search

Search notebooks

This folder contains a number of notebooks that demonstrate the fundamentals of Elasticsearch, like indexing embeddings, running lexical, semantic and hybrid searches, and more.

The following notebooks are available:

0. Quick start
1. Keyword, querying, filtering
2. Hybrid search
3. Semantic search with ELSER
4. Multilingual semantic search
5. Query rules
6. Synonyms API quick start
7. Inferencee
8. Learning to rank
9. Semantic text
10. Semantic reranking with Cohere reranker

Notebooks

0. Quick start

In the 00-quick-start.ipynb notebook you'll learn how to:

• Use the Elasticsearch Python client for various operations.
• Create and define an index for a sample dataset with dense_vector fields.
• Transform book titles into embeddings using Sentence Transformers and index them into Elasticsearch.
• Perform k-nearest neighbors (knn) semantic searches.
• Integrate traditional text-based search with semantic search, for a hybrid search system.
• Use reciprocal rank fusion (RRF) to intelligently combine search results from different retrieval systems.

1. Keyword, querying, filtering

In the 01-keyword-querying-filtering.ipynb notebook, you'll learn how to:

• Use query and filter contexts to search and filter documents in Elasticsearch.
• Execute full-text searches with match and multi-match queries.
• Query and filter documents based on text, number, date, or boolean values.
• Run multi-field searches using the multi-match query.
• Prioritize specific fields in the multi-match query for tailored results.

2. Hybrid search

In the 02-hybrid-search.ipynb notebook, you'll learn how to:

• Combine results of traditional text-based search with semantic search, for a hybrid search system.
• Transform fields in the sample dataset into embeddings using the Sentence Transformer model and index them into Elasticsearch.
• Use the RRF API to combine the results of a match query and a kNN semantic search.
• Walk through a super simple toy example that demonstrates, step by step, how RRF ranking works.

3. Semantic search with ELSER

In the 03-ELSER.ipynb notebook, you'll learn how to:

• Use the Elastic Learned Sparse Encoder (ELSER) for text expansion-powered semantic search, out of the box — without training, fine-tuning, or embeddings generation.
• Download and deploy the ELSER model in your Elastic environment.
• Create an Elasticsearch index named search-movies with specific mappings and index a dataset of movie descriptions.
• Create an ingest pipeline containing an inference processor for ELSER model execution.
• Reindex the data from search-movies into another index, elser-movies, using the ELSER pipeline for text expansion.
• Observe the results of running the documents through the model by inspecting the additional terms it adds to documents, which enhance searchability.
• Perform simple keyword searches on the elser-movies index to assess the impact of ELSER's text expansion.
• Execute ELSER-powered semantic searches using the text_expansion query.

4. Multilingual semantic search

In the 04-multilingual.ipynb notebook, you'll learn how to:

• Use a multilingual embedding model for semantic search across languages.
• Transform fields in the sample dataset into embeddings using the Sentence Transformer model and index them into Elasticsearch.
• Use filtering with a kNN semantic search.
• Walk through a super simple toy example that demonstrates, step by step, how multilingual search works across languages, and within non-English languages.

5. Query rules

In the 05-query-rules.ipynb notebook, you'll learn how to:

• Use the query rules management APIs to create and edit promotional rules based on contextual queries
• Apply these query rules by using the rule_query in Query DSL

6. Synonyms API quick start

In the 06-synonyms-api.ipynb notebook, you'll learn how to:

• Use the synonyms management API to create a synonyms set to enhance your search recall
• Configure an index to use search-time synonyms
• Update synonyms in real time
• Run queries that are enhanced by synonyms

7. Inference

8. Learning to rank

9. Semantic text

In the 09-semantic-text.ipynb notebook, you'll learn how to:

• Quickly get started with semantic search with the semantic_text field type
• Use the semantic query
• Implement hybrid search

10. Semantic reranking with Cohere reranker

In the 10-semantic-reranking-retriever-cohere.ipynb notebook, you'll learn how to:

• Use the inference API to create an inference endpoint for reranking
• Use a retriever query to search and rerank the results