Adapting Monolingual Models: Data can be Scarce when Language Similarity is High
Paper • 2105.02855 • Published
# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("GroNLP/bert-base-dutch-cased-frisian")
model = AutoModelForMaskedLM.from_pretrained("GroNLP/bert-base-dutch-cased-frisian", device_map="auto")Quick Links
Wietse de Vries • Martijn Bartelds • Malvina Nissim • Martijn Wieling
Adapting Monolingual Models: Data can be Scarce when Language Similarity is High
This model is part of this paper + code:
Models
The best fine-tuned models for Gronings and West Frisian are available on the HuggingFace model hub:
Lexical layers
These models are identical to BERTje, but with different lexical layers (bert.embeddings.word_embeddings).
- 🤗
GroNLP/bert-base-dutch-cased(Dutch; source language) - 🤗
GroNLP/bert-base-dutch-cased-gronings(Gronings) - 🤗
GroNLP/bert-base-dutch-cased-frisian(West Frisian)
POS tagging
These models share the same fine-tuned Transformer layers + classification head, but with the retrained lexical layers from the models above.
- 🤗
GroNLP/bert-base-dutch-cased-upos-alpino(Dutch) - 🤗
GroNLP/bert-base-dutch-cased-upos-alpino-gronings(Gronings) - 🤗
GroNLP/bert-base-dutch-cased-upos-alpino-frisian(West Frisian)
- Downloads last month
- 63
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("fill-mask", model="GroNLP/bert-base-dutch-cased-frisian")