The weights for similar story finding are off. First,
weights of matches found should be affected by
multipliers not just for the location the hit is found
in (target story's title/introtext/bodytext) but for
the location the word came from (edited story's
title/introtext/bodytext, take the top one if in more
Second, the length of the matched story should be
applied as an overall modifier. The longer the matched
story, the more likely it is to match any given story,
so the lower its weight should be multiplied by.
Log in to post a comment.