Research Title: Semantic Information Retrieval for Geoscience Resources
Start Date: October 2014
Various specialist domains like the geosciences have seen ontologies developed over the years. Ontology provides a specification of concepts in a domain. Its development (mostly manual) is tedious and time consuming. Yet ontologies have not attained expected levels of use especially with respect to information retrieval on the Semantic Web. Attempts to use domain ontologies to enhance information retrieval (e.g. ontology based query expansion) have yielded mixed results. While it improves retrieval in certain instances, it has not shown significant improvements over a set of retrieval tasks. Developing an effective way to utilise domain ontology for information retrieval forms the basis of this research. In ontology based information retrieval, it is already known that performance improves as overlap between terms in ontology concepts and target document collection increases. This research will also investigate ways to provide broader coverage of domain concepts through discovery of new terms to include in ontology and to align existing ontologies. The British Geological Survey are collaborators on this research which they also part-fund.
My main research objectives are:
- To understand situations where use of domain ontologies is detrimental to information retrieval.
- To develop novel approach to ontology based information retrieval that minimises the negative effects of its application.
- To evaluate the extent to which ontology based information retrieval approach developed can be applied to other domains.
- To investigate ways to achieve broader overlap between concepts described in ontology and domain vocabulary expressed in document collection of a search space.