Tuesday , December 11 2018

Analysis Support System of Open-ended Questionnaires Based on Atypical and Typical Opinions Classification

Masanori AKIYOSHI
Graduate School of Information Science and Technology, Osaka University
Yamadaoka 2-1, Suita 565-0871, Japan

Keishi KIMURA
NTT West Corporation
Baba 3-15, Chuo-ku, Osaka, 540-8511, Japan

Hiroaki OISO
Codetoys Ltd.
Nishitenma 2-6-8, Kita-ku, Osaka 530-0047, Japan

Norihisa KOMODA
Graduate School of Information Science and Technology, Osaka University
Yamadaoka 2-1, Suita 565-0871, Japan

Abstract: This paper proposes a support system for analyzing answers to open-ended questions supplied by users as mobile game content evaluation when they unsubscribe the services. The answers include useful, unexpected opinions (atypical opinions) and expected opinions (typical opinions). It is inefficient to read them all during analysis. Therefore, we propose a support system for analysis of questionnaire data related to unsubscribing. The main function of the support system is classifying the answers into typical opinions and atypical opinions, and presenting them with user interfaces. In order to grasp user tendency, typical opinions are presented as a graph showing the number of transitions. Atypical opinions are presented as cards placement on 2-dimensional plane in order to grasp opinion groups with their content.

Keywords: Classification system, natural language processing, typical pattern, questionnaire.

>>Full text
CITE THIS PAPER AS:
Masanori AKIYOSHI, Keishi KIMURA, Hiroaki OISO, Norihisa KOMODA, Cost-benefit Analysis of Decentralized Ordering on Multi-tier Supply Chain by Risk Simulator, Studies in Informatics and Control, ISSN 1220-1766, vol. 18 (3), pp. 195-204, 2009.

1. Introduction

Mobile phones and the Internet are frequently and increasingly used as marketing mediums. Users of a mobile game content system respond to a questionnaire that is intended to improve contents when they stop their subscription to the services. The questionnaire consists of multiple-choice and open-ended questions that appear on the screen of their cellular phones. Using multiple-choice questions, the questionnaire asks users to choose from some possible answers and responses that can be statistically analyzed.

Answers to the multiple-choice questions, however, are completely standardized and known from the service provider’s perspective. To tap into unexpected ideas, answers to open-ended questions must be analyzed.

There are no restrictions on open-ended questions and they are generally answered with natural language.

Therefore, the answers include an enormous amount of text data for the questionnaire analyst, and some support methods for analysis by text mining have been proposed [1][2][3][4].

However, since answers are input through cellular phones, they often include many symbols that are dependent on various kinds of terminals and grammatical mistakes, which can make them hard to understand. Additionally, since most answers are included in multiple-choice questionnaires and do not need to be read, there are a few useful opinions. Based on our previous research, a method of classifying opinions as typical or atypical is proposed [5][6], and a support system that presents two interfaces for analyzing each opinion group is designed [7].

The interface that supports the analysis of typical opinions has bar graphs that present the trend in the number of opinions in each category that includes the same opinion content. However, since the number of categories increases along with the number of opinions, the graph gets complicated. Therefore, it is necessary to freely group some categories. The function allows an analyst to discover a subscriber’s reason for unsubscribing.

The interface that supports the analysis of atypical opinions presents opinions on screen as cards in order to grasp the opinion content intuitively. Especially, this paper proposes a support system for discovering correlations between a subscriber’s properties and the singularity of their opinions.

REFERENCES

  1. YAMANISHI, K., H. LI, Mining open answers in questionnaire data, IEEE Intelligent Systems, vol.17, no. 5, 2002, pp.58-63.
  2. ROMANO, N. C. Jr., BAUER, C., CHEN, H. and NUNAMAKER, J. F. Jr., The mindmine comment analysis tool for collaborative attitude solicitation, analysis, sense-making and visualization, in Proceedings of the 33rd Hawaii Int. Conf. on System Sciences (in CD-ROM), 2000.
  3. CHEN, H., O TITKOVA, R. ORWIG, and NUNAMAKER, J. F. Jr., Information visualization for collaborative computing, IEEE Computer, vol. 31, no. 8, 1998, pp.75-82.
  4. FRASCONI, P., M. GORI, and G. SODA, Data categorization using decision trellises, IEEE Tran. on Knowledge and Data Engineering, vol. 11, no. 5, 1999 pp.697-712.
  5. HIRAMATSU, A., S. TAMURA, H. OISO and N. KOMODA, Method for atypical opinion extraction from answers in open-ended questions, in Proceedings of IEEE Int. Conf. on Computational Cybernetics (ICCC 2004) (in CD-ROM), 2004, pp.273-278.
  6. HIRAMATSU, A., S. TAMURA, H. OISO, and N. KOMODA, Support system for analyzing open-ended questionnaires data by culling typical opinions, 2004 Int. Conf. on Systems, Man and Cybernetics (SMC2004), 2004, pp.1377-1382.
  7. HIRAMATSU, A., H. OISO, T. SHOJIMA, and N. KOMODA, Man-machine interface of a support system for analyzing open-ended questionnaires, in Proceedings of IEA/AIE 2005, 2005, pp. 318-320.
  8. MATSUMOTO, Y., A. KITAUCHI, T. YAMASHITA, Y. HIRANO, H. MATSUDA, K. TAKAOKA, and M. ASAHARA, Japanese morphological analysis system ChaSen version 2.2.1, http://chasen.aist-nara.ac.jp/chasen/doc/chasen-2.2.1.pdf, 2000.