cbx-indic-profanity-detector-36M Copyright 2026 Cobuild Tech Private Limited. Licensed under the Apache License, Version 2.0 (see LICENSE): the model weights, configuration, tokenizer, scripts and documentation in this repository. Provided "as is", without warranties or conditions of any kind (LICENSE, sections 7 and 8); see "Intended use and disclaimer" in README.md. This model contains or was built from the following third-party materials: 1. mmBERT (jhu-clsp/mmBERT-small, revision abc32620dd4f6ab06f5fbe905dc25f310618e09f), the base model (its vocabulary was pruned and, for the smaller sizes, layers removed). Copyright (c) 2025 the mmBERT authors (Marone, Weller, Fleshman, Yang, Lawrie, Van Durme). MIT License: Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. 2. Training text (not distributed in this repository). The model was trained on text from these datasets, whose authors we thank and credit: - MACD (Multilingual Abusive Comment Dataset), ShareChat / Moj comments, ShareChat (hi, ta, te, ml, kn): https://github.com/ShareChatAI/MACD. Licence: no licence file in the repository; used with attribution. - DravidianCodeMix offensive language identification (OffensEval Dravidian), Chakravarthi et al. (ta, ml, kn): https://huggingface.co/datasets/community-datasets/offenseval_dravidian. Licence: CC BY 4.0. - BD-SHS, Romim et al., LREC 2022 (bn): https://www.kaggle.com/datasets/naurosromim/bdshs. Licence: CC BY 4.0. - Bengali Hate Speech, Karim et al. (bn): https://huggingface.co/datasets/rezacsedu/bn_hate_speech. Licence: MIT. - Telugu-Hatespeech, Marreddy et al., ACM TALLIP 2022 (te): https://huggingface.co/datasets/mounikaiiith/Telugu-Hatespeech. Licence: CC BY 4.0. - MOLD (Marathi Offensive Language Dataset) 1.0 and 2.0, Gaikwad, Zampieri et al. (mr): https://github.com/tharindudr/MOLD. Licence: CC BY 4.0. - RUHSOLD (Roman Urdu Hate-Speech and Offensive Language Dataset), Rizwan et al., EMNLP 2020 (ur (romanized)): https://github.com/haroonshakeel/roman_urdu_hate_speech. Licence: MIT. - Hate Speech and Offensive Language, Davidson et al., ICWSM 2017 (en): https://github.com/t-davidson/hate-speech-and-offensive-language. Licence: MIT. - HateXplain, Mathew et al., AAAI 2021 (en): https://github.com/hate-alert/HateXplain. Licence: CC BY 4.0. - FineWeb-2 (web text), Hugging Face (all 12 Indic languages): https://huggingface.co/datasets/HuggingFaceFW/fineweb-2. Licence: ODC-By 1.0. No dataset text is distributed with the model. Use of the model does not imply endorsement by any of these authors or publishers. 3. Labels. Part of the training labels were produced by large language models: Gemma 4 31B (Apache-2.0), Claude Haiku 5.5 and Claude Sonnet 5.5 (Anthropic), and gpt-5.4-mini (OpenAI).