Abstract
Recently, the safety of use of medicine have attracted a lot of attention. It is, however, not necessarily that people share the same opinion, since a medicine has a relationship with many interest groups, such as doctors, pharmacists, nurses, and patients. In this paper, a text-mining method is applied to analize the free description part of the questionnairs about the safty of use of a transdermal patch.
First, a clustering algorithm is applied to the head-words in segments of sentences, which were obtained by morphological analysis of the answers. Using the results, we clarified the differences in viewpoint on the safety of drug-use between these groups at the word level.
Second, a method of extracting the sentences of the responses, which appear more than twice in the answers of the questionnairs is proposed. The multi-web graph based on the information obtained by Japanese dependency structure analysis provide the sentences by properly sorting and connecting the links between morphemes in the graph.
Last, this method is applied to answers by each group so as to find differences in beliefs at the sentence level and extract opinions and requests which each group has.
