Improve dataset card: update task category, license, add tags and sample usage
#1
by
nielsr
HF Staff
- opened
This PR updates the dataset card for orionweller/LIMIT-small with several improvements:
- Updates
task_categories: Changed fromquestion-answeringtotext-rankingto more accurately reflect the dataset's role in benchmarking embedding-based retrieval models. - Corrects
license: Updated fromapache-2.0tocc-by-4.0, as specified in the GitHub repository's license information for non-software materials (datasets). - Adds
tags: Includedretrieval,embeddings, andtheoretical-limitationsfor better discoverability and context. - Adds Paper Abstract: The paper abstract is included as an "Introduction" section to provide a comprehensive overview of the dataset's theoretical underpinnings and purpose.
- Adds Sample Usage: Included Python code snippets directly from the GitHub README to demonstrate how to load the dataset using the
datasetslibrary and how to perform evaluation using the MTEB framework.
These changes enhance the dataset card's accuracy, completeness, and usability for the community.
Hi @nielsr , thanks for the PR.
Any chance HF can just add retrieval? It's a bit silly after so many years that retrieval doesn't have it's own task... it feels like like it's a big enough community (more so than many of the tags currently there). Let me know where I need to make a PR
orionweller
changed pull request status to
merged
Oh we do actually: https://huggingface.co/datasets?task_categories=task_categories:text-retrieval&sort=trending
Will have to adjust the prompt so that PRs like these can take that into account :D
Feel free to update the dataset card accordingly.