Developing E-mail Classification Model Using Sentiment Analysis to Improve Customer Support

Companies often utilize sentiment analysis in aiding their marketing and business strategy, customer service, and much more. More often as they reach a certain amount of users their requests and issues multiply requiring a large customer service team that would be ready to respond. The objective of this paper is to overview different techniques that use sentiment analysis in order to find the best one to build a classification model for customer support e-mails. The classification model is envisioned to harness data-driven insights for the tagging and prioritization of e-mails, gauging user satisfaction levels, and seamless integration with other processes, such as chat agents and business analytics. The critical determinant in the selection of an appropriate methodology is grounded not only in the inherent characteristics of the dataset deployed for classification model training but also in the specific domain and the anticipated research outcomes. With due consideration to these factors and related research, the present study opts for an approach where data preprocessing is prioritized, TF-IDF is used for feature extraction, and SVM for classification. In addition, two types of RNN, a deep learning method, were employed in comparison to SVM for observation, as it has demonstrated effectiveness in similar tasks.