Taxonomies in the Age of AI: Are They Still Relevant?

 Rebecca Schneider, Executive Director of AvenueCX claimed that taxonomies are still relevant. And even moreso today. 

Taxonomies: A system for organizing content according to shared characteristics, or a way of describing things. A taxonomy describes and organizes stuff. 

Organizations have large repositories that can be used to create LLMs. AI can leverage content to understand relationships between terms. 

Unstructured content can create confusing output. Content is not always high quality. And even if metadata is applied, it can vary in quality and availability. 

Well, when no money for taxonomy development, can still use AI. Your mileage may vary--can require training and retraining (and retraining), which can cost more money. 

AI is well positioned to look at relationships between terms across multiple types of content and data. In essence, AI creates a taxonomy as it processes content. But not entirely satisfied with results. 

A study found that LLMs were good with widely known domains, such as shopping, but do not perform well in specialized taxonomies, such as computer science. 

What goes into creating a taxonomy? Start with a traditional approach: business goals and objectives, what are current pain points, what content needs description, who is the audience and what is their content needs.

Use AI in the best of both worlds. For example, use AI to identify patterns and relationships. Can be used as a jumping off point, but needs review. Feed results back into the system for further training. 

Ontologies are a formal naming and definition of types, properties, and interrelationships of entities in a particular domain. Take taxonomies and their associated facets and describing relationships, you can enrich your data, which can further inform your AI engine.

An ontology can be used for a retrieval-augmented generation (RAG) solution. (the RAG space is a fast-moving target.) 

What is success? A well-tested, vetted taxonomy/ontology that is implemented. Structured content that is business ready. Clean data. Active governance for data and content. Contributions from SMEs because they are the ones who can judge the accuracy of responses.