[4Xin2] Poster session 2

[4Xin2-18] Classification of Characteristic Locations Extracted from Sentiment Analysis of Tweets

〇Daisuke Kikutani¹, Mitsuo Yoshida¹ (1.University of Tsukuba)

Keywords:Emotional information

This study addresses the evaluation of characteristics of various areas within Japan using emotional information
obtained from social media as a method. Previous research has focused on the relationship between emotional information
extracted from Twitter datasets and housing prices, reporting that areas with higher levels of happiness
correlate with property values, indicating that emotional information can be used to assess economic value. Therefore,
this research extracts areas with significant emotional fluctuations by combining tweet location and emotional
information, differentiating between positive and negative emotions, and using the calendar distinction between
weekdays and holidays for analysis. Applying ML-Ask to analyze emotional information from approximately 40
million tweets generated in 2022, areas with significant emotional fluctuations, both positive and negative, were
detected in urban and tourist locations. The results suggest the potential to extend the framework for estimating
the characteristics of specific regions from social media information through certain processing of emotional
information.