Abstract
One of the needs in adopting a crowdsourcing approach in software requirement system (SRS) is to be able to perform text analytics to gain insight or knowledge from the crowd's feedback. One of the expected text analytic tasks is to be able of analyzing the feedback automatically; such as, whether the feedback concerns about the functional requirement (FR) and non-functional requirements (NFR). To automatically do the FR and NFR identification, one can treat the problem as a text classification task. Performing automatic identification requires features (word representation) that contain sufficient information which can be used to do the identification. Thus, the types of features can be considered as an important step in performing automatic identification of FR and NFR. In this study, the RE'17 dataset challenge is used as the dataset. Using the dataset, we will like to find out the effect of word embedding against traditional features (such as bag-of-words) in NFR and FR classification. In addition, we also want to find out whether is necessary to use a complex neural classifier to obtain the best performance of NFR and FR classification. Based on the obtained results, using fastText seems to be the promising classification model since the model obtained the highest Fl-score of 92.8%.
Export citation and abstract BibTeX RIS
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.