How ?ql_score_ is computed in the latest text index implementation?

[aindlq]

I'm a bit struggling to understand the way score is calculated in the latest text index re-implementation.

From quick look, it seems that scores for text words are simply ignored and only scores for entities are taken into account. So what we get is sum of scores for the matched entity per document. So if entity occurs multiple times in the document a match will get higher score.

Is that correct?

[joka921]

Yes, Currently QLever stores scores for a word occurence in a text, but there is no way to retrieve these scores to use them for sorting and / or filtering. But as this feature (the text index and its scores) has gained more and more interest in the last weeks an months, we now have someone working on those scores, see for example #1397 .

Can you describe a concrete use case on how you would like to use which kinds of scores, this would help us design these features.