Hi, I notice that elasticsearch by default provides disjunction semantics for query terms, and therefore use dynamic pruning techniques to accelerate the query execution.
However, if the query is very long, say contains 100 tokens, it seems elasticsearch could also return the results without remarkable performance differentiation, but the bm25 score seems to change a lot. Does elasticsearch prune the query terms before the query is executed , say the practical terms are around 20 actually ?