Search - Course Run with Expired Upgrade Deadline is Higher in Relevancy Score Than Other Course

Description

Current Behavior
When I do a search for the word "Chemistry", there are some archived courses (with expired upgrade deadlines) with higher relevance scores than non-archived courses (with unexpired upgrade deadlines).

Expected Behavior
Based on my understanding of how we calculate search relevancy, the boosting that the course run would get from non-expired upgrade deadline () should result in courses with a non-expired upgrade deadline above those courses with an expired upgrade deadline.

Repro:

  1. Search for the word chemistry (click this link)

  2. The first two courses are higher in relevancy than the third one.

    1. Preparing for CLEP Chemistry: Part 1 (expired)

      1. https://www.edx.org/course/preparing-clep-chemistry-part-1-cooper-union-chem1-x

      2. https://ecommerce.edx.org/courses/course-v1:CooperUnion+Chem.1x+2T2016/

    2. Preparing for CLEP Chemistry: Part 2 (expired)

      1. https://www.edx.org/course/preparing-clep-chemistry-part-2-cooper-union-chem-2x

      2. https://ecommerce.edx.org/courses/course-v1:CooperUnion+Chem.2x+2T2016/

    3. Advanced Biorefinery (non-expired)

      1. https://www.edx.org/course/advanced-biorefinery-wageningenx-bb03x

      2. https://ecommerce.edx.org/courses/course-v1:WageningenX+BB03x+2T2017

  3. Screenshot link

Notes:

  1. Verify that existing "purchasable" boosting is correct. Consider increasing the weight for this boosting.

  2. Investigate global relevancy score (DFS Query Then Fetch)

  3. How should purchasability score compare to tf-idf score?

    1. Courses with similar term relevancy (tf-idf) should favor the available course, but an unavailable course with significantly higher term relevancy should still rank higher

    2. Could apply subject boosting (beyond tf-idf score already generated by matching on subject names that make it into search documents)

Epic Link

Activity

Show:
Patrick Cavallario
June 22, 2017, 7:43 PM

Thanks. Much appreciated.

Renzo Lucioni
June 22, 2017, 7:27 PM

I would just let folks know, though it couldn't hurt to take it for a spin and make sure it's what you expect. I wouldn't have closed this story without making sure the AC were met.

Patrick Cavallario
June 22, 2017, 5:40 PM

and I'd like to communicate this to Sasha and our partners as this is often a topic of conversation. I haven't tested these changes, but can. Do you think that is necessary? Or do you think we can just let folks know?

Renzo Lucioni
June 15, 2017, 5:52 PM

Search boosting improvements for this ticket are in production.

Lauren Holliday
June 14, 2017, 4:09 PM

fyi we are working on this to try to demote archived courses in search results

Fixed
Your pinned fields
Click on the next to a field label to start pinning.

Priority

CAT-3

Assignee

Renzo Lucioni

Sprint

None

Story Points

3

Reporter

Jasper Ho