JSON-LD Based Web API Semantic Annotation Considering Distributed Knowledge

Based on semantically annotated Web APIs, automatic Web API composition can be implemented easily. The operation can greatly improve efficiency of building a software system. However, in real world, semantic annotation for Web APIs will encounter various difficulties, because of their distribution and function diversity, such as disunited API description formats, response result with complex structure, shortage of business domain ontologies, semantic conflicts among distributed knowledge, and so on. To solve these difficulties, we propose a JSON-LD based Web API semantic annotation approach (JWASA). JWASA can assist professional developers to semi-automatically complete semantic annotation of Web APIs. A common Web API description ontology is firstly defined, including necessary vocabularies about invocation information, functional semantics, and non-functional features. Then, JWASA automatically converts a Web API description into a document in an united JSON format, and assist developers to semi-automatically embed semantic information of crucial elements of the API by means of a lightweight Linked Data format JSON-LD. Meanwhile, a semantic annotation specification is proposed to deal with various complex situations in Web API description, e.g: too many response parameters, no request parameters, etc. In addition, to improve efficiency of annotation, JWASA provides some extra operations, including automatic new ontology or vocabulary creation, automatic functional semantics extraction etc. Also, JWASA provides semi-automatically bridge rule generation algorithm, which can infer implied relationships among vocabularies (e.g: sub-class, super-class, equality). JWASA focuses on the semantic annotation of functionality of Web APIs, and can create effective semantic Web APIs for future API automatic composition. We implement a prototype system and carry out a series of experiments to evaluate JWASA on real Web APIs crawled from Internet. Experiments show that our approach is effective and efficient.