Making a Web service from OCR SDK

This article summarizes authors’ experience of developing a web service (WS) based on a document optical character recognition (OCR) software development kit (SDK). We consider issues of WS stability and performance, including: ability not to lose data under high load and after restart; ability to timely detect errors and limit their spread and duration; as well as deterministic WS behavior under conditions of parallel processing of multiple requests. High WS performance implies moderate overhead costs associated with receiving web requests and sending web responses to clients, besides of OCR engine costs itself. The described solution can be used to create a web service from any SDK which enables developer to process input documents and obtain output files from them, not necessarily in connection with optical recognition technologies.