TY - JOUR
T1 - User story extraction from natural language for requirements elicitation
T2 - Identify software-related information from online news
AU - Siahaan, Daniel
AU - Raharjana, Indra Kharisma
AU - Fatichah, Chastine
N1 - Publisher Copyright:
© 2023 Elsevier B.V.
PY - 2023/6
Y1 - 2023/6
N2 - Context: The user story is a popular artifact in agile software development. Extracting user stories is helpful for process improvement in requirements elicitation, closing limitations such as limited access, and uncovering new and unique domains. Most sources of requirements elicitation are available in natural language form. However, the approach to extracting user stories from natural language is still limited. Objective: This study aims to extract user stories from natural language. It includes identifying the aspect of who (stakeholder), aspect of what (stakeholder's wants), and aspect of why (the reason why the aspect of what exists). Method: This study used online news as a case study because information related to stakeholders and their needs is available. Aspects of who, what, and why are obtained using a rule-based approach using part-of-speech (POS) chunking, named entity recognition (NER), dependency parsing, WordNet, and BloomSoft. Result: We found that online news tends to generate requirements with hard-goals or soft-goals types. In identifying aspects of who, we succeeded in increasing the F-score value by combining stakeholder identification methods according to the characteristics of online news. We also found that PUblic REquirements (PURE), domain specificity, and WordNet lexical names can significantly improve the extraction of software-related information in identifying the aspects of what. Conclusion: This study demonstrates that information related to software requirements could arise from non-software-related artifacts such as online news.
AB - Context: The user story is a popular artifact in agile software development. Extracting user stories is helpful for process improvement in requirements elicitation, closing limitations such as limited access, and uncovering new and unique domains. Most sources of requirements elicitation are available in natural language form. However, the approach to extracting user stories from natural language is still limited. Objective: This study aims to extract user stories from natural language. It includes identifying the aspect of who (stakeholder), aspect of what (stakeholder's wants), and aspect of why (the reason why the aspect of what exists). Method: This study used online news as a case study because information related to stakeholders and their needs is available. Aspects of who, what, and why are obtained using a rule-based approach using part-of-speech (POS) chunking, named entity recognition (NER), dependency parsing, WordNet, and BloomSoft. Result: We found that online news tends to generate requirements with hard-goals or soft-goals types. In identifying aspects of who, we succeeded in increasing the F-score value by combining stakeholder identification methods according to the characteristics of online news. We also found that PUblic REquirements (PURE), domain specificity, and WordNet lexical names can significantly improve the extraction of software-related information in identifying the aspects of what. Conclusion: This study demonstrates that information related to software requirements could arise from non-software-related artifacts such as online news.
KW - Natural language processing
KW - Process improvement
KW - Requirements elicitation
KW - Software-related information
KW - User story
UR - http://www.scopus.com/inward/record.url?scp=85150020375&partnerID=8YFLogxK
U2 - 10.1016/j.infsof.2023.107195
DO - 10.1016/j.infsof.2023.107195
M3 - Article
AN - SCOPUS:85150020375
SN - 0950-5849
VL - 158
JO - Information and Software Technology
JF - Information and Software Technology
M1 - 107195
ER -