A Comprehensive Analysis of Sentiment Evaluation for Email Categorization

Email is one of the most dependable means of online communication and has become an indispensable messaging tool for most organizations and individuals. Email is reliable and most convenient way of communication as it is quick, simple and accessible from anywhere in the globe with just an internet connection. This is also able to keep a comprehensive record for future reference. Now, cyberattacks easily target the emails to fulfil their evil goals through fraudulent activities. Recently, a different problem is discovered in a customer service centre of an organization that is to understand the social sentiment of a customer regarding the product or service related with appreciations, complaints etc. So, an efficient and automatic management system is required to arrest and resolve these types of issues. This research work focuses on sentiment analysis of email contents to classify the emails by using Natural Language Processing (NLP) tools and Machine Learning (ML) algorithms. We adopted TF-IDF for feature extraction, various machine learning models for classification and observed that using TF-IDF, Extreme Gradient Boosting (XGBoost) achieves better result than other ML models with 91% accuracy.