Trained models & code to predict toxic comments on all 3 Jigsaw Toxic Comment Challenges. Built using ⚡ Pytorch Lightning and 🤗 Transformers. For access to our API, please email us at contact@unitary.ai.

SFU Opinion and Comments Corpus

A Python package to compute HONEST, a score to measure hurtful sentence completions in language models. Published at NAACL 2021.

The official code to reproduce results from the NACCL2019 paper: White-to-Black: Efficient Distillation of Black-Box Adversarial Attacks

Module for predicting toxicity messages in Russian and English

Implementing Simulation-guided Beam Search for text generation. Idea already proposed in Choo et al for Neural Combinatorial Optimization. In this repo, we use the technique for Controllable Text Generation.

Simple Multi-Language HTTP Server Text Toxicity Detector

A Drug Metabolite & Toxicity Property Predictor Based on Graph Neural Network

[ICWSM 2024] Decentralised Moderation for Pleroma and the Fediverse

For a human-centered data science assignment, I analyzed how Google's Perspective API tool detects and categorizes Black language data, also known as AAVE (African-American Vernacular English), from Twitter.

A simple Flask API which detects toxicity in text

An approach of measuring Unintended Bias in Toxicity Classification. This approach can be used to evaluate various toxic comments and messages in various public forums and talk pages. We also demonstrate how imbalances in training data can lead to unintended bias in resulting models, and therefore potentially unfair applications. We have used a …

Reproduction study of the paper "Detoxifying Text with MaRCo: Controllable Revision with Experts and Anti-Experts"

🤖 Intelligent AI Agent for Real-time Content Moderation 97.5% accuracy | Multi-stage ML pipeline | Production-ready Zero-tier filtering + Embeddings + Fine-tuned BERT + RAG

A modern, multi-modal hate speech detection web app using the Perspective API. Analyze text, images, audio, and video for toxic or harmful content in a user-friendly interface.

Python library and Orange3 widget for high-throughput screening data preprocessing, toxicity scoring, and FAIRification

Custom toxicity checks for bight

Toxicity evaluation on DeepSeek R1 and DeepSeek API

How to make multiple requests per second to Perspective API

This is the api that checks whether the given text is toxic or not. It also gives the measure of toxicity whether it is toxic, severely toxic, obscene, threat, insult or identity hate. The models were trained on the dataset from kaggle toxic comment classification challenge.