Using controlled vocabularies to train custom models.

Hi. I am am wondering how I can make the most out of controlled vocabularies offered in the formats mentioned below?

Example about an AAT subject: aventurine (quartz)

• Human-readable page about the subject at Getty's site:
  http://vocab.getty.edu/page/aat/300011154
• Human-readable page showing the hierarchical position of the subject:
  http://vocab.getty.edu/hier/aat/300011154
• Semantic URI
  http://vocab.getty.edu/aat/300011154
• Forest HTML page showing RDF triples
  http://vocab.getty.edu/aat/300011154.html
• RDF/XML
  http://vocab.getty.edu/aat/300011154.rdf
• Turtle
  http://vocab.getty.edu/aat/300011154.ttl
• NTriples
  http://vocab.getty.edu/aat/300011154.nt
• JSON
  http://vocab.getty.edu/aat/300011154.json
• JSONLD
  http://vocab.getty.edu/aat/300011154.jsonld