zake7749/DeepToxic
top 1% solution to toxic comment classification challenge on Kaggle.
Combines multiple preprocessing strategies (spelling correction, POS tagging, heavy cleaning) and embedding sources (FastText, GloVe variants) with ensemble methods across word-level and character-level RNNs/CNNs. Uses Pooled RNN and K-max CNN architectures with aggressive dropout and batch normalization for regularization, plus a binary meta-label ("bad_comment") to increase training signal diversity. Final predictions leverage low-correlation char and word-level models through bagging to achieve robust toxic comment detection.
191 stars. No commits in the last 6 months.
Stars
191
Forks
68
Language
Jupyter Notebook
License
MIT
Category
Last pushed
Feb 06, 2019
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/zake7749/DeepToxic"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related frameworks
aralroca/react-text-toxicity
Detect text toxicity in a simple way, using React. Based in a Keras model, loaded with Tensorflow.js.
charliegerard/safe-space
Github action that checks the toxicity level of comments and PR reviews to help make repos safe spaces.
jaydeepjethwa/DeTox
A web-app to identify toxic comments in a youtube channel and delete them.
DenisIndenbom/AntiToxicBot
AntiToxicBot is a bot that detects toxics in a chat using Data Science and Machine Learning...
AbhishekChamoliDeveloper/text-toxicity-detector
This library can detect toxicity in the text or string or content and in return provide you the...