Master's Thesis

A System for OCR as a Service

Guilherme Menezes Garrido Branco2025

Key information

Authors:

Guilherme Menezes Garrido Branco (Guilherme Menezes Garrido Branco)

Supervisors:

José Luís Brinquete Borbinha (José Luís Brinquete Borbinha); Flávio Nuno Fernandes Martins (Flávio Martins)

Published in

November 11, 2025

Abstract

This dissertation presents the redesign and enhancement of an open-source system providing Optical Character Recognition (OCR) as a service. The work aimed to improve the system’s flexibility, performance, and usability, to support large-scale and configurable document processing while remaining accessible to non-expert users. The improved version introduces extends input and output format support, notably accepting ZIP archives, multi-page TIFFs, and producing specialized formats such as hOCR and ALTO. A new configuration interface allows users to define, store, and reuse OCR settings at both document and folder levels, enabling batch processing and reproducible workflows. Architectural refinements enhance scalability and stability, while interface updates, including context menus, document thumbnails, and confidence-based result highlighting, contribute to a clearer and more efficient user experience. The system was evaluated through usability, performance, and accuracy assessments. In usability testing, eight participants successfully completed all task scenarios and reported an overall System Usability Scale (SUS) score of 86.25, reflecting a high level of satisfaction and ease of use. Performance testing demonstrated improved memory efficiency, enabling longer uninterrupted operation. Accuracy evaluation, conducted with digitized Portuguese cultural heritage documents, confirmed the system’s flexibility and robustness: the combination of Sauvola binarization and manual segmentation achieved an average Word Error Rate (WER) of 34%, outperforming comparable online services in search-oriented recognition tasks. The result is an open-source, configurable, and scalable OCR-as-a-Service system that effectively combines technical adaptability with user-centred design, providing a solid foundation for future research and distributed OCR processing.

Publication details

Authors in the community:

Supervisors of this institution:

Fields of Science and Technology (FOS)

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

Publication language (ISO code)

eng - English

Rights type:

Embargoed access

Date available:

October 7, 2026

Institution name

Instituto Superior Técnico