Domain classifier

[marcverhagen]

We now assume that domains are associated with data drops.

But not all documents in those drops are reliably in the same domain and we want to be able to classify new documents when they are not associated with a domain.

The code idea is to

• start from a list of publications
• extract domains via ChatGPT prompts
• find heuristics or do some minimal annotation