Telling AI Data Stories: combining insights from academic literature and RAG techniques to guide prompt design

This Google Colab notebook contains the full implementation of a Retrieval-Augmented Generation (RAG) pipeline used in the study <i>“</i>Telling AI Data Stories: combining insights from academic literature and RAG techniques to guide prompt design<i>”</i>. It includes automatic PDF text extraction, passage chunking, semantic search using sentence embeddings and FAISS, and prompt generation for ChatGPT-4. In addition to the automated pipeline, the notebook also includes examples of manually selected passages used to construct prompts. It is designed to support reproducibility and allow other researchers to adapt the approach with different articles or datasets.