Elastic AI: ELSER

Elastic IA: ELSER

ELSER is a model created by Elastic that integrates artificial intelligence capacity into searches. Previously, searches were carried out by keywords. With ELSER, searches have evolved through the use of vectors.

ELSER performs searches semantically, it is capable of using context, thanks to this approach it improves precision and personalization, increasing the user experience.

This transition from keyword-based searches to a semantic approach with AI is a big change, ELSER represents a fundamental pillar in this evolution.

ELSER uses sparse vectors, platinum/enterprise licensing is required, in addition to having 4GB on the ML node, a specific node with the ML (Machine learning) role is recommended to download and deploy ELSER.

Currently there is the ELSER V2 version offering improved precision and more efficient indexing, this is thanks to an improvement in the quality and increase in training data, there are also 2 optimized versions depending on the hardware used.

If you have used the previous ELSER model it is important to note that it is not compatible with V2, so you will need to re-index the data using a new ingest processing referencing ELSER V2.

Thanks to its integration, downloading and deploying ELSER is really easy, you just need to enter the “machine learning” menu from Kibana and select the appropriate version.