Abstract
Identifying users' implicit needs has been recognized as a key factor of product success. Thus, the importance of observing user experience (UX) for comprehensive understanding of users has been continuously emphasized. However, observing user behavior and analyzing collected data require abundant time and effort. Also, there has been little success in developing efficient techniques to assess user experiences. The aim of this paper is to analyze and interpret text data which was collected by user’s natural language more effectively and efficiently using SOM (Self-Organizing Map). Diary methods were conducted to collect users' emotional experiences on smart TV by users' natural language with twenty participants for one week. They were especially asked to describe their emotions why they happened in a certain episode. As a result, a total of 311 episodes was collected, and 268 words remained after preprocessing. Based on the co-occurrence matrix, SOM calculated the codebook which represents similarity among episodes. Similarity between episodes could be identified by analyzing a two-dimentional map, and 15 groups were classified. Characteristics of each group could be represented by a frequency of words. From the results of this paper, major patterns of TV-viewing could be identified through SOM. Moreover, implicit user values were identified by analyzing users' emotional experiences, which will help in developing both new and useful functions for the smart TV.
