We released the ability to support metadata taxonomies when extracting structured data from unstructured content and saving that data as metadata alongside content in Box with Box Extract. This capability allows teams to fully automate and operationalize their most complex, hierarchical metadata structures.
Taxonomies allow Box administrators to model reusable, hierarchical schemas (such as Country → State/Province → City) to represent multi-level choices within metadata templates. This release ensures that these sophisticated structures can now be seamlessly processed and populated with extracted data using Box Extract. Teams can now enable and extract taxonomy-backed fields already attached to your metadata templates, turning previously static, disabled fields into active, automated data streams. Instead of configuring extraction instructions for every individual level of a hierarchy, you can now provide one single set of AI instructions for the entire taxonomy field and Box Extract fully respects your pre-defined taxonomy behaviors, including:
- Single-select vs. Multi-select modes.
- Leaf-only vs. All-level selection rules, like restricting selection to a specific city or allowing selection of a state/country.
By default, Box Extract returns the selected taxonomy value alongside its full hierarchical structure, ensuring downstream users and systems have the complete context they need without complex manual mapping. Additionally, Box Extract stays perfectly in sync with your metadata template settings within the admin console. If a taxonomy field is hidden or deleted in the template itself, it is automatically removed from Box Extract. If a previously enabled field is modified or deleted, Box Extract will flag an error on the configuration page so you can quickly update your workflows.
To learn more, please see Metadata Taxonomies.