Abstract
In recent years, lots of numerical and text data are easily stored on information systems, developing information technologies. Besides, it is required to utilize such data to ameliorate gualities of medical care, spreading evidence based medicine (EBM). To acquire scientific evidence for each medical care, people often carry out information retrieval for online documents with keywords of some interest. However, these techniques have not up to par to satisfy the requirement.
In this paper, we present a text mining method to extract clinical course knowledge from medical documents, which are written in natural language. To identify medical terms in these documents, we have done morphological analysis with MEID dictionary. Then, we have applied association rule learning method to obtain association rules. After obtaining association rules, we extracted some rules, which present clinical courses about neuro-physical diseases with characteristic relationshop between medical terms. With these terms, we evaluated their usefulness comparing with important vocablary by TFIDF method. The result shows that keywords from our method can identify substance of reterieved documents as same as keywords from TFIDF method.