TicketHub: Enabling Actionable Analysis of Support Requests With NLP
Sarah Rodenbeck, Yoon Suk Uhr · Practice and Experience in Advanced Research Computing · 2023
This paper presents TicketHub, an application that uses modern Natural Language Processing (NLP) techniques to provide a dynamic, temporal, and holistic view of High-Performance Computing (HPC) support requests (tickets). While the project was originally developed to derive insights into tickets submitted to Purdue's Rosen Center for Advanced Computing (RCAC), the architecture was designed to easily integrate with additional ticket sources and platforms. This project extends, and automates, previous ticket analysis efforts by utilizing modern semantic and unsupervised NLP techniques to better understand the context of messages, while remaining robust with respect to shifting ticket topics over time.