Loading...
Publication
An approach in software requirements specification with generative AI and RSL framework
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.58 MB | Adobe PDF |
Abstract(s)
Este estudo explora o impacto da Inteligência Artificial (IA) na resolução das ambiguidades e ineficiências inerentes ao processamento da língua natural no domínio da especificação de requisitos de software (SRS). Ao aliar as capacidades da linguagem ITLingo RSL e integrá-las com ferramentas de inteligência artificial, a pesquisa visa otimizar e aprimorar o processo de extração de requisitos de engenharia de software que sejam precisos, a partir de entradas em língua natural. A abordagem propõe uma exploração na engenharia de requisitos, utilizando tecnologias de IA generativa para colmatar a lacuna entre os requisitos informais em língua natural e as suas especificações rigorosas em RSL, e vice-versa, de RSL para a língua natural. Esta abordagem visa melhorar significativamente a precisão das práticas de engenharia de software, posicionando a IA como uma ferramenta complementar nas metodologias de engenharia de requisitos. Através do desenvolvimento e avaliação de um artefacto impulsionado por IA, este estudo pretende demonstrar a viabilidade e os benefícios da automatização da geração de SRS, oferecendo uma visão sobre o futuro dos processos de desenvolvimento de software assistidos por IA. A avaliação do chatbot "GenAI4RSL" demonstrou o seu potencial para melhorar o processo de SRS através da automação da geração de requisitos estruturados. Os utilizadores consideraram o sistema intuitivo, embora tenham sido identificados alguns desafios, em particular na gestão de requisitos mais complexos. A precisão no processamento de língua natural permanece uma área a ser aprimorada. Apesar destas limitações, os participantes valorizaram a capacidade da ferramenta em reduzir o tempo despendido em tarefas repetitivas.
Em conclusão, embora o "GenAI4RSL" mostre potencial para otimizar a engenharia de requisitos, são necessárias melhorias adicionais para aumentar a precisão e adaptá-lo às necessidades específicas de diferentes perfis de utilizadores. Estas melhorias serão essenciais para uma aplicação mais ampla no campo do desenvolvimento de software.
This study explores the impact of Artificial Intelligence (AI) in addressing the inherent ambiguities and inefficiencies of natural language processing within the domain of software requirements specification (SRS). By coupling the capabilities of the ITLingo RSL language and integrating them with artificial intelligence tools, the research aims to streamline and enhance the process of deriving accurate software engineering requirements from natural language inputs. The approach proposes an exploration in requirements engineering, resourcing to generative AI technologies to bridge the gap between informal natural language (NL) requirements and their rigorous specifications in RSL and back to natural language from RSL specifications. This approach aims to improve the accuracy of software engineering practices and position AI as a complementary tool in requirements engineering. Through the development and evaluation of an AI-driven artifact, the “GenAI4RSL” chatbot, this study aims to demonstrate the feasibility and benefits of automating SRS generation, providing insights into the future of AI-assisted software development processes. The evaluation of the "GenAI4RSL" chatbot demonstrated its potential to improve the SRS process by automating the generation of structured requirements. Users found the system intuitive, though some challenges were identified, particularly in handling more complex requirements. Accuracy in natural language processing remains an area for improvement. Despite these limitations, participants appreciated the tool's ability to reduce the time spent on repetitive tasks. In conclusion, while the GenAI4RSL chatbot shows promise in streamlining requirements engineering, further refinements are needed to improve accuracy and adapt it to the specific needs of different user roles. These improvements will be essential for broader applications in the software development field.
This study explores the impact of Artificial Intelligence (AI) in addressing the inherent ambiguities and inefficiencies of natural language processing within the domain of software requirements specification (SRS). By coupling the capabilities of the ITLingo RSL language and integrating them with artificial intelligence tools, the research aims to streamline and enhance the process of deriving accurate software engineering requirements from natural language inputs. The approach proposes an exploration in requirements engineering, resourcing to generative AI technologies to bridge the gap between informal natural language (NL) requirements and their rigorous specifications in RSL and back to natural language from RSL specifications. This approach aims to improve the accuracy of software engineering practices and position AI as a complementary tool in requirements engineering. Through the development and evaluation of an AI-driven artifact, the “GenAI4RSL” chatbot, this study aims to demonstrate the feasibility and benefits of automating SRS generation, providing insights into the future of AI-assisted software development processes. The evaluation of the "GenAI4RSL" chatbot demonstrated its potential to improve the SRS process by automating the generation of structured requirements. Users found the system intuitive, though some challenges were identified, particularly in handling more complex requirements. Accuracy in natural language processing remains an area for improvement. Despite these limitations, participants appreciated the tool's ability to reduce the time spent on repetitive tasks. In conclusion, while the GenAI4RSL chatbot shows promise in streamlining requirements engineering, further refinements are needed to improve accuracy and adapt it to the specific needs of different user roles. These improvements will be essential for broader applications in the software development field.
Description
Tese de Mestrado em Informação e Sistemas Empresariais, apresentada à Universidade Aberta
Keywords
Especificação de requisitos de software RSL (Linguagem de Especificação de Requisitos) ITLingo Linguagem natural Inteligência artificial generativa Chatbot Software requirements specification RSL (Requirements Specification Language) Natural language Generative artificial intelligence
Pedagogical Context
Citation
Publisher
CC License
Without CC licence