抄録
To make reasonable plans for maintenance, analyzing past incident records is necessary. However, details of these records are usually written by natural texts and it is difficult to acquire information such as trends and distribution from them. In this paper, coding is introduced as a technique for adding semantic level structures to texts. The authors propose a text mining method for coding which facilitates users generating codes and annotating codes on sentences by showing sets of candidate sentences to users. The method utilizes clustering and machine learning techniques. The proposed method shows its efficiency thorough analyzing real incident data.
