Agatha

v5 Finalist

Agatha

Synthetic text detection service on Telegram, gradient boosting with engineered features, model agnostic.

Send the bot text and it will reply with the probability that the text was machine generated. Telegram: @the_malone_bot Telegram web: t.me/the_malone_bot

About

Agatha is a synthetic text detection service available on Telegram Messenger. In the era of accessible and powerful large language models, agatha provides users with a tool to identify and combat generated misinformation, disinformation and low quality machine content on social media and the internet at large. Agatha is easy to use; send the bot text and it will reply with the probability that the text was machine generated. It uses an in-house trained gradient boosting classifier and several engineered features to estimate the probability that a given text was generated by an LLM. Agatha is written in Python using HuggingFace, scikit-learn, XGBoost, Luigi and python-telegram-bot, supported by Flask, Celery, Redis & Docker and served via Gunicorn and Nginx. Find agatha: Telegram handle: @the_malone_bot Telegram web link: t.me/the_malone_bot GitHub: github.com/gperdrizet/llm_detector