Ontology Learning for Medical Digital Libraries

Ontologies play an important role in the Semantic Web as well as in digital library and knowledge portal applications. This project seeks to develop an automatic method to enrich existing ontologies, especially in the identification of semantic relations between concepts in the ontology. The initial study investigates an approach of identifying pairs of related concepts in a medical domain using association rule induction and inferring the type of semantic relation using the UMLS (Unified Medical Language System) semantic net. This is evaluated by comparing the result with manually assigned semantic relations based on an analysis of medical abstracts containing each pair of concepts. Our initial finding shows that the automatic process is promising, achieving a 68% coverage compared to manually tagging. However, natural language processing of medical abstracts is likely to improve the identification of semantic relations.