Abstracts

Making Sense of Versions: A Collaborative Model of Textual Variation

Beatrice Nava (Universität Wien) / Elli Bleeker (KNAW Amsterdam)

The past decades have witnessed the rapid development of computational methods to study and visualise textual variation. Still, some persistent challenges remain: (1) a lack of a shared vocabulary to describe textual variation; (2) the limited interoperability of tools and their output, and (3) the difficulty to create informative visualizations of complex cases of textual variation. Earlier this year (April 2025), an international group of scholars, developers, and interface designers came together to collectively address these challenges. Their work is continued in the working group VIDIT (Visualising and Investigating Differences in Texts). Two subgroups are devoted to designing a set of guidelines for best practices for visualisation tools and developing a shared vocabulary for expressing textual variation. In this contribution, we will elaborate upon the second objective and propose a formal classification of textual scholarship terms such as hand/agent and modification. We will argue how and why such a classification will facilitate the development and interoperability of tools by providing interoperable annotations, and as such pave the way for new research into textual variation, e.g., by studying variation across literary works, genres, or time periods. The specific aim of our talk is to present our methodology, gather feedback from the participants, and inspire them to participate in our working group.

Visualising complex textual variants with the interactive tool LERA

Marcus Pöckelmann (Freie Universität Berlin)

In this presentation the webbased tool LERA (https://lera.uzi.uni-halle.de/) is showcased, which allows the interactive collation and analysis of multiple witnesses of a text. LERAs integrated semi-automatic algorithms for comparison and distant-reading visualisations can be applied for a wide range of different text types and research questions. Recently, algorithms for cross-language comparison based on LLMs are developed within the research project COMUTE (https://www.comute-project.de/en/), which will be made available in LERAs interface as well. The presentation will also address this development and present technical details on initial results.

Multilingual Text Alignment and Editing Practices: The COMUTE Project

Sandra Balck / Brigitte Grote (Freie Universität Berlin)

The DFG-funded COMUTE project (www.comute-project.de) explores how multilingual and only partially parallel texts can be aligned using algorithmics and how these results can be applied to digital editorial workflows. The project aims to identify variation on paragraph, sentence,  phrase and word level. Examples such as Hannah Arendt’s bilingual self-revisions, and the English and German versions of Melville´s Moby Dick and Salinger´s The Catcher in the Rye demonstrate the complexity of these variations.
COMUTE develops methods for multilingual alignment and examines how researchers can analyse, review and incorporate these methods into their editorial work. This talk aims to demonstrate how algorithmic approaches can reveal patterns of multilingual variation that are difficult to detect through manual comparison. It will also discuss how such results can support editorial workflows and open up new possibilities for presenting individual texts. 

Analysing Literary Variants and Versioning: (How) Can LLMs Help? 

Gerrit Brüning / Felix Schenke (Klassik Stiftung Weimar)

The idea of systematically examining the differences (variants) among various versions of a literary text has a long history. It was already articulated by Lessing who discussed the revision of Klopstock’s Messias (Letters on the most recent literature, 1759–1765, letter 19). In the twentieth century, attempts at providing a theoretical framework emerged from a structuralist context (Mukařovský 1930; Červenka 1970). Digital text data and computational methods have raised hopes of advancing these previously scattered and programmatic approaches to a new level (Brüning 2022). However, the transition to such a level is far from straightforward: current Computational Literary Studies have not yet developed methods specifically designed to open up this field. A first attempt to apply LLMs on textual variants has, however, yielded promising results (Jannidis 2024). The presentation offers insights into experimental approaches for the analysis of textual variation within an edited corpus of Goethe’s lyric poetry. 

Tracing textual genesis through parts-of-speech-tagging of selected poems by Friederike Mayröcker (Master Thesis)

Johanna Unterholzner (ÖAW Wien)

My master thesis attempts to describe the genesis of poetry through the lens of quantitative text analysis, using parts-of-speech tagging of selected poems by Friederike Mayröcker. The main research question focuses on identifying the steps and methods that aid in modeling the text genesis with several text witnesses and revision stages and preparing them for a quantitative analysis. In addition, it investigates the extent to which quantitative text analysis focused on parts of speech can provide insights into possible editing tendencies in the case study. In light of current developments, it seems essential to reaffirm enduring scholarly standards while remaining open to newly emerging analytical potentials of the material. 

Evaluation-Oriented Generation of Text-Genetic Interpretations from Non-Digital Editions

Axel Pichler (Universität Wien)

In literary studies, there is a discrepancy between the significant effort involved in creating critical editions and their concrete use in the interpretation of literary texts. Interpretation, understood as the attribution of meaning to a text, has so far played a minor role in Computational Literary Studies (CLS) – even in the era of Large Language Models (LLMs). This paper presents initial experiments with an LLM-based approach that generates text-genetic interpretations of individual poems based on non-digital text-genetic editions. The workflow employed relies on reference data from literary studies and their argumentative core-structures. This enables the generation of interpretations that incorporate such structures and can be evaluated both in terms of their argumentative coherence and their adherence to standards of literary studies.

Challenges in Automatic Identification of Indirect Quotations between Scholarly Texts and Literary Works 

Robert Jäschke / Frederik Arnold (Humboldt-Universität zu Berlin)

As part of our project, Is Expert Knowledge Key? Scholarly Interpretations as Resource for the Analysis of Literary Texts in Computational Literary Studies, we investigate how scholarly texts, that interpret literary works, reference the literary work and how different types of references, for example, verbatim quotations, or indirect references, such as, summaries or paraphrases, can be automatically extracted. In this presentation, we first give an overview over our specific context and application of references in scholarly texts and literary works. We then present results of our investigation into the automatic identification of indirect quotations. We found that indirection quotations are hard to identify, even for humans, especially since a scholarly text necessarily references the literary work and the distinction between indirect reference and interpretation is particularly difficult. We identified a number of challenges that remain unsolved and hope for fruitful discussions around possible solutions.

Exploring gender references in Wertheriaden through network analysis

Janis Pagel (Universität zu Köln) / Marie Flüh (Universität Hamburg)

"Die Leiden des jungen Werther" by Johann Wolfgang von Goethe brought the young author almost instant, Europe-wide success and is now regarded as part of a canon of world literature. From the novel's publication to the present day, other authors have produced works that reference "Die Leiden des jungen Werther" in various ways, the so-called "Wertheriaden". In our talk, we examine manual and automatic annotations of gender references in a selection of these Wertheriaden and explore how literary character constellations related to gender differ both within the Wertheriaden and in comparison with the original. To this end, we create and analyse networks that visualise the relationships between characters with respect to their gender representations.