Towards Safer Online Platforms: Explainable and Adversarial-Resistant Toxic Comment Detection
Bibliographic record
Abstract
Toxic content on online platforms—such as hate speech, harassment, and discrimination—continues to threaten the safety and inclusivity of digital spaces. These harmful interactions negatively impact user experience and highlight the need for automated, reliable moderation tools. In this study, we present a robust and interpretable machine learning framework for toxic comment classification using a fine-tuned BERT model. To enhance transparency, we integrate SHAP-based explanations that highlight which words contribute to a classification. To improve robustness, we incorporate adversarial training with obfuscated toxic samples (e.g., misspellings or symbol substitutions) that mimic real-world evasion tactics. Additionally, we implement a rule-based filter to flag borderline cases—especially those involving ambiguous terms—for manual review, reducing false positives. A user-friendly Streamlit interface allows real-time interaction with the model, providing both predictions and visual explanations. Experiments on the Jigsaw Toxic Comment Classification datasets show that our approach outperforms standard baselines in accuracy, interpretability, and resilience to adversarial input.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.018 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".