beto-sentiment-spanish

BETO Sentiment Analysis — Spanish 🇪🇸

BERT-based sentiment analysis for Spanish text, fine-tuned on 50,000 IMDB movie reviews translated to Spanish.

Model Description

• Base model: dccuchile/bert-base-spanish-wwm-cased (BETO)
• Task: Sentiment Analysis (positive/negative)
• Language: Spanish
• Accuracy: 88% on test set (10,000 reviews)

How to Use

from transformers import BertTokenizer, BertForSequenceClassification
import torch
import torch.nn.functional as F

tokenizer = BertTokenizer.from_pretrained('vipont/beto-sentiment-spanish')
model = BertForSequenceClassification.from_pretrained('vipont/beto-sentiment-spanish')
model.eval()

def predict(text):
    inputs = tokenizer(text, return_tensors='pt', truncation=True, padding=True, max_length=128)
    with torch.no_grad():
        outputs = model(**inputs)
    probs = F.softmax(outputs.logits, dim=-1)
    pred = torch.argmax(probs).item()
    return 'positive' if pred == 1 else 'negative', round(probs[0][pred].item() * 100, 2)

print(predict("Me encantó la película, fue increíble"))
print(predict("Terrible, una pérdida de tiempo"))

Training

• Dataset: 50,000 IMDB reviews translated to Spanish
• Epochs: 3
• Optimizer: AdamW (lr=2e-5)
• Max length: 128 tokens

Results