🤖 AI Summary
Real-world online political discussions exhibit naturally sparse controversy (only 12.9% controversial instances) and severe class imbalance, posing significant challenges for reliable controversy detection.
Method: We introduce the first naturally imbalanced Reddit controversy detection dataset contextualized to Indian political discourse. We propose a topological-aware text representation method grounded in persistent homology—the first application of Topological Data Analysis (TDA) to controversy detection—and design a class-imbalance-robust evaluation metric to systematically expose performance degradation of mainstream methods under realistic data distributions.
Contribution/Results: Experiments demonstrate that integrating topological features improves model F1-score by 12.4% and enhances robustness by 37% over state-of-the-art imbalance-learning baselines, establishing new performance benchmarks for controversy detection under natural imbalance.
📝 Abstract
The detection of controversial content in political discussions on the Internet is a critical challenge in maintaining healthy digital discourse. Unlike much of the existing literature that relies on synthetically balanced data, our work preserves the natural distribution of controversial and non-controversial posts. This real-world imbalance highlights a core challenge that needs to be addressed for practical deployment. Our study re-evaluates well-established methods for detecting controversial content. We curate our own dataset focusing on the Indian political context that preserves the natural distribution of controversial content, with only 12.9% of the posts in our dataset being controversial. This disparity reflects the true imbalance in real-world political discussions and highlights a critical limitation in the existing evaluation methods. Benchmarking on datasets that model data imbalance is vital for ensuring real-world applicability. Thus, in this work, (i) we release our dataset, with an emphasis on class imbalance, that focuses on the Indian political context, (ii) we evaluate existing methods from this domain on this dataset and demonstrate their limitations in the imbalanced setting, (iii) we introduce an intuitive metric to measure a model's robustness to class imbalance, (iv) we also incorporate ideas from the domain of Topological Data Analysis, specifically Persistent Homology, to curate features that provide richer representations of the data. Furthermore, we benchmark models trained with topological features against established baselines.