Scientific workflow systems are an established means to model and execute experiments or processing pipelines. Nevertheless, designing workflows can be a daunting task for users due to the complexities of the systems and the sheer number of available processing nodes, each having different compatibility/applicability characteristics. This Thesis explores how concepts of the Semantic Web can be used to augment workflow systems in order to assist researchers as well as non-expert users in creating valid and effective workflows. A prototype workflow creation/management system has been developed, including components for ontology modelling, workflow composition, and workflow repositories. Semantics are incorporated as a lightweight layer, permeating all aspects of the system and workflows, including retrieval, composition, and validation. Document image analysis and recognition is used as a representative application domain to evaluate the validity of the system. A new semantic model is proposed, covering a wide range of aspects of the target domain and adjacent fields. Real-world use cases demonstrate the assistive features and the automated workflow creation. On that basis, the prototype workflow creation/management system is compared to other state-of-the-art workflow systems and it is shown how those could benefit from the semantic model. The Thesis concludes with a discussion on how a complete infrastructure based on semantics-enriched datasets, workflow systems, and sharing platforms could represent the next step in automation within document image analysis and other domains.