Sentiment Analysis on Tweets

Table of Contents

• Introduction
• Project Overview
• Features
• Model Architecture
• Data Interpretation and Preprocessing
• Training the Model
• Results
• Installation
• Contributing
• License

Introduction

This project is a sentiment analysis model applied to tweets. It uses machine learning and natural language processing (NLP) techniques to classify the sentiment of tweets into three categories: positive, negative, and neutral. The project demonstrates the use of various preprocessing steps, TF-IDF vectorization, and a deep learning model built with TensorFlow and Keras.

Project Overview

Sentiment analysis is a key aspect of NLP that involves determining the sentiment expressed in a piece of text. This project leverages the power of deep learning and TF-IDF vectorization to perform sentiment analysis on a dataset of tweets.

Features

• Data Preprocessing: Clean and preprocess raw tweet data, including removing stopwords and lemmatization.
• TF-IDF Vectorization: Convert the cleaned text data into numerical form using TF-IDF.
• Deep Learning Model: Train a neural network model to classify sentiments.
• Visualization: Plot accuracy and loss metrics to evaluate model performance.

Model Architecture

The model used in this project is a sequential neural network with the following layers:

• Dense Layer: 100 units with ReLU activation
• Dropout Layer: 0.3 dropout rate
• Dense Layer: 25 units with ReLU activation
• Dropout Layer: 0.3 dropout rate
• Dense Layer: 10 units with ReLU activation
• Dropout Layer: 0.3 dropout rate
• Output Layer: 3 units with softmax activation for multiclass classification

Data Interpretation and Preprocessing

We have an image depicting a dataframe and a list of features. We will utilize the 'text' feature as input and consider the 'sentiment' feature as our target variable. Our goal is to predict the likelihood of a text being categorized as positive, negative, or neutral.

The data preprocessing pipeline includes:

• Lowercasing: Convert all text to lowercase.
• Contraction Expansion: Expand contractions (e.g., "can't" to "cannot").
• Special Character Removal: Remove non-alphanumeric characters.
• Tokenization: Split text into words.
• Stopword Removal: Remove common stopwords.
• Lemmatization: Reduce words to their base or root form.

Training the Model

The model is trained using the following configuration:

• Loss Function: Categorical Crossentropy
• Optimizer: Adam with a learning rate of 0.01
• Metrics: Accuracy
• Epochs: 10

Results

The performance of the model is evaluated using accuracy and loss plots. The final trained model is able to classify tweet sentiments with a reasonable accuracy.

Installation

To run this project locally, follow these steps:

1. Clone the repository:

  git clone https://github.com/Vidit-Kushwaha/Tweet-Sentimental-Analysis.git
  cd Tweet-Sentimental-Analysis

2. Run the Jupyter Notebook:

  jupyter notebook Sentiment_Analysis_on_Tweets.ipynb

Train the model: Follow the steps in the notebook to preprocess the data, train the model, and visualize the results.

Contributing

We welcome contributions! Whether you're a seasoned developer or a curious enthusiast, there are ways to get involved:

• Bug fixes and improvements: Find any issues? Submit a pull request!
• New features: Have an idea for a cool feature? Let's discuss it in an issue!
• Documentation: Improve the project's documentation and website.
• Spread the word: Share the project with your network and help it grow!

You can follow standard python contribution guidelines.

License

This project is licensed under the MIT License. See the LICENSE file for details.