Restoring the intended structure of Hungarian ophthalmology documents

摘要

Clinical documents have been an emerging target of natural language applications. Information stored in documents created at clinical settings can be very useful for doctors or medical experts. However, the way these documents are created and stored is often a hindrance to accessing their content. In this paper, an automatic method for restoring the intended structure of Hungarian ophthalmology documents is described. The statements in these documents in their original form appeared under various subheadings. We successfully applied our method for reassigning the correct heading for each line based on its content. The results show that the categorization was correct for 81.99% of the statements in our testset, compared to a human categorization.