References from
I would like to ask about the third source's statement “From the recalled document fragments, identify the entities generated based on the fragments and sort the top 25 based on the number of text fragments associated with the entities.”
Why 25?
And what is the reason for the 2 hops as stated in all three sources?