BanglaBERT Sentiment Analysis

Overview

This repository contains code and resources for a sentiment analysis project using the BanglaBERT model from csebuetnlp/banglabert. The model is trained on the SentNoB dataset, where labels 0, 1, or 2 represent neutral, positive, and negative sentiments, respectively. The goal is to analyze sentiment in Bangla text using state-of-the-art natural language processing techniques.

Objectives

• Utilize the BanglaBERT model for sentiment analysis.
• Train and fine-tune the model on the SentNoB dataset.
• Evaluate sentiment classification performance using accuracy and other relevant metrics.

Dataset

The SentNoB dataset is employed for training and evaluation, with labels '0' for neutral, '1' for positive, and '2' for negative sentiments.
Dataset link: SentNoB

Model

The BanglaBERT model from csebuetnlp is chosen for its effectiveness in understanding and classifying sentiment in Bangla text.

Performance Evaluation:

• Evaluate the model using fundamental metrics:
  • Accuracy
  • Precision
  • Recall
  • F1-score