huggingface / huggingface/evaluate
Add Average Normalized Levenshtein Similarity (ANLS)
Open
metric request
- Dominant language
- Python
- Stars
- 2.5k
- Forks
- 341
- PR merge metrics
- No merged PRs in 30d
Description
Document visual question answering (DocVQA) models, like LayoutLM and Donut, are evaluated using the Average Normalized Levenshtein Similarity metric (ANLS for short).
Would be very useful for the community to have this metric in the evaluate library.
The metric was introduced here for DocVQA: https://rrc.cvc.uab.es/?ch=17&com=tasks.
Contributor guide
Assessment
This issue has not been assessed yet.