4chan & 8chan embeddings

TL;DR

This paper presents a large dataset of message embeddings from 4chan and 8chan's /pol/ boards, aimed at studying toxic language and improving hate speech detection.

Contribution

It introduces a comprehensive collection of 30 million messages and trained word embeddings, publicly released for research on toxic discourse.

Findings

Embeddings capture toxic language patterns

Dataset enables improved hate speech detection

Resource available for further research

Abstract

We have collected over 30M messages from the publicly available /pol/ message boards on 4chan and 8chan, and compiled them into a model of toxic language use. The trained word embeddings (0.4GB) are released for free and may be useful for further study on toxic discourse or to boost hate speech detection systems: https://textgain.com/8chan.