Text Based Emotion Recognition

Project Domain / Category

Natural Language Processing/Deep Learning

Abstract / Introduction

Emotion plays a vital role in human communication, influencing how messages are perceived and understood. This project explores Speech Emotion Recognition (SER) from text-based data, focusing on identifying emotional states such as happiness, sadness, anger, fear, and neutrality from transcribed speech. Recognizing emotions from text can enhance various real-world applications, such as improving virtual assistants, making customer service interactions more empathetic, and supporting mental health monitoring by detecting signs of emotional distress. It can also contribute to personalized content recommendations and create more engaging human-computer interactions. By enabling machines to better understand human emotions, this project seeks to bridge the gap between technology and human empathy, making digital experiences more intuitive and supportive in everyday life. The system utilizes Natural Language Processing (NLP) techniques, including sentiment analysis and deep learning models like transformers or recurrent neural networks (RNNs), to extract contextual and semantic information from the text. Preprocessing steps like tokenization, stopword removal, and word embeddings (e.g., Word2Vec or BERT) help improve model performance. The trained model then classifies the given text into predefined emotional categories. This project aims to achieve high accuracy and robustness in recognizing emotions from diverse text data, contributing to more emotionally intelligent AI systems.

Functional Requirements:

The Admin (Student) will design and develop a system capable of performing the following tasks:

·         Split data into 70% training and 30% testing data sets.

·         Assess the model's performance using standard evaluation metrics (e.g., F1-score, precision, recall) and fine-tune the model for improved accuracy.

·         Create a confusion matrix table to describe the performance of a classification model.

·         The system should accept transcribed speech text as input and classify it into emotions like hate, neutral, anger, love, worry, relief, happiness, fun, empty, enthusiasm, sadness, surprise, and boredom.

·         Provide a confidence score (optional) and handle unclear or mixed emotions gracefully.

·         Display results clearly and user-friendly.

·         A simple interface (command-line or web-based) for input and result display, accessible to users with basic tech knowledge.

Dataset:

https://drive.google.com/file/d/1CubP0qV5vttOTPf4Pm7rCKohwNig6-2F/view?usp=sharing

*You must use your VU email id to access/download the dataset.

Tools:

Use Python with NLP libraries (e.g., NLTK, spaCy) in Jupyter Notebook, VS Code, or similar environments.

Prerequisite:

Artificial Intelligence, Machine Learning, and Natural Language Processing Concepts,

Admin (student) will cover short courses relevant to the mentioned concepts besides initial documentation, i.e. SRS and Design document.