Typology
Use Typology to manage the categories the system sorts texts into. These categories make up your company's typology.
A new workspace starts with a set of starter categories that Blue Data Works loads before organizing is turned on. They appear here like any other company-wide category. Review and adapt them for your business before your texts are organized, as described in Getting started.
View categories
Open Typology to see the categories currently in use. Categories are company-wide and apply across your company. An existing project-specific category, if one is present, remains visible in the list.
Select Retired categories to see categories that are no longer in use. Retired categories are kept as history and cannot be edited from this list.
Add a category
Select Add category and enter its Name. New categories are company-wide.
You can also enter a Description. A description tells the model which texts belong in the category and which do not. Use clear, specific language so the model can make that decision well. The description is sent to the AI model with each text, so keep confidential details out of it. What Provenience can and cannot see explains what is sent.
Select Add category to save it. The new category applies from the next text received; texts already processed keep the categories they were given.
Change a description
Select Edit description beside a live category, update the Description, then select Save description. As when you add a category, the change applies from the next text received and does not re-sort texts already processed.
Retire a category
Select Retire beside a live category, then select Retire category to confirm. Texts already filed under it keep the category. New texts are no longer classified into it, beginning with the next text received.
There is no rename or un-retire option. To use a retired name again, add it as a new category.
Related
- When the model is unsure which category fits a record, it asks in the Review Queue. You can add a category there too.
- To fix a record the model categorized wrongly, use Artifacts.