A Investigation of Feminism Trends Through Sentiment Analysis Using Machine Learning and Natural Language Processing

Abstract

Introduction/Importance of Study:

One of the recent changes seen in Pakistan is the growing awareness among people, to end gender discrimination and bring equality, across various spheres. “Aurat March”, is a series of rallies that began in 2018 to mark International Women’s Day. People across the nation, comment on these rallies through social media.

Novelty Statement:

The response to the “Aurat March”, held annually since 2018, was mixed and no analysis of Twitter data had been previously done to investigate the polarity of comments, through Machine learning and Natural Language Large Language Models.

Material and Method:

For this, Sentiment analysis was performed, using Machine learning and NLP techniques, on the pre-processed data. Lexical rule-based VADER and transformer-based pre-trained Large Language Models were used to check the polarity of Twitter comments.

Results and Discussion:

The best results were achieved through RoBERTa-LARGE, which were closest to the Human Labelled Data, hence validating the accuracy of the LLM. On the other hand, VADER results were clearly far from the manually labeled results. The sentiment analysis that was applied the first time on “Aurat March” tweets, gave us satisfactory results, and we were further able to validate our research, by comparing the models’ accuracy with human-labeled results. Consequently, by analyzing the sentiments expressed on Twitter, we were able to discern the general mindset of users and gain insights into prevailing trends.

Conclusing Remarks:

This analysis provided us with a reasonably accurate gauge to assess the perception of feminism over the past few years, allowing us to evaluate whether it has garnered fame or faced defamation in the public discourse.