Automated Information Extraction from Web APIs Documentation

A fundamental characteristic of Web APIs is the fact that, de facto, providers hardly follow any standard practices while implementing, publishing, and documenting their APIs. As a consequence, the discovery and use of these services by third parties is significantly hampered. In order to achieve further automation while exploiting Web APIs we present an approach for automatically extracting relevant technical information from the Web pages documenting them. In particular we have devised two algorithms that automatically extract technical details such as operation names, operation descriptions or URI templates from the documentation of Web APIs adopting either RPC or RESTful interfaces. The algorithms devised, which exploit advanced DOM processing as well as state of the art Information Extraction and Natural Language Processing techniques, have been evaluated against a detailed dataset exhibiting a high precision and recall---around 90% for both REST and RPC APIs---outperforming state of the art information extraction algorithms.