A NLP-based system that automatically categorises and assigns technical support tickets in the context of an IT service provider.
The micro-service is based on Python 3.6 and Azure Functions, and it is potentially capable to operate with any ticketing system provided with a RESTful API (and, ideally, webhooks). The modules that compose the system are loosely coupled, so that the system can be extended easily. An example of implementation is provided for Zammad, the open-source ticketing system.
Provided with a dataset of existing tickets, it is possible to train the classifier (based on the Multinomial Naive-Bayes model) to predict metadata such as:
- problem type;
- which team the ticket should be assigned to;
- which member(s) of the team are the most relevant for the specific ticket.
The system offers various possibilities for expansion (e.g adding the functions to determine the urgency of a ticket).