A Comparative Study of Name Entity Recognition Techniques in Software Engineering Texts

Abstract

Named Entity Recognition (NER) is an essential sub-task for many important tasks in software engineering (SE), such as automated requirement analysis, opinion mining, question answering, knowledge base construction, and information retrieval. As existing domain independent approaches perform low when applied in the SE domain, recently we observe the development of different tools and techniques for NER in this domain. Despite those developments, we lack our understanding of the effectiveness of the recent deep learning based state-of-the-art models and existing popular machine learning based tools for NER in SE texts. We quantitatively evaluate the performances of seven versions of machine learning and deep learning-based tools and techniques in detecting name entities from SE texts. The evaluation results will advance our understanding in devising an improved technique for NER in the SE domain.