Avalara: Their engineering team (led by Michael Iantosca — follow him on Linkedin and Medium and visit his blog https://thinkingdocumentation.com) has pioneered using Context Graphs to turn documentation into queryable knowledge. The ontology defines the rules of the content, while the RAG system uses that metadata to ensure the documentation portal provides "governed" answers rather than hallucinations.
Hi Scott, great article on a very relevant topic. As a terminology geek, I would add that terminology fulfils many of these tasks already, and can at the same time be a great starting point for building taxonomies or ontologies. I just published a blog article on our "TAG" approach - Terminology-Augmented Generation, which goes very much in line with what you write here: https://kaleidoscope.at/en/blog/better-ai-results-through-tag-and-mcp/ . Let me know what you think!
Scott, do you have any examples of OG-RAG systems that make use of metadata or that consult an ontology?
Avalara: Their engineering team (led by Michael Iantosca — follow him on Linkedin and Medium and visit his blog https://thinkingdocumentation.com) has pioneered using Context Graphs to turn documentation into queryable knowledge. The ontology defines the rules of the content, while the RAG system uses that metadata to ensure the documentation portal provides "governed" answers rather than hallucinations.
Hi Scott, great article on a very relevant topic. As a terminology geek, I would add that terminology fulfils many of these tasks already, and can at the same time be a great starting point for building taxonomies or ontologies. I just published a blog article on our "TAG" approach - Terminology-Augmented Generation, which goes very much in line with what you write here: https://kaleidoscope.at/en/blog/better-ai-results-through-tag-and-mcp/ . Let me know what you think!