AutoTriagr

GitHub Link: https://github.com/LeeMcKusick/fs-autotriage

AutoTriagr is an unfinished app that used natural language processing to examine the contents of a ticket, evaluate it for topic and urgency, and automatically update a ticket in Zendesk with that information. The goal was to free up a large portion of the triager’s role to provide more time to work on the tickets themselves.

How it works: Ticket data is sent to the classifier app. This uses the “natural” library for NodeJS (https://github.com/NaturalNode/natural) to search the subject, description, and other relevant fields and classify it for both topic and urgency, using a logistic regression algorithm. It then outputs those values with a reliability rating. If that rating is high enough, it would be automatically updated; otherwise, it would be flagged for a human to examine manually.*

Future Plans: I believe development stopped on this project when I left my position at Finalsite.  Remaining work on this includes:

• Connecting it to Zendesk to push ticket updates.
• Continuing to update the classifier training files to improve accuracy.

* This would also happen if the predicted urgency level was high enough. We wanted to maintain a level of human oversight to avoid incorrect classification, as well as potentially urgent tickets being overlooked.