add model card
Browse files
README.md
ADDED
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: gos
|
3 |
+
tags:
|
4 |
+
- BERTje
|
5 |
+
- pos
|
6 |
+
---
|
7 |
+
|
8 |
+
Wietse de Vries β’ Martijn Bartelds β’ Malvina Nissim β’ Martijn Wieling
|
9 |
+
|
10 |
+
# Adapting Monolingual Models: Data can be Scarce when Language Similarity is High
|
11 |
+
|
12 |
+
This model is part of this paper + code:
|
13 |
+
|
14 |
+
- π [Paper](https://arxiv.org/abs/2105.02855)
|
15 |
+
- π» [Code](https://github.com/wietsedv/low-resource-adapt)
|
16 |
+
|
17 |
+
## Models
|
18 |
+
|
19 |
+
The best fine-tuned models for Gronings and West Frisian are available on the HuggingFace model hub:
|
20 |
+
|
21 |
+
### Lexical layers
|
22 |
+
These models are identical to [BERTje](https://github.com/wietsedv/bertje), but with different lexical layers (`bert.embeddings.word_embeddings`).
|
23 |
+
|
24 |
+
- π€ [`GroNLP/bert-base-dutch-cased`](https://huggingface.co/GroNLP/bert-base-dutch-cased) (Dutch; source language)
|
25 |
+
- π€ [`GroNLP/bert-base-dutch-cased-gronings`](https://huggingface.co/GroNLP/bert-base-dutch-cased-gronings) (Gronings)
|
26 |
+
- π€ [`GroNLP/bert-base-dutch-cased-frisian`](https://huggingface.co/GroNLP/bert-base-dutch-cased-frisian) (West Frisian)
|
27 |
+
|
28 |
+
### POS tagging
|
29 |
+
These models share the same fine-tuned Transformer layers + classification head, but with the retrained lexical layers from the models above.
|
30 |
+
|
31 |
+
- π€ [`GroNLP/bert-base-dutch-cased-upos-alpino`](https://huggingface.co/GroNLP/bert-base-dutch-cased-upos-alpino) (Dutch)
|
32 |
+
- π€ [`GroNLP/bert-base-dutch-cased-upos-alpino-gronings`](https://huggingface.co/GroNLP/bert-base-dutch-cased-upos-alpino-gronings) (Gronings)
|
33 |
+
- π€ [`GroNLP/bert-base-dutch-cased-upos-alpino-frisian`](https://huggingface.co/GroNLP/bert-base-dutch-cased-upos-alpino-frisian) (West Frisian)
|