Augmenting Science Journalism
Inside the SMC Lab, we’re not just observing science journalism; we’re actively shaping it. Serving as an essential segment of the Science Media Center Germany, we’re pioneering advancements in how scientific discoveries and understanding reach the public.
This blog is your front-row seat to our work, offering a firsthand look at innovative projects, profound research, and our commitment to making science and news accessible for all.
-
|
Build a Naïve and an Advanced RAG System – and Evaluate It
In the previous post, we laid the foundation of our RAG system. Based on the use case, we created a vector store, generated a synthetic test set, and defined the metrics we’ll use to evaluate performance. In this fifth part, we will build two pipelines: a simple naïve baseline and a more advanced extension. We…
-
|
Laying the Groundwork: Use Case, Vector Store, Synthetic Data, and Metrics
In the previous posts, we covered the basics: what RAG is, why it matters for journalism, and how a typical pipeline works. In the next three parts, we’ll move from concepts to code and show how a RAG bot is put together. As outlined in the previous post, when we implemented our pipeline, we followed…
-
|
How to Use RAG in Journalism?
In the previous blog post, we looked in detail at how a RAG system works step by step. In this post, we want to shift the perspective: how can RAG actually be used in the context of journalism, and especially in science & data journalism? We will look at specific use cases, discuss the trade-offs…
-
|
How Exactly Does RAG Work?
Now that we know why RAG is needed and how it helps to reduce the risk of hallucinations by compensating for the inherent limits of LLMs, we can take a closer look at how it actually works. In this series, we describe a RAG pipeline that can be broken down into four main stages: knowledge…