Currently feature collection in mjolnir runs a multi-search for each feature individually. Now that the ltr plugin is (or soon will be) available in prod and relforge it would be good to switch this over to collecting features from the ltr query as well.
|Open||None||T174064 [FY 2017-18 Objective] Implement advanced search methodologies|
|Resolved||EBernhardson||T161632 [Epic] Improve search by researching and deploying machine learning to re-rank search results|
|Resolved||EBernhardson||T162053 backend data engineering and plumbing for LTRank|
|Resolved||dcausse||T168813 Collect features from the sltr query|
I started to work on feature logging from the plugin, I think I'll continue on the plugin and use it in mjolnir.
Problem right now is that I can only obtain the explain output from a model, it means that a dummy model must be created before the fact. Logging inside the plugin should work with feature set and exisiting models.
So because both options (logging in the plugin vs explain) require some work in the plugin:
- implement logging
- extend sltr query to accept a featureset
I prefer to go with the long term approach and implement proper feature logging (unless there's some urgency in doing otherwise)?