Unveiling Altered Concern in Women-Related Tweets: A Topic Modeling Analysis

In the dynamic realm of digital discourse, understanding the evolution of conversations is crucial. With the advent of social media, there has been increased attention to discussions on women’s issues and gender equality. A significant event in this domain is the 2017 protest against sexual harassment, which originated as a powerful social campaign. This paper analyzes the impact of this protest on the tweets surrounding women’s involvement and achievements, particularly in STEM (Science, Technology, Engineering, and Mathematics) disciplines. The primary challenge was data collection, given the constraints imposed by Twitter on this process. The research was initiated by deploying the BERTweet classification model, which effectively distinguished tweets related to women. The model’s performance was rigorously evaluated with a manually classified test set, and it has showcased impressive results of accuracy (99.65%), precision (99.50%), recall (99.79%), and F1 score (99.65%), respectively. Later the prominent topics were extracted by employing Structural Topic Modeling (STM). Statistical measures show that the topic concerning women had a noticeable change in proportion during the post-2017 period. To deepen insights, topic clustering and correlation analysis were exclusively conducted on women-related tweets. Overall, this paper provides a comprehensive view of the influence of a significant event on the evolving digital discourse.

Introduction

In the digital age, social media platforms have emerged as powerful spaces for public discourse, offering insights into societal trends, beliefs, and reactions to significant events. Twitter, in particular, serves as a dynamic microcosm of public opinion, reflecting the ebb and flow of conversations on various topics. Conversations can be influenced by various factors, such as time. For instance, examining discussions from a decade ago reveals a trend vastly different from today’s conversations. In fact, if we want to focus on any particular discussion, classification proves to be an essential tool. For analyzing the online discourse, several methods have been used till now.