Taxonomy#
The taxonomy is the set of topics and document kinds the corpus is classified against. It drives the topic rows on Explore, the filters in Search and the Library, and the knowledge graph. Review it, adjust the labels, and have the classification agents apply them across the corpus so the structure users navigate reflects the real content.
Each label can carry a definition - a sentence or two saying what the label means and when it applies. Definitions show on the Taxonomy page as the vocabulary reference and are what the labelling agents classify against. Edit a label set under Manage > Taxonomy: saving it restarts every labeller that carries the set so it picks up the new labels and definitions, and the restarted labeller applies to new resources only - nothing already in the corpus is reprocessed or relabelled.
Create a set under Manage > Taxonomy (or from the Taxonomy page): give it a name - its id is derived from the name, so "Marine Region" becomes marine-region - choose whether a resource may carry one value or several, and add its labels with their definitions. Nothing carries a brand-new set, so creating one neither creates nor restarts any agent; a labeller for it comes from running analysis or the knowledge graph tools. Editing a set later restarts only the labellers that carry it.
Enrichments#
Enrichments are the structured fields generated onto each document - a real title, a summary, key takeaways and quotes of interest - designed to replace raw filenames and give every document a scannable, credible presentation on cards, in the Library and on the document page. Until the enrichment has been run over a corpus, documents fall back to their project code and file name.
The default research enrichment ships as the first enrichment. Each enrichment is a generation agent plus a schema; the portal renders whatever fields the schema defines, so adding a new lens on the corpus is a configuration change, displayed automatically.
The knowledge graph strategy#
The knowledge graph is built by an extraction agent configured with the entity types and relation examples that matter for the domain. Review and refine that strategy in management, and the graph the portal draws follows from it.