Analyzing, Classifying, and Interpreting Emotions in Software Users' Tweets

Abstract

Twitter enables software developers to track users'reactions to newly released systems. Such information, oftenexpressed in the form of raw emotions, can be leveraged to enablea more informed software release process. However, automaticallycapturing and interpreting multi-dimensional structures ofhuman emotions expressed in Twitter messages is not a trivialtask. Challenges stem from the scale of the data available, itsinherently sparse nature, and the high percentage of domainspecificwords. Motivated by these observations, in this paperwe present a preliminary study aimed at detecting, classifying, and interpreting emotions in software users' tweets. A datasetof 1000 tweets sampled from a broad range of software systems'Twitter feeds is used to conduct our analysis. Our results showthat supervised text classifiers (Naive Bayes and Support vectorMachines) are more accurate than general-purpose sentimentanalysis techniques in detecting general and specific emotionsexpressed in software-relevant Tweets.