A Transformer-Based Approach to Analyzing Public Opinion and Political Trends
Published in Proceedings of the 10th International Conference on Smart and Sustainable Technologies (SpliTech) - Volume 1, 1-6, 2025
This paper is publicly available here.
Authors:
Jon Gardeazabal Gutiérrez\(^\dagger\) \(\cdot\) Miguel Fernandez-de-Retana\(^\dagger\) \(\cdot\) Aritz Bilbao-Jayo
\(^\dagger\) These authors contributed equally to this work.
Keywords:
Political Discourse \(\cdot\) Content Analysis \(\cdot\) Natural Language Processing \(\cdot\) Public Opinion
Abstract:
Podcasts have become a significant platform for political discussion and a reflection of public opinion. This paper details the development of an NLP-driven tool designed to analyze political discourse in podcasts, with applications in smart city governance and public opinion research. The tool employs automated transcription and a RoBERTa-based classification model, trained on the Manifesto Project dataset, to categorize political topics. BERTopic is used for topic modeling, providing a structured overview of key themes. A use case illustrates the tool’s effectiveness in extracting dominant political topics from podcast episodes, demonstrating its potential to provide valuable insights for urban policy and public engagement.