Sentiment Analysis in Tickets for IT Support

Sentiment analysis has been adopted in software engineeringfor problems such as software usability and sentimentof developers in open-source projects. This paper proposesa method to evaluate the sentiment contained in tickets forIT (Information Technology) support.IT tickets are broadin coverage (e.g. infrastructure, software), and involve errors,incidents, requests, etc. The main challenge is to automaticallydistinguish between factual information, whichis intrinsically negative (e.g. error description), from thesentiment embedded in the description. Our approach isto automatically create a Domain Dictionary that containsterms with sentiment in the IT context, used to filter termsin ticket for sentiment analysis. We experiment and evaluatethree approaches for calculating the polarity of terms intickets. Our study was developed using 34,895 tickets fromfive organizations, from which we randomly selected 2,333tickets to compose a Gold Standard. Our best results displayan average precision and recall of 82.83% and 88.42%, whichoutperforms the compared sentiment analysis solutions.