How could we compute Document popularity, Query intent match and Personalization match

The techniques we could use for the computation of Terms match are described. However, for the other terms of the relevance score, I do not see infos on how we could compute them.

For the document popularity we can use the number of times the document is linked to (like google).

But for the others two, I think it would involve more machine learning which is out of scope here since we ares scoring a large number of documents.

Any thoughts ?