TY - JOUR

T1 - Sentiment analysis tools in software engineering

T2 - A systematic mapping study

AU - Obaidi, Martin

AU - Nagel, Lukas

AU - Specht, Alexander

AU - Klünder, Jil

N1 - Funding Information: This work was supported by the Leibniz Young Investigator Grant, Germany and is part of the project ComContA, 2020–2022.

PY - 2022/11

Y1 - 2022/11

N2 - Context: Software development is a collaborative task. Previous research has shown social aspects within development teams to be highly relevant for the success of software projects. A team’s mood has been proven to be particularly important. It is paramount for project managers to be aware of negative moods within their teams, as such awareness enables them to intervene. Sentiment analysis tools offer a way to determine the mood of a team based on textual communication. Objective: We aim to help developers or stakeholders in their choice of sentiment analysis tools for their specific purpose. Therefore, we conducted a systematic mapping study (SMS). Methods: We present the results of our SMS of sentiment analysis tools developed for or applied in the context of software engineering (SE). Our results summarize insights from 106 papers with respect to (1) the application domain, (2) the purpose, (3) the used data sets, (4) the approaches for developing sentiment analysis tools, (5) the usage of already existing tools, and (6) the difficulties researchers face. We analyzed in more detail which tools and approaches perform how in terms of their performance. Results: According to our results, sentiment analysis is frequently applied to open-source software projects, and most approaches are neural networks or support-vector machines. The best performing approach in our analysis is neural networks and the best tool is BERT. Despite the frequent use of sentiment analysis in SE, there are open issues, e.g. regarding the identification of irony or sarcasm, pointing to future research directions. Conclusion: We conducted an SMS to gain an overview of the current state of sentiment analysis in order to help developers or stakeholders in this matter. Our results include interesting findings e.g. on the used tools and their difficulties. We present several suggestions on how to solve these identified problems.

AB - Context: Software development is a collaborative task. Previous research has shown social aspects within development teams to be highly relevant for the success of software projects. A team’s mood has been proven to be particularly important. It is paramount for project managers to be aware of negative moods within their teams, as such awareness enables them to intervene. Sentiment analysis tools offer a way to determine the mood of a team based on textual communication. Objective: We aim to help developers or stakeholders in their choice of sentiment analysis tools for their specific purpose. Therefore, we conducted a systematic mapping study (SMS). Methods: We present the results of our SMS of sentiment analysis tools developed for or applied in the context of software engineering (SE). Our results summarize insights from 106 papers with respect to (1) the application domain, (2) the purpose, (3) the used data sets, (4) the approaches for developing sentiment analysis tools, (5) the usage of already existing tools, and (6) the difficulties researchers face. We analyzed in more detail which tools and approaches perform how in terms of their performance. Results: According to our results, sentiment analysis is frequently applied to open-source software projects, and most approaches are neural networks or support-vector machines. The best performing approach in our analysis is neural networks and the best tool is BERT. Despite the frequent use of sentiment analysis in SE, there are open issues, e.g. regarding the identification of irony or sarcasm, pointing to future research directions. Conclusion: We conducted an SMS to gain an overview of the current state of sentiment analysis in order to help developers or stakeholders in this matter. Our results include interesting findings e.g. on the used tools and their difficulties. We present several suggestions on how to solve these identified problems.

KW - Social software engineering

KW - Sentiment analysis

KW - Machine learning

KW - Systematic mapping study

KW - Systematic literature review

UR - http://www.scopus.com/inward/record.url?scp=85135527315&partnerID=8YFLogxK

U2 - 10.1016/j.infsof.2022.107018

DO - 10.1016/j.infsof.2022.107018

M3 - Article

VL - 151

JO - Information and Software Technology

JF - Information and Software Technology

SN - 0950-5849

M1 - 107018

ER -