Lee Gillam

Extracting Systems of Concepts from Text: Automatically Learning Ontologies Lee Gillam

This work investigates automating the population of knowledge bases with systems of concepts extracted from texts in arbitrary domains, normally undertaken manually by domain experts. It explores issues of terminology extraction from domain texts, the need for and use of knowledge representation, and the means by which terminology extraction and knowledge representation can be combined with international standards for terminology to produce such an initial model of an arbitrary specialist domain. A method is elaborated for identifying evidence of key domain concepts, expressed through terms used in place of and in relation to these concepts. The work presented may contribute to the Semantic Web and related initiatives by helping to overcome the well-documented and unsolved AI problem of producing an initial model of an arbitrary specialist domain from background resources without significant hand-crafting effort and involvement of a domain expert: the so-called "Knowledge Acquisition Bottleneck". This bottleneck is usually only overcome through extensive and expensive interactions with domain experts, involving a number of expert interviews.