Speaker: Deepak Rishi, Master's Candidate
Sentiment and emotional analysis on online collaborative software development forums can be very useful to gain important insights into the behaviours and personalities of the developers. Such information can later on be used to increase productivity of developers by making recommendations on how to behave best in order to get a task accomplished. However, due to the highly technical nature of the data present in online collaborative software development forums, mining sentiments and emotions becomes a very challenging task.
In this work we present a new approach for mining sentiments and emotions from software development datasets using Interaction Process Analysis (IPA) labels and machine learning. We also show that limited training data, applying distance metric learning as a preprocessing step, can lead to gains in performance metrics such a precision, recall and F1 of deep neural networks.