App search - Semantic search and machine learning NLP models

Hi there :wave:

The following webinar "Introduction to NLP models and vector search" gives a very nice demo on uploading and using NLP models to improve search relevance. However, it might seem it is only focused on core Elasticsearch:

I am wondering can that be integrated with Elastic Enterprise App Search ? Concretely, I would like to know how to integrate NLP models for semantic search into engines within App Search ?

Secondly, and putting things into context, I currently have an engine for an entity Courses which I populate through the API, I later use the Search UI provided by App Search which behind the scenes uses the API /api/as/v1/engines/courses/search.json: results would then reflect the relevance configuration done in the engine. I would like to know then: what is the approach to mix this regular engine search and the NLP _infer + _knn_search search showcased in the demo if we are building a UI that should benefit from both to improve relevance ? How do they play along together ?

Thanks in advance.

Hi Gerardo!

I think the short answer is that they are two entirely different relevancy models and they do not work together.

To do that you might want look into either using Elasticsearch directly OR using elasticsearch index engines in App Search.

I don't know what would be the best approach in your situation. Elasticsearch is outside of scope of this subforum, and Elasticsearch index engines is a technical preview feature that is not covered by support currently.

Maybe just App Search without NLP will be enough for your use-case? It has great relevance controls, so you could fine-tune the relevance to suit your needs.