Abstract
“Flaming” phenomena are increasing with the spread of self-committed media. In our previous paper, the quantitative method to evaluate time-series of “Flaming” phenomena on internet bulletin boards (BBS) was proposed. And it is suggested that “Thread” was more effective than physical time as a time unit for analyzing temporal fluctuations in the number of postings on BBS. But one of the issues with this method was the difficulty in distinguishing “Flaming” phenomena from activated discussion when postings were increased. This issue can be solved by the positive/negative discrimination of posted contents. And the polarity dictionary is used for this discrimination. However, many existing dictionaries do not reflect colloquial expressions. For this reason, these dictionaries are not suitable for discrimination of the contents posted on BBS in which such expressions are used. Therefore, in order to improve the accuracy in temporal analysis of “Flaming” phenomena, this study aimed to propose a polarity dictionary suitable for evaluation of contents posted on BBS.
This proposed dictionary was created based on the human impression evaluation of 8,775 words by 20 participants to analyze colloquial expressions. In order to evaluate the accuracy of this dictionary, the “Flaming” phenomena described in our previous paper were analyzed. In this paper, flaming index was defined as quantitative evaluation scale. First, for smoothing the time-series of “Flaming”, moving average of the flaming index by 20 “Thread” was made. Next, positive/negative index (PN index) was calculated by positive/negative discrimination using this proposed dictionary and the existing dictionary. Then, the correlation between the flaming index and the PN index based on each dictionary was analyzed, and these correlation coefficients were compared. As a result, there was a strong significant correlation between PN index based on this proposed dictionary and the flaming index. On the other hand, there was a week correlation between PN index based on existing dictionary and the flaming index. From the above results, it is suggested that “Flaming” phenomena can be analyzed more accurately using this proposed dictionary.
