NLP Emotion Classification Project App

Welcome to the NLP Emotion Classification project! This project focuses on classifying text data into six distinct emotion categories using various machine learning models. Below you'll find detailed information on how to set up the project, use the provided CLI tool, understand the project's overall functionality and results, and explore the Streamlit web application.

Table of Contents

1. Project Overview
2. Dataset
3. Preprocessing
4. Models Used
5. Results
6. CLI Tool
   • Description
   • Functionality
   • Sample Inputs and Outputs
   • Usage Instructions
7. Streamlit Web Application
   • Features
   • Usage
8. Setup and Installation
9. Contact

Project Overview

This project aims to classify text into emotions such as sadness, joy, love, anger, fear, and surprise using machine learning techniques. The dataset is sourced from Kaggle and consists of text data labeled with these six emotion categories.

Dataset

• Source: Kaggle Emotion Dataset
• Size: 416,808 entries
• Features: Text data and corresponding emotion labels

Class Labels and Emotions

Label	Emotion
0	Sadness
1	Joy
2	Love
3	Anger
4	Fear
5	Surprise

Preprocessing

The preprocessing steps include:

• Text Cleaning: Removing unnecessary characters, converting to lowercase, removing punctuations and URLs.
• Tokenization: Splitting text into words or tokens for analysis.
• Vectorization: Converting text data into numerical representations using Bag-of-Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF).

Models Used

The project implements several models to predict emotions:

1. Naive Bayes:

   • Accuracy: 74%
   • Simple probabilistic model

2. Logistic Regression:

   • Accuracy: 89.44%
   • Statistical model for binary outcomes

3. XGBoost:

   • Accuracy: 89.39%
   • Optimized gradient boosting algorithm

4. Custom Neural Network:

   • Accuracy: 88.66%
   • Multi-layer perceptron with Leaky ReLU and Sigmoid activation functions

Results

Visual Results

Model Performance

Model	Accuracy	Precision	Recall	F1-Score
Naive Bayes	0.74	0.79	0.74	0.69
Logistic Regression	0.8944	0.89	0.89	0.89
XGBoost	0.8939	0.90	0.89	0.90
Neural Network	0.8866	0.89	0.89	0.89

CLI Tool

The CLI (Command-Line Interface) tool allows users to input a sentence and get an emotion prediction using trained machine learning models.

Description

The tool interacts with the following models to predict emotions:

• XGBoost
• Logistic Regression
• Naive Bayes

Functionality

1. Text Preprocessing: Cleans input text.
2. Vectorization: Converts text into numerical format.
3. Prediction: Uses models to predict emotions.
4. User Interaction: Accepts user input and returns predicted emotions.

Sample Inputs and Outputs

• Input: "I am so happy and surprised you did this!"

  • XGBoost: Joy
  • Logistic Regression: Joy
  • Naive Bayes: Surprise

• Input: "I am already feeling frantic."

  • XGBoost: Fear
  • Logistic Regression: Fear
  • Naive Bayes: Fear

Usage Instructions

1. Run the script using Python.
2. Enter a sentence to get emotion predictions.
3. The tool displays predictions from each model.
4. Type exit to quit the tool.

Streamlit Web Application

Emotion Classifier App

Explore the Emotion Classifier app hosted on Streamlit Cloud here. This interactive web application allows users to input text and receive predictions for emotions such as sadness, joy, love, anger, fear, and surprise using machine learning models.

Features:

• Input: Enter a sentence to predict the corresponding emotions.
• Models Used: Naive Bayes, Logistic Regression, and XGBoost.
• Visualization: Displays predictions, model performance comparison, confusion matrices, class distributions, word clouds, and ROC curves.

Setup and Installation

1. Clone the repository:

   git clone https://github.com/parthrastogicoder/NLP_EMOTION_CLASSIFIER.git
   cd emotion-classification

2. Install dependencies:

   pip install -r requirements.txt

3. Download the dataset from Kaggle

4. Run the CLI tool:

   python cliapp.py

5. Explore the Streamlit app:

   streamlit run app.py

Contact

For any inquiries or suggestions regarding this project, please contact me at [email protected].