Intelligent classification of SPI practices and evidences based on NLP and semantic similarity

• Ano de defesa: 2021
• Autor(a) principal: Ecar, Miguel da Silva
• Orientador(a): Silva, João Pablo Silva da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal do Pampa
• Programa de Pós-Graduação: Mestrado Profissional em Engenharia de Software
• Departamento: Campus Alegrete
• País: Brasil
• Palavras-chave em Português: Engenharia de software; Software - desempenho; Software Engineering; Software - performance
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA
• Link de acesso: https://repositorio.unipampa.edu.br/jspui/handle/riu/6753

Software Process Improvement (SPI) consists of a set of changes in software development companies, which introduces new and improved methods, techniques, and tools. SPI initiatives generally are performed based on a reference model, such as CMMI, ISO 9001, ISO 15504, among others. One of the first steps when investing in SPI initiatives is the SPI Diagnostic. The SPI Diagnostic is generally performed manually, which demands high effort from consultants. Moreover, a high data volume is generated and must be analyzed, which is bound to subjective analysis. Since there has been a lack of automation tools to support this process, it turns SPI Diagnostic a challenging process. This work aims to propose an intelligent tool called Coptic for Practice-Evidence classification, using Natural Language Processing and Semantic Similarity. We also propose Base of Knowledge about Software Engineering Practices (Badge), which is a domain ontology that generalizes SPI resources that says “what should be done” and PStory, which is a template to write pieces of evidence. We evaluated Badge through a focus group session. We evaluated PStory through an exercise and questionnaire with industry professionals. We evaluated Coptic by a quasi-experiment with PStories evaluated by industry professionals. As an outcome, Coptic presented satisfactory results using the initial corpus. We conclude that Coptic presents a valuable result in terms of providing support to professionals in performing a SPI Diagnostic. Badge introduces a domain ontology that differs from the related proposals in literature and has value to SPI initiatives. We also concluded that PStory introduces a simple way to write pieces of evidence, and Coptic provides support to SPI Practices-Evidence matching process. Key-words: Coptic, Badge, PStory, Software Process Improvement, SPI.