[3Xin4] Poster session 1

[3Xin4-11] Analysis of sentiment words similarity network with ubiquitous community extraction

〇Fumitaka Iwaki¹, Takaomi Yokosuka¹, Tatsuji Takahashi¹ (1.Tokyo Denki university)

Keywords:association, network, clustering

Plutchik's emotional circle model has been used in many fields such as natural language processing and cognitive science, but its structural validity has not much examined. We consider a possible way of examining it through comparison, namely comparison with the structure of a similarity network of emotion words humans generate. Since the emotional circle includes various mixed emotions, soft clustering analysis is considered appropriate. We used a method called Markov chain modular decomposition (MDMC), which is capable of extracting ubiquitous communities. We conducted an experiment in which we asked participants about the similarity of all 48 ordered pairs of emotion words presented in the set of basic and applied emotions developed by Plutchik. Based on the results, we created a similarity network of emotion words, and analyzed its structural features. Furthermore, based on this analysis, we have examined the structural validity of the emotional rings by comparing them through ubiquitous community extraction.