Abstract
Companies often carry out questionnaire(s) in order to design marketing strategies or to grasp consumers' trends. It has been difficult to design a set of questions properly. For example, if respondents take the meaning of two or more questions in a questionnaire as same/similar, these questions could be redundant. It is, however, difficult to know which meaning respondents will take beforehand. On the other hand, it is possible to analyze the views of respondents to the questions, and how the set of questions is proper after the questionnaire. This paper proposes a method of N-mode Singular Value Decomposition(SVD) for cosine similarity tensor and shows that respondents' understanding of meanings of questions, which are not found by the conventional method, can be discovered.
