Abstract
In this study, the authors developed a text mining technique including Document Frequency-Inverse Minutes Frequency (DFIMF), a new word selection method, to grasp changes in discussion from the minutes of public work planning processes. As the result of applying the developed technique together with Term Frequency-Inverse Document Frequency (TFIDF) to the minutes of the Yodo River Watershed Committee, it was first illustrated that DFIMF was more effective than TFIDF in terms of comprehending contents of discussion in the committee. It was also revealed that the themes of discussion in the committee had shifted first from "water demand management", "citizen participation" and "ecosystem integrity" to "dams" and "flood control", then "citizen participation" again, and "estimated high-water level" and "flood control", being finally followed by "review of the plan's progress".