Analyzing Persian Twitter Sentiments on the Arbaeen Walk: A Comparative Study of LDA and BERTopic with the Arbaeen Tweets Dataset

Didehkhani, Zeynab

doi:10.22060/ajmc.2025.23870.1316

Analyzing Persian Twitter Sentiments on the Arbaeen Walk: A Comparative Study of LDA and BERTopic with the Arbaeen Tweets Dataset

Articles in Press

Document Type : Special Issue: MLKD 2024

Author

Zeynab Didehkhani

Independent Researcher, Tehran, Iran

10.22060/ajmc.2025.23870.1316

Abstract

As the world becomes increasingly interconnected, social media platforms

like Twitter play a pivotal role in shaping public discourse. One significant

topic that has emerged in recent years is the Arbaeen walk, a religious and political

movement that has attracted global attention. This paper has two main objectives:

first, to construct the Arbaeen Tweets Dataset, curated specifically for sentiment analysis;

and second, to analyze Persian tweets related to the Arbaeen walk from 2021 to

2022. By employing natural language processing (NLP) techniques—namely Latent

Dirichlet Allocation (LDA) and BERTopic—we aim to uncover prevalent themes and

sentiments, referred to as “topics” in topic modeling. The Arbaeen Tweets Dataset

consists of 2,622 colloquial Persian texts from Twitter, labeled for sentiment analysis.

Our findings indicate that while LDA demonstrates slightly superior quantitative

performance, BERTopic offers greater coherence in topic interpretation.

Keywords

Text Mining

Topic Modeling

Dataset construction

Foot-pilgrimage

Subjects