README.md

---
library_name: transformers
tags:
- bangla
- bangla-classifier
- multiclass-classifier
- text-classifier
datasets:
- SayedShaun/sentigold
language:
- bn
metrics:
- accuracy
base_model:
- csebuetnlp/banglabert
pipeline_tag: text-classification
---

# Bangla Binary Text Classifier

## Description

This is a **Bangla binary sentiment classification** model, fine-tuned on top of [`csebuetnlp/banglabert`](https://huggingface.co/csebuetnlp/banglabert). The model was trained using the [**SayedShaun/sentigold**](https://huggingface.co/datasets/SayedShaun/sentigold)

---

## How to Use

```python
from transformers import pipeline

pipe = pipeline("text-classification", model="SayedShaun/bangla-classifier-multiclass")

response = pipe("ডেলিভারি ম্যান খুব যত্ন সহকারে পণ্যটি ডেলিভারি করেছে")
print(response)
# Output: [{'label': 'LABEL_0', 'score': 0.9503920674324036}]
```

## Tags
```
{"SP" :0, "WP": 1, "WN": 2, "SN": 3, "NU": 4}

SP: Strongly Positive
WP: Weakly Positive
WN: Weakly Positive Negative
SN: Strongly Negative
NU: Neutral
```

## Result
| Training Loss | Validation Loss | Accuracy  | Precision | Recall   | F1 Score  |
|---------------|-----------------|-----------|-----------|----------|-----------|
| 0.820600      | 0.916846        | 0.646714  | 0.649295  | 0.642749 | 0.643535  |


# Source Code
Source code can be found in `files and versions` as `finetune.py`