ZamAI-Sentiment-Pashto

Model Description

Sentiment analysis model for Pashto text classification and emotion detection.

This model is part of the ZamAI (زمای) project - an advanced Afghan AI assistant designed to understand and communicate in Pashto, English, and other Afghan languages.

Key Features

Multi-class sentiment classification
Emotion detection
Cultural context understanding
Social media text analysis
Real-time sentiment scoring

Use Cases

Social media monitoring
Customer feedback analysis
Content moderation
Market research
Opinion mining

Model Architecture

Base Model: cardiffnlp/twitter-roberta-base-sentiment-latest
Architecture: roberta
Task: sentiment-analysis
Languages: Pashto (ps), English (en)

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load model and tokenizer
model_name = "tasal9/zamai-sentiment-pashto"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Generate text
prompt = "سلام! زه د افغانستان په اړه پوښتنه لرم:"
inputs = tokenizer.encode(prompt, return_tensors="pt")

with torch.no_grad():
    outputs = model.generate(
        inputs,
        max_length=200,
        temperature=0.8,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id
    )

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Training Details

Dataset: ZamAI Pashto Dataset (tasal9/ZamAI_Pashto_Dataset)
Training Method: Classification fine-tuning
Epochs: 6
Batch Size: 8
Learning Rate: 2e-5

Performance

The model has been trained on conversational Pashto data and shows strong performance in:

Natural conversation flow
Cultural context understanding
Mixed language handling (Code-switching)
Afghan cultural knowledge

Limitations

Primary focus on Pashto and English
May require further fine-tuning for specific domains
Performance may vary with complex technical terminology

Ethical Considerations

This model is designed to respect Afghan and Islamic values, promoting positive and constructive conversations while avoiding harmful or inappropriate content.

Citation

@misc{zamai_zamai_sentiment_pashto_2024,
  title={ZamAI ZamAI-Sentiment-Pashto: Advanced Pashto Language Model},
  author={ZamAI Team},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/tasal9/zamai-sentiment-pashto}
}

Contact

For questions, suggestions, or collaboration opportunities, please reach out through the ZamAI project.

Built with ❤️ for the Afghan community

tasal9
/

ZamAI-Sentiment-Pashto