Reproduce “LLM-Assisted Topic Reduction for BERTopic on Social Media Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wannes Janssens, Matthias Bogaert, Dirk Van den Poel
A framework integrates BERTopic with Large Language Models (LLMs) to address the issue of excessive and overlapping topics generated from noisy social media data. This approach consistently enhanced topic diversity and often improved coherence compared to traditional reduction methods, yielding more interpretable topic models.
Key insights
- LLM-based topic reduction consistently and significantly enhanced topic diversity across diverse social media datasets compared to BERTopic's built-in reduction methods.
- The impact on topic coherence was more nuanced; while often improved, it was sensitive to dataset characteristics, sometimes underperforming for highly sensitive content.
- Iterative LLM prompting offers finer control over the number of topics and effectively manages LLM input token limits and processing complexity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.