Implementing semantic search with sparse vectors
In this recipe, we’ll introduce another vector search strategy – the sparse vector. In Elastic, it’s implemented through a proprietary model developed by search and machine learning experts at Elastic, called Elastic Learned Sparse EncodeR (ELSER). This model gives you the ability to perform semantic search right out of the box. We will learn how you can apply it to our dataset and what the benefits of leveraging such a strategy are versus dense vector search.
Getting ready
To follow along in this recipe, you’ll need to meet the following requirements:
- Make sure that you have completed the Implementing semantic search with dense vector recipe, as we will rely on the same dataset and Search Application
- As we will use the ELSER v2 sparse embedding model, which requires Elastic Stack version 8.11, make sure to have an Elastic deployment running in at least this version.
- Also, be mindful...
