Open-sourced Indic profanity detector covers 12 Indian languages, runs at 4ms per comment on CPU
sudalairajkumar · x · 2026-11-19
CobuildX-AI open-sourced an Indic profanity detection model family (Apache-2.0), built as step one for filtering text before fine-tuning Indic LLMs.
- Coverage: 12 Indian languages + English, handling native scripts, romanized text, and code-mixed input; based on mmBERT-small.
- Three sizes (CPU inference): 17M at 4ms per comment (F1 0.88), 36M at 8ms (F1 0.89), 58M at 14ms (F1 0.89).
- Takeaway: the 17M model is 3.5× faster than 58M for only 0.01 less F1; all sizes beat a keyword list (F1 0.70), with 3.5–4% false-positive rates.
More from Research
- SE3 Labs Releases First Open YAM Real-Robot Teleop Dataset: 500 Bimanual Episodes — paigeinsf · 2026-11-19
- 1.5 years under review: researchers urge releasing AI papers immediately after submission — carlosdponx · 2026-11-19
- Berkeley's StreamEmbed indexes your screen history on-device, 3.7x faster at matched accuracy — berkeley_ai · 2026-11-19
- NAACL RAF opens 2026-2027 grant proposals for NLP initiatives, due May 6 — keviv9 · 2026-11-19
- MAVE reconstructs editable 3D blockouts from film shots, works inside Blender — bdsqlsz · 2026-11-19
- NAACL 2027 Student Research Workshop opens CFP, papers due January 11, 2027 — keviv9 · 2026-11-19