Detecting Urgency Status of Social Media Based Customer Support Requests

Abstract

"Customer support platforms of the digital product-based companies are frequently handling high number of service requests (support tickets) receiving from their product consumers on a daily basis. Sometimes it is practically impossible to identify the support requests that needs to be attended on a high priority, without properly inspecting details in the support ticket. As of today, most of the companies in the customer service domain follows a manual approach to evaluate and categorize the customer support tickets at the initial stages of the support process.

After spending a considerable amount of time for evaluate and classify the support tickets, they are dispatched to the responsible parties in the support platform to further investigation and finally customer service agents are reaching out to the customer. This is a very time-consuming process, and also identified to be one of the major reasons for support platforms taking longer period of time to attend to the customer reported issues. The author of this research project presents an efficient business solution to overcome this problem by utilizing the urgency detection research area in NLP domain. Author introduces a novel, minimally supervised learning technique which will only require a limited number of properly labelled data to train a high performing deep learning classifier which is capable of identifying the support tickets / messages in different levels of urgency.

Furthermore, author is focused on enhancing the performance of this classification model by extending its architecture to allow accepting a higher volume of data input and generate the classification results in a smaller amount of time. Author is also carrying out a comprehensive evaluation of the developed model with benchmarking it against the top performing text classification models introduced in recent research projects in the natural language processing domain."