Model Card for Arabic Sentiment Muhannedsh
Model Details
Model Description
This is a fine-tuned BERT-based Arabic sentiment analysis model, specifically adapted from the aubmindlab/bert-base-arabertv02 model. It has been fine-tuned for binary sentiment classification tasks (positive vs. negative sentiment) and achieves excellent performance on the validation set.
- Developed by: Muhanned Shaheen
- Model type: BERT-based model for sequence classification
- Language(s) (NLP): Arabic
- License: Apache 2.0
- Finetuned from model:
aubmindlab/bert-base-arabertv02
Model Sources
Training Metrics
- Training Loss: 0.0315
- Training Accuracy: 99.31%
- Training F1-Score: 99.28%
Validation Metrics
- Validation Loss: 0.2464
- Validation Accuracy: 92.24%
- Validation F1-Score: 92.89%
Uses
Direct Use
This model is intended to be used for binary sentiment analysis tasks in Arabic. It can classify Arabic text into positive or negative sentiment.
Downstream Use
The model can be fine-tuned further for other tasks in Arabic text classification or sentiment analysis.
Out-of-Scope Use
The model is not recommended for tasks involving non-Arabic text or for sentiment analysis with more than two classes.
Bias, Risks, and Limitations
Recommendations
Users should be aware that the model's performance is tied to the data used during fine-tuning. Biases in the dataset could affect predictions.
How to Get Started with the Model
from transformers import AutoTokenizer, AutoModelForSequenceClassification
model_name = "muhannedshaheen/Arabic_Sentiment_Muhannedsh"
# Load the model and tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)
# Example
text = "هذا المنتج رائع جدا"
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)
print(outputs.logits)
- Downloads last month
- 22