Using Retrieval Augmented Generation to Build the Context for Data-Driven Stories

Data Storytelling (DS) is building data-driven stories to communicate the result of a data analysis process effectively. However, it may happen that data storytellers lack the competences to build compelling texts to include in the data-driven stories. In this paper, we propose a novel strategy to enhance DS by automatically generating context for data-driven stories, leveraging the capabilities of Generative AI (GenAI). This contextual information provides the background knowledge necessary for the audience to understand the story's message fully. Our approach uses Retrieval Augmented Generation (RAG), which adapts large language models (LLMs), the core concept behind GenAI, to the specific domain required by a data-driven story. We demonstrate the effectiveness of our method through a practical case study on salmon aquaculture, showcasing the ability of GenAI to enrich DS with relevant context. We also describe some possible strategies to evaluate the generated context and ethical issues may raise when using GenAI in DS.