As the call for papers foregrounds, the TEI has been the standard for modeling textual information for over 40 years. This short paper reports on the staticization work of the Yellow Nineties (Y90s), a project that has relied on the TEI for 20 years. The leading resource for the study of this emerging publishing genre, the Y90s site offers digital editions of the complete print runs of eight magazines produced between 1889 and 1905. The site’s critical introductions and peer-reviewed biographies enable users to gain insight into the magazines’ material formats, alternative modes of production, intermedial connections, and transnational networks of contributors, as well as their involvement in contemporary social movements, such as environmentalism, feminism, sexual rights, socialism, and nationalist revivals. A world renowned resource, whose data has been shared across projects for scholarship and teaching (Hughes; Mahoney and Abrams), the Y90s also “demonstrates best practice for digital Victorian site [content] in a manner matched by few peer projects. The site shows that such projects can focus in a rigorous manner on primary content, while also meeting international standards for digital development, preservation, documentation, and transparency” (Wisnicki, 981).

The project, has undergone a number of technological transformations: from plans for ColdFusion, to TEI and HTML underpinned by a MySQL database, and then WordPress. In order to ensure sustainability, the Y90s is taking up the Endings Project’s static site development methodology (Holmes and Takeda). The Endings-informed staticization process has not only given us the opportunity to consider Humanities-appropriate FAIR compliance, but also the occasion to engage best data development practices developed by values aligned organizations including CWRC, Homosuarus, and RTPP, with the goal of having Yellow Nineties 3.0 live up to our billing as a project that meets the “international standards for digital development, preservation, documentation, and transparency.”

This presentation proposes a TEI-compliant structured model for kibyōshi, a genre of classical Japanese books, through the integration of TEI encoding and IIIF image resources.

Kusazōshi were illustrated fiction published in Japan from the mid-seventeenth to the nineteenth century. Among them, kibyōshi emerged as a distinct subgenre in the late eighteenth century. Each work typically consists of a short narrative of approximately 10 to 30 pages. With the exception of the preface and postscript, nearly all pages are illustrated, and the main text and dialogue are written in the blank spaces of the images, creating a close interdependence between visual and textual elements.
Kibyōshi incorporates diverse themes—including theater, folktales, moral instruction, waka poetry, customs, and Buddhism—expressed through both image and text. Despite their limited length, these works contain remarkable thematic variety.
Although the TEI Guidelines provide various methods for encoding relationships between images and text, these approaches have not been sufficiently systematized. As a result, multiple encoding practices coexist, hindering interoperable processing and viewer development.
This presentation reports on a practical implementation of TEI-based structuring for kibyōshi in conjunction with IIIF image resources. Image information was encoded using elements such as <facsimile>, <graphic>, and <surface>, while specific regions were defined with <zone> and linked to textual components via the @facs attribute, enabling bidirectional referencing.
This implementation demonstrates that correspondences between image regions and textual elements can be consistently encoded within the TEI framework. The resulting model offers a reusable approach applicable to other early modern illustrated works in which image and text function in a mutually complementary manner.

This paper will demonstrate some ways that the rich information modeled in TEI documents can be used for text and network analysis. The Women Writers Project (WWP), a long-term research project focused on text encoding and early women’s writing, has recently begun developing methods for using encoded texts in computational analyses. As part of the WWP’s work on making word embedding models more accessible for humanists, we published a set of routines that use XSLT and XQuery to prepare TEI files for computational text analysis. These routines provide a set of fine-tuning options that are targeted to the needs of natural language processing applications. For example, the contents of notes can be placed either at their points of anchor or in a separate section in the plain-text output files; the transformation routines also enable researchers to select and omit any elements whose contents are likely to distort results in trained models. In a current project studying the impacts of women on the development of the sciences, the WWP team has built two new transformation routines, both in XQuery. One expands on the plain-text generation process described above to allow selection of researcher-defined “documents” for topic modeling, based on how texts’ structures are modeled in the markup. The other uses bibliographic markup to extract information on co-citation within texts and produce edge lists for network analysis. This paper will share some insights that the WWP has developed through our efforts to realize the potential of TEI documents for computational analysis. It will also offer some strategies for projects interested in taking up such work with their own data, with approaches for adapting existing resources, making the most of provisional solutions, and iterating between data retrieval and research design.

Digital scholarly editions, while differing in editorial goals, encoding depth, and source material, share many operational steps—digitisation, OCR, validation, transformation, publication, versioning, and archiving. Often, these steps are either done manually and/or implemented in project-specific scripts.

This paper presents a workflow automation tool freeing editors and developers of infrastructure from the necessity to perform these repeating tasks themselves. It combines several well-established tools to create flexible workflows that can easily be adapted to a projects needs while still being rooted on the same operational principles, combining the need to adjust a workflow to a project’s needs with a sustainable and scalable infrastructure.

The workflow integrates OCR and layout recognition, post processing, schema validation, version control, publication, and long-term archiving, allowing for fully automated flows and manual intervention as needed. Workflow orchestration is implemented using Prefect, a lightweight, Python-based framework that enables event-driven and sequential execution of discrete tasks. Each step is explicitly defined, logged, and documented, ensuring traceability and reproducibility as well as enabling projects to add or modify steps as required.

Such a workflow enforces consistent TEI structures through automated validation and controlled transformations, supporting interoperability across platforms and facilitating reuse—and consistent tagging is an important cornerstone in creating sustainable editions.

Automation has two important impacts besides reducing repetitive tasks: large amounts of texts can be easily converted to TEI—which might be prohibitively time consuming when done manually–, and it means that a basic TEI representation of a text can be easily and quickly realised, improving access to written heritage.