add documentation for reranking #261
Merged
+118
−0
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: HenryL27 <hmlindeman@yahoo.com>
jonfritz
requested changes
Feb 28, 2024
| @@ -0,0 +1,119 @@ | |||
| # Reranking | |||
|
|
|||
| Second-stage reranking is a technique to use AI to gain significant search relevancy (we've seen +5-10% recall in the top 5) for your queries. OpenSearch 2.12 introduces the [Rerank processor](https://opensearch.org/docs/latest/search-plugins/search-pipelines/rerank-processor/), contributed by Aryn, to perform this operation. For more information, visit the [OpenSearch reranking documentation](https://opensearch.org/docs/latest/search-plugins/search-relevance/reranking-search-results/). | |||
There was a problem hiding this comment.
I don't think we should give percentages on improvements. Let's also remove "significant" - let's just say it can improve results, given there is a wide variance.
| @@ -0,0 +1,119 @@ | |||
| # Reranking | |||
|
|
|||
| Second-stage reranking is a technique to use AI to gain significant search relevancy (we've seen +5-10% recall in the top 5) for your queries. OpenSearch 2.12 introduces the [Rerank processor](https://opensearch.org/docs/latest/search-plugins/search-pipelines/rerank-processor/), contributed by Aryn, to perform this operation. For more information, visit the [OpenSearch reranking documentation](https://opensearch.org/docs/latest/search-plugins/search-relevance/reranking-search-results/). | |||
There was a problem hiding this comment.
Can we say "Sycamore uses OpenSearch's Rerank processor [link], contributed by Aryn...
|
|
||
| ## Usage | ||
|
|
||
| In order to use the rerank processor, you first need to register and deploy a [Text Similarity Model](https://opensearch.org/docs/latest/ml-commons-plugin/custom-local-models/#cross-encoder-models). If you're using the quickstart containers, one such model comes deployed already. You can get its id (let's call this `reranker_id`) with |
There was a problem hiding this comment.
They aren't called Quickstart containers. It's just Sycamore, so instead say "Sycamore includes NAME OF MODEL, and you will need to get it's unique ID to use for reranking. To get its ID...
| } | ||
| ``` | ||
|
|
||
| Now, create a pipeline with the rerank processor in it: |
There was a problem hiding this comment.
"create a new search pipeline (link to the docs where we talk about this)..."
| @@ -0,0 +1,119 @@ | |||
| # Reranking | |||
|
|
|||
| Second-stage reranking is a technique to use AI to gain significant search relevancy (we've seen +5-10% recall in the top 5) for your queries. OpenSearch 2.12 introduces the [Rerank processor](https://opensearch.org/docs/latest/search-plugins/search-pipelines/rerank-processor/), contributed by Aryn, to perform this operation. For more information, visit the [OpenSearch reranking documentation](https://opensearch.org/docs/latest/search-plugins/search-relevance/reranking-search-results/). | |||
There was a problem hiding this comment.
For more information, visit the Opensearch reranking documentation and Search Pipelines documentation NEED LINK.
| } | ||
| ``` | ||
|
|
||
| You can compose processors in pipelines. For example, to create a search pipeline that performs hybrid search, reranking, and RAG: |
There was a problem hiding this comment.
search pipelines. For example, to create a pipeline...
| } | ||
| ``` | ||
|
|
||
| The blocks of processors are ordered. We recommend reranking before doing RAG since it's important for RAG that the LLM gets the best results. |
There was a problem hiding this comment.
recommend adding reranking before the RAG processor...
| } | ||
| ``` | ||
| > Note: reranking can have sizeable latency - the processing required grows linearly with the number of search results. We recommend using this feature only in a high-resource environment. |
There was a problem hiding this comment.
Reranking can add latency to your query -
| > Note: reranking can have sizeable latency - the processing required grows linearly with the number of search results. We recommend using this feature only in a high-resource environment. | ||
| ## Theoretical motivation |
There was a problem hiding this comment.
Put this at the top, in the introduction.
Signed-off-by: HenryL27 <hmlindeman@yahoo.com>
jonfritz
approved these changes
Feb 28, 2024
|
|
||
| Second-stage reranking is a technique to use AI to improve search relevancy for your queries. Sycamore uses OpenSearch's [Rerank processor](https://opensearch.org/docs/latest/search-plugins/search-pipelines/rerank-processor/), contributed by Aryn, to perform this operation. For more information, visit the [OpenSearch reranking documentation](https://opensearch.org/docs/latest/search-plugins/search-relevance/reranking-search-results/) and [search pipelines documentation](https://opensearch.org/docs/latest/search-plugins/search-pipelines/index/). | ||
|
|
||
| ## Theoretical motivation |
Signed-off-by: HenryL27 <hmlindeman@yahoo.com>
bohou-aryn
pushed a commit
that referenced
this pull request
Mar 13, 2024
* add documentation for reranking Signed-off-by: HenryL27 <hmlindeman@yahoo.com> * incorporate PR edits Signed-off-by: HenryL27 <hmlindeman@yahoo.com> * remove theoretical motivation header Signed-off-by: HenryL27 <hmlindeman@yahoo.com> --------- Signed-off-by: HenryL27 <hmlindeman@yahoo.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.