G. Héja, M. Gergely, G. Surján, P. Varga, and A. Szekeres (Hungary)
Knowledge Representation, Natural Language Processing, Medical Documentation, Medical Ontologies
431 oto-rhino-laryngological discharge reports have been automatically transformed to structured XML files. Domain ontology of anatomical concepts of the human ear - based on SNOMED - has been created and is being integrated to OpenGalen. To identify these concepts in the discharge reports syntactical analysis and word pattern matching methods have been used. In the best case 66% recall and 97% precision could be achieved. According to our experiments the general Hungarian grammar has to be augmented with special rules to be able to cope with medical language. By evaluating the assertions about the concepts in the reports, the attributes of the concepts have been collected. Both statistical and NLP based analysis has been used for this purpose.
Important Links:
Go Back
