In recent years, the analysis of approval desire in social networking services (SNSs) has garnered significant research attention. However, identifying specific words that influence the perception of approval desire remains challenging. Additionally, limited research has objectively examined whether posts evoke approval desire from the readers' perspective. This paper introduces a framework for identifying and quantifying approval desire in SNS posts. Machine learning, statistical analysis, and text mining techniques were employed to analyze the presence of approval desire. Two metrics were developed to quantify approval desire: the proportion of posts exhibiting approval desire and the strength of approval desire per tweet. Additionally, linguistic feature analysis was conducted to identify approval desire in SNS posts, and statistical hypothesis testing was used to examine post characteristics. In this study, a total of 7078 tweets were collected from 20 Japanese Twitter posting users, of which 1200 were used as training data. Comprehensive studies on Twitter posts reveal that the proposed machine learning approach achieves a maximum coincidence rate of approximately 67%. Moreover, although the posting frequency of the top five individuals associated with high approval desire is lower, their posts are characterized by a higher number of words. These findings suggest that classifying tweets using machine learning and quantifying approval desire provide valuable insights into the characteristics of post content and the behavior of posters in this context. • Approval desire strength per tweet serves as a quantification metric. • SNS posters with stronger approval desire post less frequently but with more words. • Quantifying approval desire provides insights into posting behavior. • Linguistic analysis of SNS posts reveals characteristic content features.
Murata et al. (Fri,) studied this question.