metadata
language:
- bn
- en
tags:
- sentiment-analysis
- bengali
- code-mixing
- multilingual
datasets:
- Swarnadeep-28/bn_code_mix_sentiment_dataset
metrics:
- accuracy
- f1
Bengali Code-Mixed Sentiment Analysis Model
Model Description
A multilingual sentiment analysis model configured for Bengali code-mixed text (Bengali-English code-switching). Classifies text into four sentiment categories:
- Negative: Expresses negative sentiment
- Neutral: Expresses neutral sentiment
- Positive: Expresses positive sentiment
- Mixed: Contains mixed or ambiguous sentiment
Usage
from transformers import pipeline
classifier = pipeline(
"sentiment-analysis",
model="OsamaBinLikhon/bengali-code-mix-sentiment"
)
result = classifier("আজকের দিনটা খুবই ভালো গেছে!")
Training Data
- Dataset: Swarnadeep-28/bn_code_mix_sentiment_dataset
- Samples: 16,012 training, 2,001 validation, 2,002 test
Performance
Configured for Bengali sentiment analysis with 4-class classification.
Citation
@misc{bengali-sentiment-model,
author = [MiniMax Agent],
title = [Bengali Code-Mixed Sentiment Analysis Model],
url = [https://huggingface.co/OsamaBinLikhon/bengali-code-mix-sentiment],
year = [2025]
}
License
Apache License 2.0