The article proposes a conceptual framework for achieving automated software evolution (ASEv) for context-rich intelligent applications. The framework consists of five key dimensions:
Context Sources (C): This dimension encompasses various data sources from the software engineering lifecycle, intelligent applications, and their environments. The underlying idea is that the more relevant contextual details are collected, the better decisions can be made in generating the final software products.
Data Modalities (D): This dimension covers the diverse data formats, including text, images, videos, audio, and sensor data, that can be leveraged for multimodal learning.
Multimodal Learning (M): This dimension explores different techniques for multimodal learning, including white-box, black-box, and gray-box models, as well as data preprocessing and fusion methods.
Key Features of ASEv (K): This dimension identifies the essential components of the learning process, such as dynamic code generation, system integration, automated bug detection and correction, and continuous integration and deployment.
Products (P): This dimension focuses on the final outcomes of the ASEv process, which can include new or updated features, programs, applications, and system configurations.
The article also introduces a Selective Sequential Scope Model (3S model) to categorize existing and future research on ASEv based on the coverage of software engineering phases and multimodal learning tasks. The framework and 3S model are then applied to analyze several ASEv-related studies, demonstrating their utility in understanding the current state of the research and identifying potential areas for future exploration.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Songhui Yue at arxiv.org 04-09-2024
https://arxiv.org/pdf/2404.04821.pdfDeeper Inquiries