Offer support for Learning To Rank Opensearch plugin
As an AI engineers, developer using OpenSearch for ML applications,
I want to incorporating machine learning in relevance ranking
So that I can better improve my application with proper ranking-> recommendation
Viktor Trykolenko
shared this idea
-
Hi, thanks for your comment
Neural search has been on our radar for a while and also the search processor project.
We hit a few problems OS 2.9 so we are targeting OS 2.10. The upstream OpenSearch 2.10 will be out later Sept and we will adopt it as soon as possible in Aiven for OpenSearch.
-
semantic word
commented
v2.9.0 of the OpenSearch has marked Neural Search as production ready and out of box.
Previously Lucene had a limit of 1,024 dimension vectors but OpenAI provided embedding with 1,536 dimensions. With OpenSearch 2.9, we can now use embedding created using OpenAI as FAISS support now has 2,048 dimension vector support.
It would be nice for Aiven to update OpenSearch to v2.9 as I am sure the new AI-powered semantic search and ML model support in this release is a big news.
-
The OpenSearch community has moved a fork of an equivalent Elasticsearch plugin under official opensearch project. We can observe the development of this plugin here:
https://github.com/opensearch-project/opensearch-learning-to-rank-base
