EmbSE (original) (raw)

Proceedings of the XXXIII Brazilian Symposium on Software Engineering, 2019

Abstract

The representation of contexts is essential in tasks involving Natural Language Processing (NLP). In the field of software engineering, classifying similar texts within a specific context has been a complex task, considering the informality and the complexity inherent of the texts produced through many software development processes (e.g. agile methods). Word embeddings capture semantic and syntactic information about unique words, allowing them to be represented in a dense and low-dimensional format. This property makes the embeddings vectors an important input feature for machine learning algorithms that aim to classify texts. Although there has been much research around the application of word embeddings in several areas, up to this moment, there is no knowledge about studies that have explored its application in the creation of a specific model for the domain of the area of software engineering. Thus, this article presents the proposal to generate an embedding model, called embeddings model for software engineering (EmbSE), which can recognize specific and relevant terms in the software engineering context. This model can be used as the main entry in the classification of several textual artifacts generated during the software development project process. The results are promising, presenting a 48% improvement in the mAP values for the EmbSE concerning the model trained on the generic corpus. This reinforces the hypothesis that a model of this nature can bring significant improvements in the classification of texts of the area.

Andrey Pimentel hasn't uploaded this paper.

Let Andrey know you want this paper to be uploaded.

Ask for this paper to be uploaded.