Quepid is our free tool for relevancy tuning – available as a free hosted service and as open source code. In certain relevancy cases, you may only care about the ordering of a subset of documents within the entire result set. One particular use case may be in e-commerce where you want certain products to be boosted and others to be demoted in the ranking algorithm. While this was doable in previous versions of Quepid, it could get difficult to manage the list of rated documents, especially if the list went many pages deep into the results.
To help solve this problem we worked with Kobler to introduce several new features to Quepid. We strived to make it easier to work with your corpus of rated documents in the main case view, inside of Explain Missing Documents and within the scorers themselves. Here’s a short video showing the new feature in action:
As a Holiday present to the Quepid community you can look forward to the following changes:
- Migrated scoring to an asynchronous pipeline
- Added “Show Only Rated” toggle to the main
searchResult
listing - Support for showing only rated documents within snapshots
- Explain Missing Documents modal now defaults to showing all of your rated documents when you first open it up
- Added
eachRatedDoc
function to ScorerFactory to iterate over rated documents in scoring - Provided
refreshRatedDocs(k)
in ScorerFactory for loading a custom amount of rated documents up to k - Fixed stuck “Updating queries” message when creating a new case
We are happy to share these new features and fixes with the community and we offer our sincere thanks to Erik Bugge and the folks at Kobler for funding the Only Rated feature.
If you’d like to chat about using Quepid do join Relevance Slack and the #quepid channel, or feel free to contact us for help with your relevancy tuning challenges.