That’s correct.
What is the best way to test autotagger in 5.3? Do I create a content graph and import it? If so, what should its structure be? A taxonomy, flat list…?
Here’s a possible setup for testing (assuming you already have a taxonomy that you want to tag against):
1. Create an ontology called “Document Ontology”.
- Add SKOS Core as an Include.
- Create a class “Document” and some attributes like “title” and “content”.
- Make “title” a subclass of “label” (from RDF Schema)
- Add a relation “topic” with range “Concept” (from SKOS Core)
2. Create an ontology called “Corpus” that includes the “Document Ontology”. Create some Document instances with titles and content, but without topics—the topics will be added in Tagger. (In a production deployment, this graph would be populated by some sort of import script that imports the documents from the content repository where they live.)
3. In the Server Administration under EDG Configuration Parameters, enable “Document Ontology” as a Tagger Property Graph.
4. Create a Content Tag Set with “Corpus” as content graph, “Document Ontology” as tag property graph, and your taxonomy as the concept vocabulary. Pick “topic” as the default tag property and “Document” as the root content class.
This setup should allow the use of Tagger for manual tagging to create some training data.
Continue with the AutoClassifier setup (Maui Server, etc.) as described in the documentation.
Hope that helps,
Richard