Tweet Sentiment Analysis

Overview

This project implements a Naive Bayes classifier for sentiment analysis on tweets. The classifier is trained on a dataset containing labeled tweets categorized as positive, negative, or neutral sentiments. After training, the classifier can analyze new tweets and classify them into one of these sentiment categories.

Table of Contents

• Introduction
• Installation
• Usage
• Files
• Data Files
• Algorithm Details
• Customization

Introduction

The project consists of two main files: run.py and template.py. The run.py script contains the main functionality to train, test, evaluate accuracy, and analyze tweets using the Naive Bayes classifier implemented in template.py.

Installation

To run the project, follow these steps:

1. Clone the repository:

   git clone https://github.com/mobinbr/tweet-sentiment-analysis

2. Navigate to the project directory:

   cd tweet-sentiment-analysis

3. Install the required dependencies:

   pip install -r requirements.txt

Usage

1. Prepare your data:
   Ensure you have the following CSV files:
   train_data.csv: Contains labeled training data.
   eval_data.csv: Contains evaluation data for calculating accuracy.
   test_data_nolabel.csv: Contains unlabeled test data to be classified.

2. Run the training script:

   python run.py

This will train the Naive Bayes classifier on the provided training data, evaluate its accuracy using the evaluation data, and generate a classification result for the test data.

3. Analyze tweets:
   After running the script, you can analyze new tweets by calling the classify method of the trained classifier. For example:

   test_string = "I love playing football"
   print(nb_classifier.classify(preprocess(test_string)))

Files

• run.py: Script for preprocessing data, training the classifier, evaluating accuracy, and analyzing tweets.
• template.py: Implementation of the NaiveBayesClassifier class for sentiment analysis.

Data Files

train_data.csv: Labeled training data containing tweets and sentiment labels. eval_data.csv: Labeled evaluation data for calculating accuracy. test_data_nolabel.csv: Unlabeled test data for testing the classifier.

Algorithm Details

The sentiment analysis algorithm is based on the Naive Bayes classifier. Key components include:

Preprocessing: Tweets are preprocessed to remove noise, such as special characters and stopwords, and tokenized into words for analysis.

Training: The Naive Bayes classifier is trained using labeled tweet data. It calculates the likelihood of each word occurring in each sentiment category and determines the prior probabilities of each category.

Classification: Given a new tweet, the classifier calculates the probability of it belonging to each sentiment category based on the trained model and assigns it to the category with the highest probability.

Customization

You can customize the project by adjusting parameters in the run.py file, such as file paths for data, and by modifying hyperparameters in the template.py file, such as smoothing techniques or feature selection methods in the Naive Bayes classifier. Experiment with different datasets and preprocessing techniques to improve the accuracy of sentiment analysis.