@Starcross@mk.absturztau.be I haven't used Quibblo, but surely the snippet from the source code would be machine readable, since whatever source the server sent with the HTML is probably machine generated. Since you have category_id given, I think the best way to approach this problem would be to parse the snippet to categorize by category_id, which you can do with a simple script. Machine learning doesn't seem to be the right tool for this to me because your task doesn't require the interpretation of language; you're matching a set of inputs by the category_id field they have in common, so it could be done autonomously by just comparing category_ids.

If you were trying to categorize the quiz answers by what categories they would likely fit into, like if you only had the quiz answers but not the category_ids, then that might be the right situation for a language model, in my opinion. Even then, I think it should be a purely categorizing AI model; even in this situation, the task itself doesn't require processing a natural language prompt or generating a response, it only needs to process the language of the quiz answers.