Skip to content

Risk repository: Switch to tree hierarchy for classification - #59

Merged
rgambee merged 8 commits into
mainfrom
rr/classify-tree-hierarchy
Aug 9, 2026
Merged

rgambee merged 8 commits into
mainfrom
rr/classify-tree-hierarchy

Conversation

@rgambee

@rgambee rgambee commented Aug 9, 2026

Copy link
Copy Markdown
Member

This switches the classification stage from using the older (sub)category hierarchy to a more flexible tree hierarchy, which allows arbitrary levels of nesting. The evaluation helper is already using it.

There's a conversion stage needed for the existing extraction stage (which is currently unused), as well as for the risks from the first iteration of the repository, which we use as ground truth for evaluating the classifier.

There's also a new script for running the classifier by itself against the first iteration's risks for easier tuning.

rgambee added 8 commits August 8, 2026 17:04
Concurrency now happens at the risk level rather than the document
level. This makes it simpler to process pre-extracted risks from
Airtable.
Instead of classifying risks extracted by the pipeline, it operates on
risks stored in Airtable from the previous iteration of the risk
repository.
@vercel

vercel Bot commented Aug 9, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
airi-pipelines Ready Ready Preview Aug 9, 2026 6:28pm

Request Review

@rgambee
rgambee merged commit f93a7e2 into main Aug 9, 2026
3 checks passed
@rgambee
rgambee deleted the rr/classify-tree-hierarchy branch August 9, 2026 18:29
rgambee added a commit that referenced this pull request Sep 23, 2026
This switches the classification stage from using the older
(sub)category hierarchy to a more flexible tree hierarchy, which allows
arbitrary levels of nesting. The evaluation helper is already using it.

There's a conversion stage needed for the existing extraction stage
(which is currently unused), as well as for the risks from the first
iteration of the repository, which we use as ground truth for evaluating
the classifier.

There's also a new script for running the classifier by itself against
the first iteration's risks for easier tuning.

This branch was successfully deployed

1 active deployment
Preview — 3d70cc1f Deployed Aug 9, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant