Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Improve quran.com retrieval from Vectara #83

Open
waleedkadous opened this issue Nov 17, 2024 · 0 comments
Open

Improve quran.com retrieval from Vectara #83

waleedkadous opened this issue Nov 17, 2024 · 0 comments
Assignees
@abdullah-alnahas

Comments

@waleedkadous
Copy link
Collaborator

Guidance from domain expert:

  • Retrieve 100 results
  • First, rerank them with Slingshot, our multilingual reranker.
  • Then, eliminate any results with a score less than threshold, t. You would use a user defined function (UDF) to do this, and take advantage of reranker chaining. If you're not sure what value to use for, t, 0.5 is probably the best choice.
  • Then resort the remaining results for diversity using the MMR re-ranker.
  • Feed the top 5 or 10 results into your LLM.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@abdullah-alnahas abdullah-alnahas

Labels
None yet
Projects
Ansari Work
Status: Ready for work
Milestone
No milestone
Development

No branches or pull requests
2 participants
@waleedkadous @abdullah-alnahas