Just made an algorithm adjustment to clew that hopefully should improve some results significantly: instead of multiplying everything into one score, I'm now using multiple sort critions; specifically, at the moment, I'm ranking by number of keywords matched, then by the score I was previously using.
This is gonna take more refinement but is the first step in the direction I needed to go next. It should also be the groundwork of being able to tell whether results are actually matching the query or only matching a small subset of it.
Add comment