Flair: Spanish CoNLL-2002 official results

[aconneau]

What is the official result of Flair on the Spanish CoNLL-2002 dataset?
Couldn't find it here.

Thank you!
Alexis

[alanakbik]

Hi @aconneau we haven't done experiments on Spanish CoNLL-02, but would be happy to post numbers if someone in the community does the experiments!

[stefan-it]

@aconneau I'll add the results in my experiments repo the next days.

I already did some experiments with the recently added NER fine-tuning code in 🤗/Transformers, see them here for comparison :)

[stefan-it]

@aconneau without hyper-parameter tuning and using the default parameters (as used in the English CoNLL training example):

| Embeddings | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg.
| -- | -- | -- | -- | -- | -- | --
| Flair Embeddings (Test) | 85.1 | 86.21 | 86.44 | 86.41 | 85.92 | 86.016
| Word Embeddings + Flair Embeddings (Test) | 86.8 | 87.52 | 87.63 | 87.29 | 87.7 | 87.388

[stefan-it]

Btw: just found your "Unsupervised Cross-lingual Representation Learning at Scale" paper - really great work, I'm excited to integrate XLM-R into Flair!

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.