Towards effective semantic annotation for mobile and edge services for Internet-of-Things ecosystems

As a popular and typical representative of services, Web APIs have been widely used to provide mobile and edge services in Internet-of-Things (IoT) ecosystems. As a combination of Web APIs, mashups are also popularly used to provide more complicated services in IoT ecosystems. Tags are one of the core attributes of Web APIs, and semantic annotation or automatic tagging takes a core role in understanding and management of the large volume of services. But currently, semantic annotation heavily relies on manual work and costs a lot of time and human resources. There have been some annotation methods for services, but the existing methods suffer from low accuracy and other problems. In this paper, we propose a neural semantic annotation framework for automatic tagging for mobile and edge services in IoT ecosystems. We propose to accomplish semantic annotation as the multi-label classification problem, and develop an automatic annotation framework based on multi-head self-attention mechanism. We crawled two large real-world datasets and gave a deep analysis of the distributions of annotations. We performed sufficient experiments on the two datasets and the experimental results demonstrate that our framework yields superior performance in semantic annotation for mobile and edge services.