Classify Data with Hierarchical Taxonomies

//todo: everything

Hierarchical Taxonomy Collections in Conscia

We've seen in previous sections the capabilities of a Taxonomy Collection to oversee a Collection and dynamically label its Records. We've also seen a Hierarchy that procedurally builds relationships between its own Records. However, there's no reason these two capabilities can't both apply at the same time - and create a Hierarchical Taxonomy.

The benefit this unlocks is that rather than assign a String as the tag to a Record, we can tag that record with a Hierarchy Path. This can enrich the targeted Collection in a robust way, enabling the targeted Collection to receive dynamic updates as the backing Hierarchy is manipulated.

Use cases for this capability include managing the categories of a product catalog, org charts and people management, customer segmentation, access management, species categorization, knowledge graphs, and many more.

Create a Hierarchy Collection...

Establishing a Hierarchy is detailed in the previous section.

...Then Make it a Taxonomy

Once we have a Hierarchical Collection we can right-click it and select Set as Taxonomy. We specify the field being applied (Hierarchy Path is a natural choice) as well as the Collection it targets and the column in that Collection where tags will be applied (which need not exist previously). Once complete, we can build a tree of related Records in the Hierarchical Taxonomy, then apply their addresses programmatically to a target Collection, giving nuanced data regarding the relationships at play in the data store managed by DX Graph.

This can enable implicit relationships within the target Collection, where two nodes that have the same Hierarchy Path are considered equivalently in terms of that attribute; and things like "cousin groups" can be explored through a walk of the DX Graph output ("Node X has an A > B > C Path and Node Y has an A > B > D Path. Therefore, we can highlight that they both B and they both A, but differ in a key way").

Making inferences using this data may be instinctual to users familiar with the data set, but it would be unwise to assume that intuition in Large Language Models (LLMs) and other AI services. By establishing a logical hierarchy between Records in the target Collection (that can be easily kept up-to-date with no code required), DX Graph can form the backbone of a Retrieval-Augmented Generation (RAG) system that lets an Agent (external like a chatbot, or internal like a copilot) be interviewed and answer questions about what can be a very complex or niche dataset with a high degree of competence.