Naive Bayes Classifier for 3-Label Sentiment Classification

This repository contains the homework assignment for the AI course, which we implemented, focusing on a Naive Bayes classifier for sentiment classification with three labels: Positive, Negative, and Neutral. The assignment included training, evaluation, and testing subsets of the dataset.

Table of Contents

• Introduction
• Dataset
• Assignment Details
  • Naive Bayes Classifier Structure
  • 3-Label Classification
• Evaluation

Introduction

This homework assignment aimed to help students understand and implement a Naive Bayes classifier for sentiment analysis. The task was to classify text data into three sentiment labels: Positive, Negative, and Neutral. We provided students with training and evaluation subsets of the dataset and expected them to implement a Naive Bayes classifier with Laplacian smoothing.

Dataset

The dataset was divided into three subsets:

• Train Set: Used for training the model.
• Eval Set: Used for evaluating the model during development.
• Test Set: Used by TAs to evaluate the final submissions after the deadline.

Assignment Details

Naive Bayes Classifier Structure

The homework PDF we provided explained the structure of a Naive Bayes binary classifier. Students were required to extend this structure to handle three labels. The key components included:

• Tokenization: Breaking down text into individual words or tokens.
• Probability Calculation: Calculating the probability of each word given a class.
• Classification: Using the calculated probabilities to classify new text data.

3-Label Classification

Students needed to adapt the Naive Bayes classifier to handle three sentiment labels:

• Positive
• Negative
• Neutral

Evaluation

After the submission deadline, we evaluated the models using a test set.