MVP: Inspire Recommendation System (IRS)

[ParthS007]

1. 50k es documents
2. Do we need to clean them, if yes keep the required metadata only else skip
3. Create embeddings
4. Metric - Performance, Compute usage if any, etc...
5. have them on jupyter notebook for results checking by others.
6. decide the final model and check the results again.
7. experimental feature release and clean up

[ParthS007]

• Models: https://huggingface.co/sentence-transformers

[ParthS007]

All progress here now: https://github.com/cern-sis/inspire-recommendation-system