Open michelole opened 5 years ago
@khituras other.json
(used to downrank matches on comorbidities in CT exclusion criteria) also has a negative boost. However, it doesn't happen together with negative_boosters.json
(used only for BA).
Worst case scenario we could try a must_not
clause. Update: they don't seem to improve results either.
match_all
could be switched to a match on any "cancer" or "gene"/"mutation".
This is being solved by reverting back to an ES 5.4 index (see also #84).
Used for
negative_boosters.json
,match_all.json
, andother.json
.negative_boosters.json
andother.json
should be migrated to query boosting. Caveat: static boosting might lead to different results (better? worse? same?)match_all.json
: boost = 0other.json
could also be migrated to ashould_not
query.Test with 2017 and 2018 data. Validate on internal 2019 GS.