Microservice-Based Data Management and Processing: A Modular Workflow for Automatic Text Recognition and Beyond

The FLOW project develops modular, microservice-based workflows for machine learning-driven data processing in the Digital Humanities. By separating key tasks like preprocessing, training, inference, and evaluation into independent components, it provides scalable solutions for automatic text recognition (ATR). Centered around GitHub and its CI/CD capabilities, these microservices support streamlined processes tailored to diverse research needs. Researchers can configure the services and trigger automated pipelines via GitHub Issues, without requiring any coding knowledge. The project aims to implement a real-world text recognition pipeline and demonstrate the architecture’s potential to lower technical barriers while supporting open science practices.