PDE5 should give PDE5A first?

[cbizon]

With an input of PDE5 (autocomplete on or off), i might expect to get PDE5A. PDE5 is an exact synonym of PDE5A (it was the name for a long time).

But the top few hits are things like this:

{
    "curie": "PUBCHEM.COMPOUND:16066664",
    "label": "Pde5-IN-42",
    "synonyms": [
      "Pde5-IN-42",
      "CHEMBL1094365"
    ],

If you tokenize on - then there's an exact match to a token, but that seems not as good as the exact match to a synonym?

[gaurav]

Hmm, good catch. I'll try to increase the boost for the synonym exact match and see if I can fix this.

[gaurav]

As of PR #183, our results have improved: lots of pde5 and a few pde5a's -- with taxon filtering this should be pretty much done, but I'll mark it as needing testing for now.

[gaurav]

It looks like we can't fix this with only boosts, because we only have a few "pde5" preferred label matches, and at the moment we promote those far higher than the many matches with either "pde5" as a synonym or "pde5*" as a preferred label. We can use boosts to boost the human gene so that it appears above equivalently scored entries, and you can use filters to request only human genes. I'll see how much I can improve this in #211 before closing this.