Uploaded image for project: 'Jackrabbit Oak'
  1. Jackrabbit Oak
  2. OAK-8119

Let similarity search rerank use distance as exact score

    XMLWordPrintableJSON

    Details

    • Type: Task
    • Status: Closed
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 1.12.0, 1.10.2
    • Component/s: lucene
    • Labels:
      None

      Description

      Currently the reranking algorithm for feature vector similarity search uses an additive scoring mechanism so that the final score is the sum of the original LSH query and the query - document feature vector distance. It turns out that since LSH is supposed to approximate the fv distance it'd be more effective to just use the exact distance as the final score instead of performing the mentioned sum.

        Attachments

          Activity

            People

            • Assignee:
              teofili Tommaso Teofili
              Reporter:
              teofili Tommaso Teofili
            • Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: