The experimental findings on ontology evolution with text resources

Main Researchers: 
Chunnan She
Research Areas: 
Ontology has emerged as one of the key components in knowledge intensive information systems and the semantic web. Ontologies need to evolve as the underlining knowledge domains change. High costs and skill requirements of ontology evolution motivate the development of a framework for assisting ontology engineers and domain experts in this task. We propose and attempt to validate a new method for ontology evolution with textual resources. The proposed method exploits the power of relation extraction from text and the view of the evolution of an ontology as graph operations. A preliminary experiment was designed with the corpus prepared from MEDLINE abstracts and a base ontology from a branch in the 2009 SNOMED CT international release. A gold standard based on a newer version of the same branch served as the comparison group. The prototype system based on the framework achieved 60% of the recall. The prototype also demonstrated that it is able to suggest the position which new concepts should be placed in the concept hierarchy. Finally, the sources of changes are preserved by displaying the source text of new concepts in the input corpus.