Extraction of Medical Knowledge from Japanese Text

概要

論文の詳細を見る
With the increase in knowledge-based medical computer programs, automatic extraction of medical knowledge from natural language sources would be necessary. We studied the Japanese medical resources to find out the literal and syntactic patterns where the relationship between symptoms and their relevant body parts is represented. We initially assumed that the patterns can be classified in three ways. We evaluated our hypothesis by examining the sentences in the sections for clinical manifestations of infectious diseases from Japanese electronic medical textbook. Our experiments have shown 66% recall and 92% precision by using these three patterns only. Though our method is based on a simple technique, we think that our method would be useful to acquire various knowledge from Japanese medical textbooks in the future.
国立情報学研究所の論文