Hierarchical clustering of free form questionnaires using domain knowledge

In this paper, we propose a new statistical analysis method of free form questionnaires for discovering unexpected themes by artificial intelligence technique. Then, we use hierarchical clustering method through similarity measure based on domain knowledge. We applied our method to an actual data of a free form questionnaire to university students, and we obtained that the semantic average metrics between every pairs of two words within the same clusters are condensed more than conventional one without domain knowledge.