Skip to main navigationSkip to main content
The University of Southampton
Centre for Risk Research

Gender Differences in Topics of Breakup Posts on Social Media Event

Time:
14:00 - 15:00
Date:
23 October 2020
Venue:
Online, MS Teams

For more information regarding this event, please email Dr Mario Brito at M.P.Brito@soton.ac.uk .

Event details

Abstract:

In this study, we explored the gender differences in terms of the words used in the relationship breakup posts on social media. To this end, in total 4,142 breakup posts were collected from Dcard(the biggest anonymous social media in Taiwan). We firstly used the TF-IDF algorithm to extract the key nouns from the posts of the males and females. The results showed that in the top 50 key nouns of each gender, 45 out of them were shared by both genders. The gender differences in the breakup posts cannot be revealed in the word level. Therefore, we tried to seek for any gender differences in the topic level. To this end, we applied the hierarchical Dirichlet process mixture model to extract the topics in the posts of each gender. For each gender, all nouns could be separated as different sets (i.e., topics), each of which was fit by a probability model (e.g., Beta distribution), generated from the base model in the Dirichlet process. In order to constrain the models generated for these two genders, each base model was generated from another common base model. The results showed that in total 10 topics were extracted. Among these topics, one topic was the biggest for either gender, which can be realized as a general description of the relationship breakup, including the nouns like friends, quarrels, home, pressure, mobile phones, ex-girlfriend, etc. There was one particular topic exclusively for the females, including the nouns like company, gifts, sick, breakfast, relationship, mood, etc. This topic implies that females particularly care about the mentally and physically caring from the boyfriend. The main findings are as follows. First, topics are better able to distinguish two genders than keywords, especially when they are talking about the same theme. Second, caring matters in relationship particularly for females.

Speaker Biography:

My name is Lee-Xieng Yang. I am now an associate professor at National Chengchi University in Taiwan. I received my PhD degree from University of Western Australia in 2004. The major research interest of mine is investigating human cognitive functions, including categorization, function learning, working memory, and forecasting. In order to understand how cognitive functions work, I do not only conduct behavioral experiments, but also develop computational models (i.e., neural network models or Bayesian models) to account for human cognitive functions. I also explore to what extent individuals’ performance in complex economic situation (e.g., earned profits in the games of beauty contest or double auction) can be explained by their fundamental cognitive functions (e.g., working memory capacity). Data science is another research interest of mine. Specifically, I am developing methods to analyze the data collected on the web pages and social media with the techniques such as web scraping, text mining (or text analysis), Chinese word segmentation, LDA, and other related ones. Recently, I have focused on applying the Dirichlet process in Bayesian analysis to solve some problems in text analysis (e.g., finding out the common and unique topics of different groups in their written documents for the same theme) and to establish a hierarchical model to account for higher level processing in categorization and function learning.

Recent papers and article:

Yang, L.-X., & Shu, C.-F. (2019). A text analysis approach to analyzing gender differences in breakup posts on social media, Chinese Journal of Psychology, 61, 209-230.

Yang, L-X. (2018). Application of Internet Methods in Psychology. In Chen, S.-H. (Eds.) Big Data in Computational Social Science and Humanities. NY, USA: Springer.

Tai, C.-C., Chen, S.-H., & Yang, L.-X. (2018). Cognitive ability and earnings performance: Evidence from double auction market experiments. Journal of Economic Dynamics & Control, 91, 409-440

Speaker information

Dr Lee-Xieng Yang,National Chengchi University,Associate Professor Department of Psychology

Privacy Settings