Deep Learning-Based Code Auto-Completion for Distributed Applications


Distributed computing has been gaining a continually increasing interest over the past years in research and industrial communities. One of the significant objectives of distributed computing is to provide the infrastructure for performing tasks on independent systems. Utilizing this approach in software development can reduce costs. Consequently, there has been an increasing interest in distributed applications. However, distributed applications need to meet main features, such as scalability, availability, and compatibility. In this context, service-based systems provide an architecture that can support mentioned features. Nevertheless, current services use various technologies and languages, which bring complexity to development. This work aims to facilitate web service development by introducing a deep Learning-based code auto-complete model. This model is used in the toolkit called SmartCLIDE, which provides features to accelerate development using Artificial Intelligence and cloud deployment. The contribution of this work can fall into two steps: First, the top web APIs from a benchmark web service data-set has been identified. Afterward, a data optimization approach has been proposed to systematically augment and improve available web service codes. Second, the service code auto-completion model has been trained, which takes advantage of text generation trends and deep learning methods. The experimental results on web service codes demonstrate that the proposed approach outperforms another general-purpose code-completion model.



Comments are closed

Recent Tweets