Modular AI for the Digital Humanities: Applications of MPAI Standards to DH Research Workflows

As Artificial Intelligence technologies, such as Large Language Models, become increasingly widespread in the domain of Social Science and Humanities, the lack of a common approach to managing complex AI-based workflows becomes a critical barrier to scalability, reproducibility, and interoperability. This paper explores how the standardization technologies developed by Moving Picture, Audio and Data Coding by Artificial Intelligence (MPAI) can address this gap and provide modular, implementation-agnostic mechanisms suited to describing and streamlining research workflows in the domain of Digital Humanities. DARIAH.it, the Italian node of Digital Research Infrastructure for Arts and Humanities, strategically supports this standardization initiative and will provide two use cases, derived from the Digital Philology Hub and the MetaFAIR Ecosystem, respectively. A third use case originates from the 'LiLa - Linking Latin' project, which aims to provide interoperability between lexicons, texts, and Natural Language Processing tools for Latin by adopting the Linked Data paradigm. By adopting interfaces and workflows defined in terms of the MPAI standardization procedure, Social Science and Humanities researchers can construct modular, scalable, and interoperable research environments that promote long-term data reusability and methodological transparency.