RAG

RAG Chatbot for 4chan data analysis.
Published

July 5, 2025

Analysis with BERT and RAG Chatbot

For detecting sentiments, topics, and previous conversations for crime analysis, I did the following:

  1. Trained BERT for recognizing possible tones.
  2. Developed a Retrieval Augmented Generation (RAG) Chatbot which is able to fetch similar messages from an indexed database and answer based on that context.

This makes text analysis very easy as it can find patterns based on vector embedding of text. Additionally, it can caption images too and describe connections.

We can automatically scrape 4chan too and perform automatic analysis. It prints a summary and topics discussed in the end. The default board is /pol/ on 4chan.

or we can upload images and chat with it

here model is zephyr 7 billion parameters ,finetuned on mistral.