Discovering and Categorising Language Biases in Reddit

Understand

We present a data-driven approach using word embeddings to discover and categorise language biases on the discussion platform Reddit.

• As spaces for isolated user communities, platforms such as Reddit are increasingly connected to issues of racism, sexism and other forms of discrimination.
• Hence, there is a need to monitor the language of these groups.
• One of the most promising AI approaches to trace linguistic biases in large textual datasets involves word embeddings, which transform text into high-dimensional dense vectors and capture semantic relations between words.