Conversation
- Add new guide for choosing between retrieval approaches (BM25, semantic, hybrid) - Covers when to use each strategy and how they compare - Add navigation entry in toc.yml Provides guidance for users deciding which retrieval strategy to use for their Elasticsearch search implementation.
- Add comprehensive guide comparing full-text, semantic, hybrid search, and reranking - Helps users choose the right retrieval approach for their use case - Includes decision matrix and next steps guidance - Add navigation entry in toc.yml
Compare full-text, semantic, hybrid search, and reranking approaches Include decision matrix and resource considerations Add navigation entry already present in toc.yml This PR adds documentation to help users choose between different retrieval strategies in Elasticsearch. The guide covers: - Full-text search (BM25): for exact-match and keyword queries - Semantic search: for conversational and natural language queries - Hybrid search: combining BM25 and semantic retrieval with RRF - Reranking: second-pass scoring for improved precision Changes: - New file: solutions/search/choose-retrieval-strategy.md - Navigation: Entry already exists in solutions/toc.yml (line 16) Generative AI disclosure: Tool(s) and model(s) used: - Claude Sonnet 4.6 (Anthropic) - ChatGPT Deep Research (OpenAI, model 5.4)
✅ Vale Linting ResultsNo issues found on modified lines! The Vale linter checks documentation changes against the Elastic Docs style guide. To use Vale locally or report issues, refer to Elastic style guide for Vale. |
There was a problem hiding this comment.
Pull request overview
Adds a new Solutions Search documentation page to help users choose between retrieval strategies in Elasticsearch (full-text/BM25, semantic, hybrid, reranking), and wires it into the Search approaches navigation.
Changes:
- Adds a new guide:
choose-retrieval-strategy.mdwith explanations, a comparison table, and “next steps” links. - Updates
solutions/toc.ymlto include the new page under Search approaches. - (Unintended) Alters the Full-text TOC structure around “search relevance” pages due to a removed TOC entry and mis-indented children.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
solutions/toc.yml |
Adds the new page to navigation, but currently breaks nesting for the existing full-text “search relevance” section. |
solutions/search/choose-retrieval-strategy.md |
New retrieval strategy selection guide; needs table syntax correction and front matter alignment with section conventions. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| - file: search/full-text.md | ||
| children: | ||
| - file: search/full-text/how-full-text-works.md | ||
| - file: search/full-text/text-analysis-during-search.md |
There was a problem hiding this comment.
The TOC nesting looks broken: search/full-text/text-analysis-during-search.md is now given children: that point to search/full-text/search-relevance/*, and the search/full-text/search-relevance.md entry was removed. This will misplace the “Search relevance optimizations” pages in navigation (and likely make search-relevance.md unreachable). Restore the - file: search/full-text/search-relevance.md item and move these children back under it, leaving text-analysis-during-search.md without those children.
| - file: search/full-text/text-analysis-during-search.md | |
| - file: search/full-text/text-analysis-during-search.md | |
| - file: search/full-text/search-relevance.md |
| | Situation | Recommended approach | | ||
| |---|---| | ||
| | Queries use specific terms, identifiers, or codes | Full-text | | ||
| | Queries are conversational or use natural language | Semantic | | ||
| | Mixed query types in the same application | Hybrid | | ||
| | You need both exact-match precision and semantic recall | Hybrid | | ||
| | You want higher precision over a retrieved candidate set | Add reranking | | ||
| | Multilingual corpus, no translation layer | Semantic | | ||
|
|
There was a problem hiding this comment.
The comparison table uses || at the start of each row (|| Situation | Recommended approach |), which doesn’t match the table syntax used elsewhere in these docs and will render incorrectly (extra empty column / malformed table). Use standard Markdown table pipes (|) for the header and each row.
| --- | ||
| navigation_title: Choose a strategy | ||
| description: Compare full-text search, semantic search, hybrid search, and reranking to decide which approach fits your use case. | ||
| applies_to: | ||
| stack: ga | ||
| serverless: | ||
| elasticsearch: ga | ||
| --- |
There was a problem hiding this comment.
This page’s front matter defines applies_to but omits the products: list that is present in other solutions/search/* pages (for example solutions/search/using-openai-compatible-models.md:1-13 and solutions/search/search-approaches.md:1-12). If the docs build relies on products for filtering/badging, add an appropriate products list here for consistency.
- Restore search-relevance.md as parent entry with correct nesting - Add products list to front matter for consistency - Fixes navigation error: 'Could not find Search relevance optimizations'
🔍 Preview links for changed docs |
alaudazzi
left a comment
alaudazzi
left a comment
There was a problem hiding this comment.
A couple of editing suggestions, otherwise LGTM.
|
|
||
| # Choose a retrieval strategy | ||
|
|
||
| Elasticsearch supports several retrieval strategies. This page helps you understand the differences and choose the right approach for your use case. |
There was a problem hiding this comment.
| Elasticsearch supports several retrieval strategies. This page helps you understand the differences and choose the right approach for your use case. | |
| {{es}} supports several retrieval strategies. This page helps you understand the differences and choose the right approach for your use case. |
| - **Full-text search**: keyword and lexical matching using BM25 | ||
| - **Semantic search**: meaning-based matching using vector embeddings | ||
| - **Hybrid search**: BM25 and semantic search combined | ||
| - **Reranking**: a second-pass scoring step applied on top of retrieval results |
There was a problem hiding this comment.
| - **Full-text search**: keyword and lexical matching using BM25 | |
| - **Semantic search**: meaning-based matching using vector embeddings | |
| - **Hybrid search**: BM25 and semantic search combined | |
| - **Reranking**: a second-pass scoring step applied on top of retrieval results | |
| - **Full-text search**: Keyword and lexical matching using BM25 | |
| - **Semantic search**: Meaning-based matching using vector embeddings | |
| - **Hybrid search**: BM25 and semantic search combined | |
| - **Reranking**: A second-pass scoring step applied on top of retrieval results |
Summary
Add comprehensive guide for choosing a retrieval strategy in Elasticsearch.
This PR adds documentation to help users choose between different retrieval strategies. The guide covers:
Changes
solutions/search/choose-retrieval-strategy.mdsolutions/toc.yml(line 16)Generative AI disclosure
Did you use a generative AI (GenAI) tool to assist in creating this contribution?
✅ Yes
Tool(s) and model(s) used: