BERTimbau base for Semantic Textual Similarity

This is the neuralmind/bert-base-portuguese-cased model finetuned for Semantic Textual Similarity with the ASSIN 2 dataset. This model is suitable for Portuguese.

• Git Repo: Evaluation of Portuguese Language Models.
• Demo: Portuguese Semantic Similarity

Full regression example

from transformers import AutoModelForSequenceClassification, AutoTokenizer, AutoConfig
import numpy as np
import torch

model_name = "ruanchaves/bert-base-portuguese-cased-assin2-similarity"
s1 = "A gente faz o aporte financeiro, é como se a empresa fosse parceira do Monte Cristo."
s2 = "Fernando Moraes afirma que não tem vínculo com o Monte Cristo além da parceira."
model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
config = AutoConfig.from_pretrained(model_name)
model_input = tokenizer(*([s1], [s2]), padding=True, return_tensors="pt")
with torch.no_grad():
    output = model(**model_input)
    score = output[0][0].detach().numpy().item()
    print(f"Similarity Score: {np.round(float(score), 4)}")

Citation

Our research is ongoing, and we are currently working on describing our experiments in a paper, which will be published soon. In the meanwhile, if you would like to cite our work or models before the publication of the paper, please cite our GitHub repository:

@software{Chaves_Rodrigues_eplm_2023,
author = {Chaves Rodrigues, Ruan and Tanti, Marc and Agerri, Rodrigo},
doi = {10.5281/zenodo.7781848},
month = {3},
title = {{Evaluation of Portuguese Language Models}},
url = {https://github.com/ruanchaves/eplm},
version = {1.0.0},
year = {2023}
}