Email Urgency Classifier Using Natural Language Processing and Naïve Bayes

Abstract

Emails are today's most commonly used means of communication. It is one of the Internet's greatest inventions. Billions of emails are transferred daily for various purposes. The Harvard Business Review states that after an interruption (e-mail or other), it takes 20 minutes to regain full focus. This is backed by the argument of Loughborough University that when interrupted by phone, tasks take 33 percent longer to complete. Hence, people are constrained on how to manage their time effectively; this is because a piece of time-critical information may be received and needed to be worked upon urgently.  Large organizations receive thousands of emails every day; they cannot decide on which email to respond to first because the emails are already in a hierarchy, based on the time they were received. Hence, emails that are time-critical or require urgent response may be pushed to the bottom of the list. The aim of this paper is to present a model that can interpret human language and efficiently discern between urgent and not urgent emails. When emails are sent, the model prioritizes which emails should be responded to by the receiver based on the level of urgency and importance of the content of the email. The tools used to classify the emails into urgent and not urgent are Natural Language processing and Naïve Bayes for text classification. The model is created using Python programming language on Spyder (a Python Integrated Development Environment (IDE)) and Scikit-learn (an efficient Python data mining and analysis tool). The result obtained is the percentage (polarity) of the urgency of the email, hence urgent emails are placed on top of the inbox list. This paper contributes to knowledge by the creation of a dataset on Kaggle. There is unavailable dataset for this project, I had to create a new one