(EAI-375): Ingest snooty docs facets and meta #558
Closed
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
nlarew
reviewed
Nov 18, 2024
| const facetAndMetaNodes = findAll( | ||
| node, | ||
| ({ name }) => name === "facet" || name === "meta" | ||
| ) as (SnootyFacetNode | SnootyMetaNode)[]; |
There was a problem hiding this comment.
This works for now. Longer term if we plan to maintain this Snooty AST glue code then it might be nice to move to Zod parsing for these instead of type casts.
nlarew
approved these changes
Nov 18, 2024
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Outdated suggestions cannot be applied.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.
Jira: https://jira.mongodb.org/browse/EAI-375
Changes
metaandfacetsfields asPage.metadataNotes
facets.tomlfiles. There's a DOP ticket to capture thefacets.tomlhttps://jira.mongodb.org/browse/DOP-5182Experiment Results
Initial experiment
Experiment compares using the ingestion pipeline with the new snooty metadata to the previous baseline. The results can be found here: mongodb-chatbot-retrieval/experiments/mongodb-chatbot-retrieval-snooty-metadata
The results actually show a very slight decrease in search quality as a result of these changes:
Follow up experiment
After upgrading embedding model and preprocessor. Even worse results. Can be seen here in Braintrust.
BinaryNDCG@5 goes from 51.77% -> 46.93%
Next Steps
Based on the results, I think we should not ingest and chunk this metadata. Instead, we should ingest it so it's present in the
pagescollection, but doesn't get included in theembedded_content. Will create follow-up PR for that.