The project investigates early modern Greek and German language printed publications from the 18th century, particularly those connected to the cultural and intellectual networks of Southeastern Europe and the monastic communities of Mount Athos. By bringing together Greek and German language newspapers produced in Vienna and Greek printed books from the 16th to 18th centuries, it examines the gradual shift from the polyvariant medieval manuscript tradition to standardized printed forms. This transition not only transformed textual production and reading practices but also reshaped how knowledge circulated across linguistic, regional and cultural boundaries.

The central goal of the project is to build a digital research environment that enhances access to these historical materials while preserving the interpretive richness of small, specialized datasets. The corpus, which consists of rare books, newspapers and Athonite print culture, represents a body of material whose value lies precisely in its specificity. Rather than relying on large, uniform datasets, the project foregrounds the careful, contextual handling of highly heterogeneous sources, where orthographic variation, unique typographic features and complex historical scripts require close scholarly attention and tailored computational approaches. Here, digitization becomes not merely a technical process but a collaborative interpretive practice, engaging both human expertise and machine assistance (Terras et al. 2013).

To support this vision, the project integrates automated transcription technologies with expert guided corrections. Transkribus is used for handwritten text recognition (Kahle et al. 2017), while Kraken enables the training of custom HTR models designed specifically for historical Greek and German typefaces (Kiessling 2019). These tools allow the system to accommodate the distinctive materiality of early modern print: ligatures, idiosyncratic punctuation, varied orthographic regimes and the interplay of multiple languages on the same page. The resulting transcriptions undergo further processing through a custom NLP pipeline capable of lemma-based search and contextual retrieval of inflectional patterns. Models from the Classical Language Toolkit facilitate the analysis of historical Greek (Johnson et al. 2021), while adapted spaCy tools support the normalization and lemmatization of German texts from the 18th century, capturing linguistic forms that differ significantly from contemporary usage. The project also incorporates geographic analysis as a means of reconnecting texts with the spaces that shaped them. A dedicated geocoding module, enhanced through large language models such as LLaMA, Mistral and Kimi K2 (Touvron et al. 2025; Jiang et al. 2023; Mistral AI 2023), identifies place references and links them to openly accessible knowledge bases including GeoNames, OpenStreetMap and Pleiades. This spatial dimension fosters engagement with the material as part of a broader historical landscape, enabling scholars to trace networks of circulation, influence and mobility across regions and empires.

All processed texts, images and metadata, catalogue information, rights details, and derived annotations, are stored in MongoDB, a non-relational database chosen for its scalability and suitability for complex, multimodal data. The project emphasizes openness, with the infrastructure designed to support long-term accessibility and transparent, reusable workflows. Through this, it aims to foster a collaborative research ecosystem in which scholars working on early modern Southeast European history, philology and book culture can interact with newly digitized materials, share insights and contribute to expanding the dataset.

Supporting this broader scholarly effort is the Scriptura Europae Flask-based application, a multimodal platform designed specifically for historical document processing. Built with Flask, it provides a structured workflow for image upload, automated transcription, linguistic analysis and data visualization. The frontend, implemented in Tailwind CSS, offers a responsive interface that foregrounds clarity and usability while enabling real-time interaction with the processing pipeline. Authentication and role-based permissions ensure secure handling of sensitive archival material, while detailed activity logging preserves transparency in all interactions.

The Scriptura Europae application integrates multiple recognition strategies. Kraken HTR, supported by more than fifteen specialized historical models covering Latin, German, French, Greek and Czech scripts, serves as the primary engine for processing manuscripts and complex early printed materials. Tesseract provides an alternative route for more modern or clearly printed documents (Smith 2007). AI-based enhancement, using the Groq API, refines raw transcriptions by expanding abbreviations, resolving ambiguous characters and improving readability while maintaining historical integrity. This feature illustrates how human judgment and machine assistance can work in tandem: normalization is guided by contextual prompts and metadata, and users are encouraged to inspect and evaluate the results.

For linguistic analysis, the platform employs multilingual spaCy models and custom normalizers for medieval Latin, Greek, historical German and French. These tools support tokenization, morphological analysis, named-entity recognition and lemma extraction (Honnibal et al. 2020). By generating inflectional variants and handling historical spelling conventions, the system facilitates cross-textual queries that account for linguistic complexity rather than flattening it. Texts processed in this way can be mined for patterns, but they also remain open to interpretation, inviting scholars to consider ambiguity, variant forms and the material traces of scribal or typographic practice.

A geolocation module further enriches analysis by identifying place names within texts and converting them into coordinate data linked to interactive maps. This integration opens pathways for exploring spatial relationships within small, historically significant corpora and it is designed to encourage researchers to examine movement, territory and geopolitical context embedded in early modern media.

The platform’s architecture places strong emphasis on secure, isolated file management, user accountability and transparency. Every upload, download and processing step is recorded, allowing for reproducible research and clear provenance of all data transformations. The system is designed to degrade gracefully: when certain AI models or specialized tools are unavailable, alternative workflows ensure continuity without compromising user experience.

Together, the research project and the Scriptura Europae platform constitute an integrated effort to re-engage with early modern print culture through contemporary digital methods. They bring multilingual historical sources into dialogue with advanced computational tools while recognizing the interpretive, ethical and cultural stakes of such work. By combining specialized datasets, expert knowledge and AI-driven processing, the project offers a model for how digital humanities can both preserve and revitalize complex cultural archives, opening them to new forms of inquiry and participation