Turkish issue report classification in banking domain [Bankacılık alanında Türkçe yazılım hata raporu sınıflandırması]

Official URL: https://dx.doi.org/10.1109/SIU49456.2020.9302232

Users report the problems they encounter while using a software product with software issue reports. It is important that they are assigned to the correct software team or developer so that they are resolved quickly. Incorrect assignment may increase solution times, thus causing customer dissatisfaction. Past studies suggest to use text classification techniques to automatically assign issue reports. In this study, software issue reports written in Turkish, obtained from an industrial case in the banking sector are classified by applying deep learning techniques on word embedding representation, and the results are compared with our baseline model, which is applying Support Vector Machines (SVM) on top of the bag of words (BOW) model. In our study, best results are obtained when words are presented with BOW model and classes are predicted with the SVM algorithm.