Objectives and Results

Technologies that enable machines to imitate cognitive processes, recognize patterns, and understand human languages have become an integral part of many areas of society. Artificial Intelligence (AI) has also found its way into the digital humanities (DH), influencing how objects are studied, data is collected, and relations are presented. Especially in the editing and analysing of historical sources, which often require complex comparative and contextualising work, AI has the potential to significantly streamline and accelerate research.

Yet, the development of AI models capable of reliably assisting in the understanding and editing of historical texts in bygone languages remains at an early stage. Here, research and development encounter the fundamental problem that large AI models are trained using languages for which extensive digital source material is available and continually produced. Languages with small or extinct speaker communities (low-resource languages, or LRLs) leave fewer digital traces and are frequently neglected in AI development. Historical texts present additional difficulties due to their lack of grammatical and orthographic standardisation.

This gap is addressed by EdiKILex. Over the coming years, the team will use Early New High German (c. 1350–1650) as a case study to develop an AI that can be used effectively by both non-specialists and specialists. Early New High German is a particularly suitable LRL for this pioneering research, because an increasing number of digital editions of texts in this language provide training material. Most importantly, the Early New High German Dictionary (FWB) offers a comprehensive basis for lexicographical commentary.

EdiKILex pursues four core objectives:

1. The Open-Source Software (sustainably hosted by the HAB Wolfenbüttel) to facilitate the reading and analysis of digitally edited texts in Early New High German. It will offer a digital environment where tailored cultural-historical, pragmatic, and semantic explanations for each word can be generated. The digital FWB and selected editions are being used to develop a comprehensive text-knowledge-system that will enable the application to provide effective assistance to users with different interests (Hybrid Intelligence).
    
2. The AI’s expansion to include an intuitive interface that can also be used as a tool for producing new digital editions in the future, with EdiKILex assisting in the time-consuming task of identifying and annotating Late Medieval and Early Modern German word forms.
    
3. The prospective transfer of the model and methods to texts in other LRLs, such as Old French or Celtic languages, based on insights into problems and solutions gained during the development of EdiKILex.
    
4. A theoretical contribution to the broader discussion on the opportunities and risks of AI and Hybrid Intelligence in academic and socio-technological contexts, not least with a focus on sustainability.

EdiKILex will be user-focused from the outset. Needs assessment and usability testing in both research and teaching are therefore integral to the development process. EdiKILex will thus create not only a versatile tool for exploring early modern languages and knowledge cultures, but also a platform for interdisciplinary dialogue on the future of the Humanities and broader science communication.

The tools developed, as well as results of research, will be made available here. In the meantime, news about the project will be published under News and Events and Publications and Press.