Performance Improvement of Drug Effects Extraction System from Japanese Blogs

Information disclosed to the public by patients is very important for people who are suffering from same illness because such information can be a source of knowledge and encouragement. Our aim is to make a system that extracts, organizes and visually represents information from patients' blogs. As the first step, the purpose of this paper is to extract descriptions of the effects caused by taking drugs as a triplet of expressions - drug name, object of change, and its effect - from illness survival blogs. However, conventional extraction methods are not suitable since these blogs are written in free natural language. Therefore, this paper proposes a method to extract the triplets using specific clue words and parsing the results. An evaluation experiment confirmed that medication usage information can be extracted with high accuracy using our proposed method, in comparison to existing methods. Moreover, recall was improved by combining our proposed method and a baseline system.