I am trying to create an information retrieval system which can benefit from user feedback (either implicit, through e.g., click-through data) or explicit (e.g., binary feedback on irrelevant suggested results).

More specifically, my plan is to use an ElasticSearch index to reduce the space by providing a 'bucket' of candidate relevant results and then re-rank the results in the bucket by a more sophisticated ML model that can benefit from feedback (both implicit and explicit, as described above).

My sense is that for the ML component I should use a Learning to Rank method, but other colleagues have suggested that I should take a look in Reinforcement Learning instead.

I want to know what is your viewpoint on the matter, i.e., which of the above two methods (Learning to Rank vs Reinforcement Learning) seems more relevant for the task?

  • $\begingroup$ Feels like a "It depends" kind of problem. So, if you can, please, share some information that may help to make such a decision. $\endgroup$
    – oW_
    Oct 12 '20 at 19:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.