osiria's picture
Update README.md
bacda33
|
raw
history blame
2.76 kB
metadata
license: apache-2.0
language:
  - it
widget:
  - text: >-
      una fantastica [MASK] di #calcio! grande prestazione del mister e della
      squadra
    example_title: Example 1
  - text: >-
      il governo [MASK] dovrebbe fare politica, non soltanto propaganda!
      #vergogna
    example_title: Example 2
  - text: >-
      che serata da sogno sul #redcarpet! grazie a tutti gli attori e registi
      del [MASK] italiano #oscar #awards
    example_title: Example 3


  
    Model: BERT-TWEET
    Lang: IT
  

Model description

This is a BERT [1] uncased model for the Italian language, obtained using TwHIN-BERT [2] (twhin-bert-base) as a starting point and focusing it on the Italian language by modifying the embedding layer (as in [3], computing document-level frequencies over the Wikipedia dataset)

The resulting model has 110M parameters, a vocabulary of 30.520 tokens, and a size of ~440 MB.

Quick usage

from transformers import BertTokenizerFast, BertModel

tokenizer = BertTokenizerFast.from_pretrained("osiria/bert-tweet-base-italian-uncased")
model = BertModel.from_pretrained("osiria/bert-tweet-base-italian-uncased")

Here you can find the find the model already fine-tuned on Sentiment Analysis: https://huggingface.co/osiria/bert-tweet-italian-uncased-sentiment

References

[1] https://arxiv.org/abs/1810.04805

[2] https://arxiv.org/abs/2209.07562

[3] https://arxiv.org/abs/2010.05609

Limitations

This model was trained on tweets, so it's mainly suitable for general-purpose social media text processing, involving short texts written in a social network style. It might show limitations when it comes to longer and more structured text, or domain-specific text.

License

The model is released under Apache-2.0 license