I attempted and failed creating one—nevertheless it was when LLMs weren’t a factor!
At first, data graphs (KGs) sounded daunting—not the idea, however the strategy of establishing one.
I’ve tried establishing a data graph earlier than and failed.
Graphs are undoubtedly the most effective methods to symbolize advanced relationships. They’ve many makes use of, corresponding to suggestion techniques and fraud detection. Nonetheless, the one which caught my curiosity probably the most was info retrieval.
I began utilizing data graphs to construct higher RAGs.
RAGs don’t essentially demand data graphs. They don’t demand a database in any respect. So long as you possibly can extract related info from a big pool and move it to the context of an LLM, RAGs work.
You could possibly construct a RAG with an internet search as its info retrieval technique or use a vector retailer to learn from its semantic textual content search options.
Should you use a graph database to retrieve contextual info, we name it GraphRAG.
This isn’t a put up about GraphRAGs (Maybe in a future put up). That is concerning the building of information graphs itself utilizing LLMs. However it’s value mentioning…