SEHGN: Semantic-Enhanced Heterogeneous Graph Network for Web API Recommendation

With the growth of cloud computing, a large number of innovative mashup applications and Web APIs have emerged on the Internet. The expansion of technology and information presents a significant challenge to the discovery of Web APIs from multiple service ecosystems. Various Web API recommendation m...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on services computing Vol. 17; no. 5; pp. 2836 - 2849
Main Authors: Wang, Xuanye, Xi, Meng, Li, Ying, Pan, Xiaohua, Wu, Yangyang, Deng, Shuiguang, Yin, Jianwei
Format: Journal Article
Language:English
Published: IEEE 01.09.2024
Subjects:
Application programming interfaces
Biological system modeling
Computational modeling
Ecosystems
heterogeneous network
mashup creation
Mashups
multi-semantic enhance
semantic embedding
Semantics
Vectors
Web API recommendation
ISSN:1939-1374, 2372-0204
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the growth of cloud computing, a large number of innovative mashup applications and Web APIs have emerged on the Internet. The expansion of technology and information presents a significant challenge to the discovery of Web APIs from multiple service ecosystems. Various Web API recommendation methods have been proposed for Mashup creation, but most either treat different feature factor interactions equally or solely rely on requirements for API recommendation. These approaches face several challenges such as API compatibility dependencies, ambiguous definition and boundary dilemmas of APIs, and sparse API invocation records. In this work, we propose a Semantic-Enhanced Heterogeneous Graph Network(SEHGN) for Mashup creation. To address the above deficiencies, we design a multi-semantic aggregator to capture semantic associations between features to encode multiple node-edge relationships. Then, we introduce a semantic embedding component to generate text embedding vectors for mashups and APIs to learn global and local semantic information about text documents at different levels of abstraction. Finally, we fuse the output vectors to obtain a list of candidate Web APIs. Experiences are performed on real datasets, and statistical results show that SEHGN outperforms state-of-the-art models in terms of overall and long-tail Web API recommendations.
ISSN:1939-1374
2372-0204
DOI:10.1109/TSC.2024.3417323