Details for: Sarschar M. Pipeline for Automated Code Generation from Backlog Items...2025

Sarschar M. Pipeline for Automated Code Generation from Backlog Items...2025

Type:

Other > E-books

Files:

1

Size:

13.61 MiB (14275363 Bytes)

Uploaded:

2025-02-06

By:

andryold1  

Seeders:

6

Leechers:

6

Info Hash:

D6B1E564B079BC47DE1525D604727F65B43095B6

Get This Torrent  PLAY/STREAM Movies/TvShows  Download Anonymously

Textbook in PDF format This book investigates the potential and limitations of using Generative AI (GenAI) in terms of quality and capability in agile web development projects using React. For this purpose, the Pipeline for Automated Code Generation from Backlog Items (PACGBI) was implemented and used in a case study to analyse the AI-generated code with a mix-method approach. The findings demonstrated the ability of GenAI to rapidly generate syntactically correct and functional code with Zero-Shot prompting. The PACGBI showcases the potential for GenAI to automate the development process, especially for tasks with low complexity. However, this research also identified challenges with code formatting, maintainability, and user interface implementation, attributed to the lack of detailed functional descriptions of the task and the appearance of hallucinations. Despite these limitations, the book underscores the significant potential of GenAI to accelerate the software development process and highlights the need for a hybrid approach that combines GenAI's strengths with human expertise for complex tasks. Further, the findings provide valuable insights for practitioners considering GenAI integration into their development processes and set a foundation for future research in this field. Large Language Models (LLM), are models that are trained on a large amount of data to understand and generate human language. Their large size enables them to do a wide range of tasks, like text generation and translation, but also image and code generation. Code generation, also known as program synthesis, is the task where “a natural language description of a code snippet is fed as input, and the model is expected to generate the corresponding code snippet as output”. This natural language input is called prompt. The prompt is then broken down into small chunks that vary in size by a process of transformer-based LLMs called ‘tokenization’. The result is the initial prompt’s information divided into so-called ‘tokens’. GenAI is already used in multiple different tools to enhance the coding experience. There are IDE-Tools like GitHub Copilot, Amazon CodeWhisperer or TabNine. These are installed as a plugin in source-code editors and deliver code completion suggestions based on the current code file within the context of the repository code. For web development, there are tools that can specifically generate user interfaces (UI). Anima for example transfers mockups into React code, while v0 by Vercel turns natural language into React components via Few-Shot prompting. The AI Website Builder by Teleport HQ aims to build static websites from wireframes and prompts. Finally, two tools are to be published within 2024 that take a more holistic approach on the GenAI usage in the software development life cycle. The first tool is GitHub Copilot Workspace, which is scheduled for release in 2024. It is currently a research prototype developed with the aim of assisting developers in implementing tickets of GitHub repositories. There are no official statements, but it is likely that it uses Codex as LLM, similar to GitHub Copilot itself. The second tool is Codegen. There is currently limited information on this tool, but similar to Copilot Workspace, its aim is to automatically solve issues from ITS like Jira, Linear or GitHub Issues. It states to use GPT-4 as a LLM. The results regarding the quality and capability of AI-generated code, along with their practical implications, aim to aid practitioners in the decision-making process of adapting GenAI into their software development life cycle and further provide reference points for future research. Introduction Theoretical Background Related Work Method Implementation Results Discussion Summary References

Get This Torrent  PLAY/STREAM Movies/TvShows  Download Anonymously

1. Sarschar M. Pipeline for Automated Code Generation from Backlog Items...2025.pdf13.61 MiB