openreview-expertise
openreview-expertise copied to clipboard
openreview
Expertise modeling for the OpenReview matching system
This PR no longer attempts to find and remove all old instances of publications in the Redis cache, and instead sets an expiration date whenever inserting into Redis.
This PR adds additional startup code to Flask and Celery that marks all jobs running at the time of a server restart with ERROR which will cause process functions to...
This PR catches if the provided invitation does not exist - in this case, the API will treat the job as if no invitation was provided.
There is a pattern where profiles with dots in the IDs in the match group will not be included in the score file
In the case where the server is forced to restart with jobs still in the queue, jobs are still marked in Redis as running and will queue themselves back into...
The expertise API was built without taking into account retrieving sparse scores for large conferences. This option is available in the API but the backend does not search for this...
Some error messages from the expertise model can be made into something user friendly, we can detect the model errors and translate them into an error that the user can...
When retrieving all publications as submissions, we only use iterget with no checking of the paperhash and expertise selection. Make the same checks that happen when retrieving publications as expertise.
We need to have access to a dashboard to monitor the Celery queue. We need to see the jobs that are failed, queued, complete and running. The endpoint could be...
Metadata
Owner
Metadata
Expertise modeling for the OpenReview matching system