate OQE with the language model to improve the search performance. The third examines how to manipulate such semantic-based document classification to improve the smoothing accuracy. The role of ontology in the research is to provide formally described domains of interest that serve as context, to enhance system query effectiveness.