Paraphrase Roberta Es

🚀 Prompsit/paraphrase-roberta-es

This model enables the evaluation of paraphrases for a given phrase, fine - tuned from a pre - trained model.

🚀 Quick Start

This model is designed to answer the question: Is "phrase B" a paraphrase of "phrase A"? It's important to note that the model deals with phrases rather than sentences, so it doesn't expect punctuation marks or long texts. The resulting probabilities correspond to two classes: 0 for "Not a paraphrase" and 1 for "It's a paraphrase".

✨ Features

• Fine - tuned Model: This model is fine - tuned from the pretrained "PlanTL - GOB - ES/roberta - base - bne".
• Project - based Development: It is built under a TSI - 100905 - 2019 - 4 project, co - financed by the Ministry of Economic Affairs and Digital Transformation from the Government of Spain.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

Basic Usage

import torch

from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("Prompsit/paraphrase-roberta-es")
model = AutoModelForSequenceClassification.from_pretrained("Prompsit/paraphrase-roberta-es")

input = tokenizer('se buscarán acuerdos','se deberá obtener el acuerdo',return_tensors='pt')
logits = model(**input).logits
soft = torch.nn.Softmax(dim=1)
print(soft(logits))

The output of the above code is:

tensor([[0.2266, 0.7734]], grad_fn=<SoftmaxBackward>)

In this example, since the probability of 1 (It's a paraphrase) is 0.77 and the probability of 0 (It is not a paraphrase) is 0.22, we can conclude that "se deberá obtener el acuerdo" is a paraphrase of "se buscarán acuerdos".

📚 Documentation

Evaluation Results

The model was evaluated using a test dataset of 16,500 pairs of human - tagged phrases. The following metrics were obtained:

metrics={
 'test_loss': 0.4869941473007202, 
 'test_accuracy': 0.8003636363636364, 
 'test_precision': 0.6692456479690522, 
 'test_recall': 0.5896889646357052, 
 'test_f1': 0.6269535673839184, 
 'test_matthews_correlation': 0.49324489316659575, 
 'test_runtime': 27.1537, 
 'test_samples_per_second': 607.652, 
 'test_steps_per_second': 19.003
 }