Expanding the evalutation example to include the recall argument about pyvespa

[kkraune]

Hi @sephib thanks for submitting. Thiago will get back on this, it will take some days as he is not working on this right now

[thigm85]

Yes, it makes sense when you want to rank only specific document ids. I will add an example as suggested.

[thigm85]

Hi @sephib, I am not sure if using recall in this use-case makes sense. When you call evaluate_query you want to evaluate the results returned by a specific query_model for a given query.

Using recall here will contaminate the evaluation results by potentially including documents that wouldn't otherwise be returned by the query model. Am I missing something?

[sephib]

you are also correct -
but on the other hand - sometimes you want to confine the evaluation to a set of documents - especially when working with NDCG metric.