This plenary talk examines the future of open scholarship in the age of AI, arguing that openness must be understood not simply as access to research outputs, but as a broader commitment to public accountability, infrastructural justice, and epistemic plurality. Beginning from the perspective of digital humanities in Quebec, the talk emphasizes that knowledge is always shaped by its material and technical conditions: archives, metadata, platforms, interfaces, corpora, standards, and tools do not merely transmit scholarship, but actively structure what can be seen, asked, and interpreted. This insight is especially urgent in a context where scholarly communication is increasingly mediated by commercial platforms, proprietary infrastructures, and AI systems. The talk situates open scholarship within debates on open access, open social scholarship, and “open humanities.” Drawing on the INKE partnership and related work, it argues that openness in the humanities must include participation, reuse, public engagement, sustainability, multilingualism, and contestation. Humanities scholarship requires infrastructures that support not only dissemination, but also interpretation, collaboration, accessibility, and long-term stewardship. It must be reinvented as a public, critical, multilingual, and accountable project: one capable of shaping AI-mediated knowledge infrastructures rather than merely adapting to them.

The <foreign> element has been part of the Text Encoding Initiative (TEI) standard from its earliest editions. Defined as identifying “a word or phrase as belonging to some language other than that of the surrounding text,” its purpose is both functional—to be able to identify words that do not belong to the declared primary language of the text—and descriptive, providing a semantic equivalent for marking phrases that, in the Western print tradition, are usually signalled by typographical difference (e.g. italics). It is also, as the TEI Guidelines make clear, an element of last resort: since every element in XML can specify the language of its content using the @xml:lang attribute, <foreign> is arguably meant to be used when there is no more specific element on which to place @xml:lang.

Yet, as we will suggest, the <foreign> element is neither semantically neutral nor purely functional. Emerging from cross‑project discussions of shared challenges, this co-authored paper examines the practical, interpretive, and political challenges that arose out of attempts to represent editorial decisions in multilingual texts. Drawing on case studies of distinct projects spanning various linguistic, cultural, and historical contexts, this paper describes how the <foreign> element, while perhaps technically appropriate, was, at best, conceptually insufficient and, at worst, actively working against the project's social and political aims. As we will outline, for texts where languages co‑exist, overlap, and mutually constitute the content, the label “foreign” risks reinscribing hierarchical language ideologies, especially in contexts shaped by colonial histories, racialization, and language suppression. We will also describe the ways that our projects have taken different approaches to better represent the multilingual nature of the text and the critical decisions that inform the final editions.

To be sure, this paper is not an attempt to do away with the <foreign> element completely. Rather, in focusing on the <foreign> element and its drawbacks, we hope not only to argue for changes to the TEI standard, but also showcase what we see as a generative moment of friction between the TEI standards and editorial practice—a case study in the blurry lines between textual interpretation and technical protocols that underscores how seemingly neutral technical features carry with them significant assumptions about shared norms and practices. In doing so, this paper hopes to demonstrate how critical collaboration and dissent, to use Julia Flanders’ framing, not only advances the TEI standard and editorial practice, but ultimately testifies to the critical and interpretive value of digital scholarship.

Problématique

En contexte numérique, le texte édité se comporte vite comme un objet stabilisé, indexé, cité, réutilisé, parfois même adopté comme « référence ». Cette stabilisation transforme une chaîne de choix interprétatifs en résultat apparemment évident, alors même que la numérisation multiplie les opérations qui pèsent sur le sens (segmentation, normalisation, encodage, alignement, visualisation). En philologie classique, pourtant, cette chaîne est le cœur du travail. Les textes issus de la tradition manuscrite sont instables et lacunaires, et l’apparat critique signale des variantes sans toujours expliciter la logique des arbitrages ni les cadres linguistiques, métriques, stylistiques et stemmatiques qui orientent une leçon plutôt qu’une autre. La transition vers le numérique rend ce décalage plus aigu lorsque des catégories computationnelles figent ce qui demeure graduel, contextuel et argumentatif dans l’atelier.

La thèse défendue est que l’unité critique fondamentale d’une édition numérique n’est pas seulement le texte et ses variantes, mais la décision éditoriale elle‑même, conçue comme un objet adressable, traçable et discutable. L’objectif n’est pas de mécaniser le jugement philologique, mais de le rendre vérifiable, auditable dans le temps, comparable entre projets et interrogeable par requêtes. Un modèle argumentatif minimal, inspiré de Toulmin, déplace ainsi l’édition d’un « texte et apparat » vers une « édition‑argument », où les raisons, objections et qualifications sont éditées au même titre que la leçon.

Méthodologie

Une décision est formulée comme une conclusion liée à une cible textuelle stable. Elle s’appuie sur des données (témoins, tradition éditoriale, cohérence linguistique, plausibilité métrique, vraisemblance stylistique), explicite une garantie comprise comme règle d’inférence, documente des appuis épistémiques, enregistre des contre‑lectures et situe un degré de certitude. La proposition est démontrée sur une édition‑prototype de l’Anthologia Graeca, centrée sur un court ensemble d’épigrammes funéraires, choisi pour la densité des cas limites.

Le travail s’appuie sur Philographos, atelier web créé pour le projet, conçu pour lier chaque intervention au dossier de justification qui l’accompagne et pour en conserver les versions. L’architecture articule deux couches : une couche textuelle sérialisée en TEI‑XML (texte établi, segmentation, apparat) et une couche argumentative sérialisée en JSON‑LD, où les décisions et leurs relations forment un graphe. L’alignement repose sur des identifiants stables, notamment des URN, de sorte qu’un locus précis puisse être visé, audité, comparé et révisé sans dépendre d’un affichage particulier.

Conclusion

La contribution originale est en deux temps. D'abord, un schéma de données réutilisable pour éditer des décisions comme objets évaluables, et un principe d’architecture (texte hiérarchique d’un côté, justification réticulaire de l’autre) compatible avec la lecture, l’audit et la pérennisation. Le dispositif rend visibles des médiations souvent implicites, permet de cartographier des zones d’incertitude et d’explorer des désaccords, tout en assumant qu’une part du sens résiste au calcul. Formaliser la décision ne supprime pas cette résistance, mais la rend situable et discutable, et oblige l’édition numérique à porter explicitement la responsabilité de ses transformations.

The computer and the artifact are not the same kind of object. They differ in many ways such as materiality, scale, durability, and mode of encounter. As Robinson and Solopova (1993) argue, digital systems cannot offer transparent access to the source. Therefore, editorial work consists of a sequence of interpretive decisions; this condition becomes unavoidable in the digital world, where ambiguity cannot remain unresolved. A system must decide what each mark is in order to register it at all, even when such decisions distort the source. As Barbara Cassin (2016) asserts, there is no equivalence in translation, only negotiation, and that process always exposes loss. Drawing from this framework, this paper asks what is lost when monumental stone crosses in the Visionary Cross Project are translated into data, and how a 3D edition might respond responsibly to those losses.

The Visionary Cross Project emerged from a shared interest in developing a new kind of digital scholarly edition, one that not only preserved Old English texts but also integrated them meaningfully with cultural artifacts. Specifically, the project focuses on 3D visualizations of monumental Old English stone crosses. These artifacts are significant not only for their inscriptions and iconography but also for their connection to Old English religious poetry, which shares thematic and visual motifs with the crosses. The goal of the Visionary Cross Project is to move beyond traditional print/static digital editions by linking these 3D models with annotated texts, translations, and scholarly commentary, and thus creating a 3D edition that facilitates deeper engagement for scholars and the public.

While the scans capture details such as texture, colour, and weathering with accuracy, aspects such as scale and physical presence resist translation. Encountering the crosses digitally offers little sense of their mass, height, or bodily impact, as monumentality is necessarily flattened into a screen-based interaction. Place, similarly, is abstracted as the crosses circulate as mobile digital objects detached from their setting. Editorial interventions and technological limitations further induce untranslatability. As storage and hosting limits encourage us to display either smaller models or divided forms of the cross, each approach confronts us with the reality of making a 3D edition. Additionally, as we deal with linguistic data in the form of runic inscriptions, which resist stable representation: retaining the runes preserves visual and historical specificity, transliteration introduces approximation, and translation imposes interpretive constraints. By treating untranslatability not as a failure but as a productive condition, this paper positions the 3D edition as a site where loss, mediation, and interpretive choice are made visible. In doing so, it argues for a model of digital scholarship that acknowledges the limits of this translation while using them to generate new forms of critical engagement with material, textual, and cultural heritage.

References

Cassin, Barbara. "Translation as paradigm for human sciences." The journal of speculative philosophy 30, no. 3 (2016): 242-266.

Robinson, Peter, and Elizabeth Solopova. "Guidelines for Transcription of the Manuscripts of the Wife of Bath’s Prologue." The Canterbury Tales Project Occasional Papers 1 (1993): 19-52.

This paper reflects on the current transformations of information retrieval (IR) systems, recommendation systems (RS), and automated literature review tools. Despite their potential impact on innovation and discoverability in science, the role of these systems remains largely invisible. The integration of AI systems into all phases of IR suggests that new, discrete research practices —such as those described by Clavert and Muller —are becoming entrenched without full awareness of their impact on knowledge production. In this paper, I will update the stakes of scholarly literature discoverability through a state-of-the-art review of AI research assistants, assessing their influence on the scientific ecosystem and emerging discrete research practices. I will also propose avenues for reflection, design suggestions, and architectural frameworks for recommendation and information retrieval systems tailored to address these major challenges.

The reliability of AI Large-Language Models used in academic research is a growing topic in the recent AI conversational landscape. A 2024 study by Liao et al. reported that 81% of survey respondents had already incorporated LLMs into their research workflow. Two common problems cited with LLM-based research are "hallucinations" (plausible but nonetheless fictional results) and so-called "temporal shifts" that can occur when LLMs are trained on one corpus from a particular period of time but are deployed on texts from a different time period (Ushio; Rosin). These temporal problems -- shifting word definitions and different discursive language practices -- can thwart the LLM's ability to generate useful results. This paper is an attempt to understand and rectify those two problems by building a local (that is, not cloud-based) retrieval-augmented generation (RAG) system using the open source model, Ollama.

Our test-case corpus is a selection of historical brewing texts. We have chosen this corpus for two reasons: 1) very few of these texts are available online as either plain text or XML files and so are not easily accessible for off-the-shelf LLM training; and 2) this well-bounded corpus highlights both hallucinations and temporal shifts quite dramatically. The discourses surrounding the brewing industry have shifted considerably over the XXX years of our corpus, making these texts a robust and well-bounded arena for experimentation.

The language variations in this corpus are numerous and vexing enough for humans -- from the oft-repeated historical difference between "beer" (with hops) and "ale" (without) to a more recent distinction between "ale" (using top-fermenting yeast) and "lager" (using bottom-fermenting yeast). Or the confusing use of the terms "porter" and "stout" that were once merely adjectives but have evolved into distinctively different beer styles. Moreover, across the centuries, the backdrop of science has also added and subtracted terminology: from 17th-century Newtonian physics (the "break") to 18th-century chemistry (which saw fermentation as an acid-base reaction) to 19th-century biology (when yeast was first identified as an organism that digested sugar and produced alcohol as a by-product). The description of the brewing process is always framed by a time-shifting understanding of the science behind the art.

A significant limitation for this project has been acquiring reliable plaintext versions of historical texts. Our corpus is still small, but is growing quickly and we hope very soon to have a much larger, and more robust, corpus to serve as the repository for our local LLM.

Translation Studies (TS) appears to occupy a peripheral position in the Humanities and Social Sciences, as evidenced by the fact that the CFP focuses on the theorization of translation without explicitly considering the discipline that squarely focuses on translation – and its corollaries such as the untranslatable. Indeed, the CFP doesn’t account for the many voices from contemporary TS that have explored the untranslatable and the relationships between translation and (physical/digital) materiality (Littau 2016), translation and embodiment (Ivancic and Zepter 2022), translation and cultural specificity (Al-Ghadeer et al. 2025), translation and online/digital media (Desjardins 2017, 2019, 2020), translation and the Internet of Things (Desjardins 2025), and translation in the era of artificial intelligence (AI). In fact, in the Routledge Handbook of Translation Technology and Society (Baumgarten and Tieber 2025), a chapter is dedicated to TS and the Digital Humanities (Tanasescu 2025). Therefore, to contextualize our presentation, we overview some key contemporary debates located at the TS/DH nexus, including the concept of the untranslatable/translatable, both metaphorically and literally. How might we conceptualize what is translatable/untranslatable in an era marked by surveillance capitalism, algorithmic curation, and non-human communication/translation? This leads into the presentation’s second part and focus, which is multilingual, translational, and algorithmic trends within the online creator economy, with specific attention given to Canadian creators and influencers. The content created for social platforms is key to DH, including as the source text for cultural analysis and analytics. Questionnaire data from a SSHRC-funded project on translation and multilinguality within the Canadian creator economy will be presented, elucidating multilingual practices among Canadian creators: do they use AI to translate their content? Do they self-translate? What do they find resists translation or facilitates translation and what novel strategies might they employ? Further, given that online engagement and monetization are often premised on ‘posting for the algorithm’, what algorithmic strategies are employed? Building on McCulloch’s work in Because Internet (2019), linguist and influencer Adam Aleksic’s book states: “[…] the internet has ushered in an unprecedented linguistic upheaval. We’re entering an entirely new era of etymology heralded by the invisible forces driving social media algorithms”. If, as Aleksic’s work suggests, “communication is changing in both familiar and unexpected ways,” from the use of emojis to the way different generations talk about taboo subjects, how do translation and multilinguality operate in these (arguably) transnational settings and what lessons might we draw from these multimodal and multilingual encounters? With a focus on systemic inequities, how does translation—linguistic, cultural, and algorithmic—shape how new content is created in digital platforms? Finally, we hope to dispel the idea that English is the turnkey language for engagement on the social internet, which runs parallel to ideas presented in the preface of the Dictionary of Untranslatables (Cassin et al. 2014) (English version of le Dictionnaire des intraduisibles), including “exploring the places where languages touch” and revealing “the limits of discrete national languages and traditions”. After all, as tech journalist Taylor Lorenz (2025) reports: “the global [multilingual] influencer era is upon us”.

This presentation reports the findings of a comprehensive environmental scan conducted by the ETCL (Electronic Textual Cultures Lab) at the University of Victoria, mapping the "encounters and entanglements" between generative AI and Open Social Scholarship (OSS). Moving beyond the initial hype cycle, we offer a detailed topography of how AI is reconfiguring knowledge production across three critical axes: the Open (the paradox of open access as training data), the Social (crises of provenance and algorithmic governance), and the Scholarly (methodological augmentation and epistemic risk). We synthesize key themes emerging from the literature, including the tension between commercial speed and community consensus, the invisibility of AI labor, and the rise of "centaur" workflows that blur human and machine authorship. The session concludes by offering navigational principles for scholars and librarians, arguing that OSS values—reciprocity, care, and public good—are essential for governing these increasingly synthetic information ecosystems.

Les environnements d’écriture véhiculent des modèles épistémologiques, des visions du monde et des paradigmes d’émergence du savoir (Delannay 2025). Or, les pratiques contemporaines d’écriture académique convergent vers une poignée de formats propriétaires et de plateformes proposant des solutions par défaut (Vitali-Rosati 2024 ; Guichard 2014). Dans le champ des sciences humaines et sociales (SHS), cette situation constitue un appauvrissement majeur de la réflexion épistémologique : une part essentielle du geste scientifique implique la production et la structuration du texte (Kembellec 2019 ; Bert 2017), qui se trouvent de facto déléguées à des dispositifs contrôlés par des entreprises dont les objectifs et les valeurs ne coïncident pas avec ceux de la recherche. Cette proposition présente Stylo, un éditeur de texte développé depuis 2017 par le Laboratoire sur les écritures numériques de l’Université de Montréal, fondé sur des principes d’ouverture, de pérennité et rendant possible une appropriation critique des infrastructures de production du savoir. Il est aujourd’hui déployé dans l’écosystème de la TGIR Huma-Num, où il soutient les workflows éditoriaux de nombreuses revues et de communautés de recherche en SHS.

Au cœur de Stylo se trouve un principe simple : une source unique, composée de trois formats ouverts et interopérables, permet la production de multiples rendus. Le balisage Markdown pandoc-flavoured assure la structuration sémantique du texte. Le YAML encode les métadonnées : les schémas disponibles sont l’aboutissement d’un travail en lien avec l’écosystème des revues francophones et les acteurs du monde académique pour modéliser les objets textuels de la production scientifique. Le fichier BibTeX gère les références bibliographiques. Ce triptyque, grâce à Pandoc, sert à la production de plusieurs formats de sortie (HTML, PDF via LaTeX, XML-TEI, TEI-Commons) qui conservent la structuration sémantique et permettent la circulation d’un même texte entre des chaînes éditoriales diverses et spécialisées.

De récents développements ont permis la modélisation et l’implémentation de structures de métadonnées entièrement personnalisées, ainsi que la multiplication des artefacts éditoriaux de sortie (thèses et livres augmentés, carnets de recherche et sites de revues). Cela permet d’interroger les relations entre la pluralité des modes de production et circulation des connaissances en SHS, la diversité des modèles épistémologiques des communautés savantes et la composition matérielle des environnements d’écriture. Cette proposition montrera comment Stylo incarne cette articulation en proposant un cadre méthodologique unifié mais modulaire.

L’engagement de Stylo est d’abord théorique : il s’agit de réfléchir à ce qu’est l’écriture scientifique aujourd’hui et de l’implémenter dans un environnement numérique. En défendant des standards et des formats ouverts, il garantit la pérennité de l’accès aux contenus, et offre une alternative à l’hégémonie des solutions commerciales. Il est aussi méthodologique, en favorisant la structuration sémantique des contenus par les auteurs·ices. Il est enfin social, parce qu’il encourage la collaboration de plusieurs expertises au sein de communautés de recherche et d’édition, tout en visibilisant chaque implication. En définitive, l’engagement de Stylo favorise un rapport critique et maîtrisé aux technologies d’écriture, tout en restant ouvert et adaptable.

Generative AI has introduced a new layer of uncertainty into academic life. For many scholars, GenAI appears simultaneously as powerful engines of efficiency and as destabilizing forces that may intensify publication competition, disrupt authorship norms, and complicate assessment integrity. Existing research documents widespread uptake of generative AI and recurring concerns around ethics, workload, and academic integrity, but it typically treats anxiety as a stable attitude, abstracts away from institutional context, and rarely examines how experiences differ across academic rank.

This project starts from the premise that AI-related anxiety is structurally mediated by career stage, such that pre-tenure and tenured faculty inhabit systematically different “risk landscapes” even within the same institution. Integrating the Technology Acceptance Model (TAM) and Cognitive Appraisal Theory (CAT), the study examines how faculty make sense of, experience, and respond to generative AI. From TAM, it incorporates perceived usefulness and ease of use; from CAT, it focuses on appraisals of GenAI as threat versus challenge, along with perceived control and available resources. Rather than treating GenAI as neutral tools, the study conceptualizes them as stressors whose affordances shape emotional responses and coping strategies. It further argues that appraisal–anxiety–coping pathways are conditioned by structural vulnerability associated with tenure status (pre-tenure versus tenured).

The study compares how faculty at different career stages appraise GenAI experience, AI‑related state anxiety, and adopt coping strategies in research and teaching contexts. It asks: (1) How do pre‑tenure and tenured faculty differentially appraise GenAI as threats versus challenges? (2) How do perceived usefulness and ease of use shape coping responses such as problem‑focused experimentation, workflow redesign, selective disclosure, avoidance, or deferral? (3) How do faculty themselves understand the consequences of these appraisal–anxiety–coping configurations for workload, well‑being, integrity governance, and perceived fairness in academic evaluation?

Methodologically, the project employs a qualitative-comparative online survey targeting permanent academic staff at the University of Alberta. Participants are invited to describe recent, concrete GenAI-related episodes in their academic work and to reflect on their appraisals, emotional responses, perceived control and resources, and coping actions. Open‑ended narrative responses constitute the core data, supplemented by a small set of structured items capturing career stage, GenAI exposure, broad appraisal categories, and coping tendencies. Data will be analyzed using thematic analysis with a hybrid deductive–inductive codebook aligned with the integrated TAM–CAT framework, enabling systematic comparison across career stages while remaining sensitive to emergent, institution‑specific themes.

The study aims to produce a fine‑grained account of how structurally vulnerable and structurally protected faculty experience and manage AI‑related pressures within a single institutional setting. It seeks to inform that rank‑sensitive, adaptation labour, and emotional burdens may be unevenly distributed. It offers a portable, theory‑driven qualitative framework for examining AI anxiety, resilience, and equity in academic work- one that speaks directly to debates about automation, authorship, and the governance of scholarly labour in an era of ubiquitous generative AI.

Digital humanities initiatives now span multiple generations of scholars, infrastructures, and institutions. As projects, platforms, and partnerships mature, questions of leadership transition and succession have become central to the future of the field. Many project founders have devoted time, energy and intellectual engagement to ensure that their initiative survives shifts in technology over time and endures as key scholarly infrastructure but have done relatively less to facilitate the successful transition and succession of people as they retire, change institutions, pursue new research projects or step back for other reasons. At present, little is written beyond anecdotes and cautionary tales to provide guidance. This paper aims to add to that body of knowledge through the examination of a series of success stories from digital humanists who have navigated various transitions and successions to ensure projects are transferred, transformed and sustained over time.

This paper will present the newly migrated and redeveloped DigitalDonne as a case study in the necessity of placing sustainability at the center of scholar-led, open-access digital resource development. After fifteen years at Texas A&M University (https://digitaldonne.tamu.edu/), where it was built with substantial external funding and institutional support, DigitalDonne suddenly found itself in search of a new home. With no faculty member remaining to champion this important resource, its host institution could no longer commit to its ongoing support. This situation is far from uncommon as mature digital projects age, faculty retire, and hosting units within universities face increasing resourcing constraints. In assuming responsibility for the site, we determined to do so in a way that will minimize the chances of a similar crisis down the road when its new champion of this resource at USask (Nelson), eventually, retires from his faculty position. In this presentation, we will present the strategic choices we made in our redevelopment of the site in the interest of sustainability. A key point of departure for our thinking were the principles set out by the Ending Project for ensuring digital longevity (https://endings.uvic.ca/principles.html), chiefly 4.1: “[n]o dependence on server-side software.” A common point of resistance in university hosting support is the database. Such was the case with DigitalDonne, which in its pre-migration state relied on databases to coordinate and serve content. A second important point of reference was established best practice in web development aimed at ensuring the legibility, maintainability, and transferability of responsibility to future developers who might inherit responsibility for this resource. We begin by elaborating on these two considerations, then describe the strategic methodological and procedural choices made during redevelopment and conclude by discussing the results of these decisions as embodied in the final product, including a brief demonstration.

Digital humanities research which is both accessible and public facing, is also susceptible to unintended and unexpected changes, yet these changes—the degradation of online projects which occurs gradually, constituting an understated and misunderstood complex problem—strike at the heart of the “irreducibility” of the aesthetic, those “untranslatable” elements or components that contribute to the holistic nature of artistic and/or cultural understandings and works. “Irreducibility”—a key word and concept across the entirety of Cassin, et al’s, work on the untranslatable, carries two competing definitions: the technical resistance to being broken into constituent parts, and the metaphysical transcendence of those parts (OED’s “invincible, insuperable” etc); we argue that both definitions intersect in the public awareness and the desire to protect and preserve culturally open web-based DH projects. However, while recommendations are in place for creating stable and long-lasting resources, they can be difficult and problematic to implement for projects that have been running for some years.

Our previous work has addressed digital preservation questions. In this presentation, we align with the conference theme by exploring the “untranslatable” features of projects: how has the digital representation of documents introduced biases, reductions, and simplifications—affecting our preservation efforts and current practices. How can aspects that resist translation be preserved?

We have been analyzing projects published in the DH Book of Abstracts for 18 years (2007 to 2025), analyzing their changes and shifts over time. We store periodic snapshots for each project as WARC files, examine functioning and non-functioning components, extract their text using Python’s BeautifulSoup], and retrieve HTTP response codes and headers with Requests library. We do this to classify the snapshots by their year of publication into groups depending on the status of its server response. We also plan to apply deep learning, real-time computer vision and AI-assisted software development on cases that are human-perceivable, but currently not algorithmically identifiable by machines. Computer vision is a way of giving intelligent machines ‘eyes’ to see data, bringing our methodology into new visual dimensions of access and interpretation. Our purpose is to provide systematic methods to sustain and preserve culture and the digital scholarly infrastructure in the humanities over time, preventing their degradation and decay.

It is our intention to also preserve the researcher’s intent and perspectives. Through this process, focused on ethical/academic data collection and feature extraction, we can visually track the changes of these projects to their potential points of abandonment—furthering the conversation on the preservation of knowledge. We conclude that identifying and collecting features from online projects is feasible and can be utilized to develop metrics to identify changes in content and knowledge in the digital humanities. At the same time, we are invested in preserving the untranslatable aspects of DH research, and we would like to expand this discussion with other members of the community. We aim to provide insights and preservation strategies that could be applied to other online projects in the digital humanities, ensuring a long-term future so that more widespread access and exploration of these resources becomes possible.

Introduction

Au XVIIIe siècle, la reprise d'un texte passe rarement par la citation littérale d’un passage. Elle se manifeste plutôt par des paraphrases, c'est-à-dire des formes d’appropriation reformulante qui peuvent inclure la reprise d’une idée avec autre lexique, le déplacement d’images, le réagencement rhétorique, ou l'emprunt d’un tour stylistique. Or, dès qu’on cherche à repérer ces phénomènes à l’échelle d’un corpus, on se heurte à une tension méthodologique : les opérations nécessaires à la comparaison automatique (extraction, nettoyage, segmentation, normalisation) renforcent la comparabilité, mais risquent d’effacer précisément les indices de contexte et de style qui rendent une parenté interprétable.

Cette communication présente un travail de détection de paraphrases à partir des Caractères de La Bruyère, comparé à un corpus d’œuvres du siècle des Lumières et de la période (un sous-ensemble exploratoire incluant Montesquieu, Diderot et Crébillon, appelé à être réévalué et élargi). Nous proposons un protocole d’extraction, de nettoyage et de comparaison de passages destinés à produire des listes de candidats, puis à affiner les méthodes en fonction d’une validation experte. La thèse défendue est que la détection de « paraphrases » n’est pas un simple problème de correspondance. Plutôt, chaque choix de représentation numérique du texte configure ce qui devient visible, ce qui se déforme et ce qui disparaît dans l’analyse.

Méthodologie

Les textes sont nettoyés et segmentés en fenêtres textuelles comparables, puis évalués au moyen de plusieurs familles de similarité afin de capter des proximités de nature différente. Les proximités lexicales reposent sur des représentations de type bag-of-words avec pondération TF‑IDF ; la recherche de motifs et de patrons phraséologiques utilise des n‑grammes ; enfin, des représentations vectorielles fondées sur des modèles de type BERT (sentence-transformers) permettent de proposer des rapprochements plus conceptuels. Les résultats sont agrégés en classements de paires candidates, accompagnés d’une trace complète du paramétrage (taille de fenêtre, normalisation, métrique, seuil) afin de rendre l’expérience reproductible et la discussion méthodologique possible.

Évaluation et annotation

L’évaluation ne se limite pas à une métrique globale. Un échantillon de paires est validé par lecture experte et annoté selon le type de transformation observé (reprise lexicale, substitution, réorganisation, amplification, déplacement de point de vue, emprunt de tournure). Cette étape sert à calibrer les seuils, mais aussi à qualifier les erreurs typiques : faux positifs dus à des patrons linguistiques et topoi partagés et faux négatifs liés à des réécritures profondes, à l’ironie ou à des variations de registre. Le résultat est un modèle plus précis, et des données annotées, tant positivement que négativement.

Conclusion

Les conclusions attendues sont doubles. D’une part, une cartographie de proximités hiérarchisées entre œuvres, utile pour formuler des hypothèses de circulation et d’influence sans confondre proximité numérique et preuve. D’autre part, une analyse des compromis introduits par chaque chaîne de traitement : ce que l’on gagne en comparabilité et ce que l’on paie en appauvrissement des signaux stylistiques. La contribution originale réside dans l’articulation entre chaîne de traitement reproductible, validation qualitative et interprétation littéraire, afin de faire de la distance textuelle un outil d’enquête critique plutôt qu’un verdict automatique.

Les modèles d'intelligence artificielle multimodaux prétendent résoudre la traduction entre médias artistiques en réduisant textes, images et sons à des vecteurs dans un espace latent. Pourtant, loin de résoudre l'intraduisible intersémiotique, cet aplatissement vectoriel le révèle à cause des compensations nécessaires au passage d’un langage verbal ou même médial à un autre.

La traduction intersémiotique, passage d'un système de signes à un autre (Jakobson, 1959), permet d’observer l'intraduisible dans les humanités numériques. Chaque médium dépend de matérialités spécifiques. Ces systèmes rendent l'équivalence parfaite structurellement impossible (Bennett, 2019 ; Tekgül Akın & Kiran, 2023).

Je prendrai ici deux cas de traduction. Le premier est une traduction de langue : The Raven d’Edgar Allan Poe par Baudelaire. Le second une traduction de langage médial : les deux itérations de Théorème de Pasolini (1968). Le second cas est d’autant plus intéressant que l'auteur a simultanément créé le roman et le film, négociant différemment les mêmes thèmes selon les affordances (Murray 2012) de chaque médium, ce qui révèle l’importance de la matérialité en tant que la somme des caractéristiques physiques des matériaux et technologies avec l’intervention de l’artiste (Hayles 2002).

Les modèles multimodaux tentant des traductions intersémiotiques similaires opèrent par compensation probabiliste. Cette compensation produit des artefacts intéressants, comme des « artefacts cognitifs » (Queiroz & Atã, 2019). Les « hallucinations », approximations et choix génératifs révèlent ce qui ne peut être aplati (media collapse : Impett & Offert 2026) dans l'espace latent.

Ma méthodologie combine analyse comparative et expérimentation : j'examine les négociations créatives chez Baudelaire et Pasolini, puis soumets des segments équivalents à des modèles multimodaux pour observer les stratégies de compensation algorithmique. Cette comparaison montre que l'aplatissement vectoriel révèle les spécificités médiatiques par contraste.

Cette recherche contribue à une compréhension critique des modèles multimodaux en humanités numériques. L'intraduisible n'est pas une défaillance technique, mais une effet corolaire de la création artistique. Les projets de numérisation et de remédiation doivent reconnaître que la traduction intersémiotique implique des pertes, des gains ainsi que des transformations créatives.

Efforts to preserve software for future research tend to primarily deal with locally executed legacy software, ranging in complexity from fairly simple word processors, to complex video games with more stringent requirements for execution speed and accuracy. In recent years, emulation is the method most often used to accomplish this, highlighted by multiple scholars for its accessibility, scalability, and consistent ability to preserve the primary research value of complex software objects. Emulation is conducted through “computational, technical processes that use one system to reproduce the functions and results of another system,” generally aiming to reproduce a legacy system in its entirety to allow the operation of software originally designed for that system.

However, in the case of Massively Multiplayer Online games (MMOs), preserving the game’s software and network functionality through emulation presents a barren, lifeless world that gives little indication of how the game was once played. MMOs are online multiplayer games that are designed around large numbers of concurrently connected players (ranging from hundreds to tens of thousands), and the player cultures developed within these playerbases have substantial effects on how the game is played. Due to this, MMOs are complex digital interactive systems that have social dependencies in the form of numerous concurrent players, posing further barriers to their effective preservation atop the more common considerations for local multiplayer games or online multiplayer games of smaller scale.

Ethnographic approaches to preserving MMOs have been suggested by multiple scholars, aiming to study the groups and complex player cultures within MMOs that are still alive. Additionally, after (or, more contentiously, before) an MMO’s publisher has shut down its official servers and rendered it ‘dead’, users (or sometimes a single dedicated user) may begin hosting their own server (typically called a ‘private server’) for the MMO that others can then connect to, allowing the MMO to be playable again.

Private servers are often overlooked as effective preservation tools due to their legal hurdles; however, though they do often infringe on digital copyright law, the City of Heroes: Homecoming private server acquired official rights to operate from the original publisher of the MMO and exists legally, serving as an important proof of concept of a legal private server. Because some maintain playerbases in the thousands, which can begin to recreate the original experience of playing the MMO, private servers are rife with potential as preservation tools. This paper will explore this potential, as well as its general feasibility, through a review of the literature pertaining to MMO preservation, especially where the use of private servers applies, alongside discussion of recent approaches, and synthesis and analysis of both the existing literature and online primary sources pertaining to private servers. I will also include analysis of select private servers of dead MMOs that are currently active and their communities, aiming to assess the plausibility of institutionally hosted private servers of dead MMOs for preservation, a novel means of approaching the preservation of software with large-scale social dependencies.

Fondé en 2020, le Collectif Obèle est une initiative de recherche-création qui effectue des performances littéraires dans des environnements vidéoludiques. S'inscrivant dans la pratique des arts littéraires, les initiatives du Collectif Obèle peuvent être décrits comme une forme de littérature numérique en ceci qu'ils utilisent des jeux vidéo comme matière, moteur et méthode d'écriture de textes générant un discours au sein de et sur ceux-ci.

La démarche de recherche-création adoptée par le Collectif Obèle emploie une approche nommée transécriture, où le processus de traduction intersémiotique entre deux médias porte attention aux codes et attributs formels des deux médias mis en dialogue de manière à engendrer une réflexion sur les intersections entre ceux-ci. Pour chaque projet, ce travail de réflexion, qui peut convoquer des notions issues de différentes disciplines (études littéraires, études médiatiques, études des jeux vidéo), aboutit à la production d'artefacts au sein des jeux sur lesquels les membres du Colelctif jettent leur dévolu, accompagné d'un appareillage critique prenant la forme de communication dans des colloques et de publications scientifiques.

Tout au long de la démarche de recherche-création, un travail rigoureux et circonspect d'écriture est effectué par les membres du Collectif Obèle. Épousant l'approche des « cycles heuristiques » de Louis-Claude Paquin, le processus, qui alterne entre phases d'idéation et de prototypage jusqu'à l'obtention d'un résultat final, est documenté au sein de plusieurs formulaires correspondant à des phases de travail et d'élaboration du projet.

Ma présentation accomplira trois objectifs :

- Dans un premier temps, elle présentera le concept de transécriture, en portant une attention spécifique aux enjeux de tranduction intersémiotique qu'elle soulève en considérant les manières dont les environnements numériques des jeux vidéo dans lesquels se déploient les projets du Collectif Obèle façonnent et conditionnent les propriétés textuelles des discours engendrés par le volet créatif de nos démarches.

• Dans un deuxième temps, elle présentera succinctement la méthodologie de recherche-création employée par le Collectif Obèle depuis ses débuts afin d'expliquer en quoi l'écosystème numérique dans lequel elle est effectuée favorise et cloisonne la collaboration entre membres
• Dans un troisième temps, quelques exemples de résultats des démarches du Collectif Obèle seront présentés.

N.B. En complément de cette communication, une proposition de démonstration des projets du Collectif Obèle lors du congrès CSDH/SCHN a également été soumise

(Paper 1) From Repository to Database and Back: Distributed Authority in Ian Lancashire’s Online Libraries

Ian Lancashire produced three main repositories of texts: UTEL (the University of Toronto English Library), RPO (Representative Poetry Online), and LEME (the Lexicons of Early Modern English). All three repositories came into being partly, if not wholly, because of Lancashire’s academic generosity: allowing the world to access important works of English literature and the history of the English language. At a late stage of UTEL’s development, Lancashire and his assistants envisioned the site as “our English Department’s Web site” (Douglas et al, 1999, 18) and an on-line simulacrum of the department and its teaching and research goals (18). The authority of the web site is an extension of the authority of the scholar. By linking to online syllabi for English literature courses taught at the university, the authority of the web site becomes distributed rather than residing in one web master. RPO’s development shows a tension between three motivations: being true to the original goals of Representative Poetry in making available the best English-language poetry, representing multiple English language world cultures through their poetry, and representing Lancashire’s own particular tastes in poetry. Authority here shifts from the pre-post-modernist canon to inclusivity and personal tastes. LEME grew from being a digitization of the important dictionaries of the Early-Modern period to being a digital repository of all English dictionaries, dual-language dictionaries, and assorted lists of words from the period. The authority of the prominent lexicographers is joined with idiosyncratic collectors of words. In all three cases, the authority inherent in the original project is relaxed to become much more catholic in nature. While UTEL was and remained a repository of texts and links, RPO transformed from a repository of texts to “a dynamic website built upon a relational database wherein all the poetic data and metadata are stored” (Plamondon, 2012, 2). RPO was designed to be a database for the digital analysis of the poems, in addition to a site for reading, studying, and enjoying excellent poetry. LEME was constructed on similar principles, where the texts are not just available for reading and searching, but also for integration with text analysis tools. The authority and stability of the text gives way through the database design to the authority of the poetic line and of the lexicon word entry. These are the building blocks upon which RPO and LEME were built. With the most recent edition of RPO and the second version of LEME, these building blocks have been hidden. The authority returns to the texts. The sites, while (potentially) still built on relational databases, have obscured most of the functionality that comes with taking advantage of the relational database foundation. The databases designed for reading, searching, and integration with text analysis tools have returned to being repositories of texts. This paper will trace the shifts in authority in RPO and LEME, arguing that the authority of the repository obscures and neglects the value inherent in the distributed authority contained in their underlying database designs.

(Paper 2) In their own words: Ian Lancashire, Cognitive Stylistics, and the Return of the Author

Ian Lancashire (1942-2025) was known to many as an eminent scholar who worked at the intersections of digital humanities and early modern English drama (Dramatic Texts and Records of Britain, 1984) and historical lexicography (Lexicons of Early Modern England). But another major area of research for Lancashire was Cognitive Stylistics, which Lancashire defined below:

Cognitive Stylistics underscores how every utterance is stamped with signs of its originator, and with the date of its inception…. Cognitive Stylistics determines what those traces may be, using concordances and frequency lists of repeated phenomena and collocations: together, partially, these conceivably map long-term associational memories in the author's mind at the time it uttered the text. … The length of fixed phrases, and the complexity of clusters of those collocations, are important quantitative traits of individuals (Lancashire 2004).

Starting in the early 1990s, Lancashire explored how computer-assisted text analysis could reveal and differentiate not merely the distinctive style of a text, but how it could enable the development of a neuroscientific interpretation of creativity and the cognitive processes of authorship (Lancashire 1993, 1995, 2004). From the beginning, Lancashire’s work suggested that cognitive stylistics marked the return or rebirth in literary studies of ‘the author,’ a personage that had been declared dead by theorists like Roland Barthes. Lancashire’s analyses “search for the author’s brain function in his text. … Is the author who died three hundred years ago truly dead in his work, or can he be partly recognised in them?” (Lancashire 2010). This question is fully explored in Lancashire’s 2010 study, Forgetful Muses: Reading the Author in the Text, which ranges over the works of numerous authors, from Caedmon to Margaret Atwood. Working with computational linguist Graeme Hirst and others, Lancashire went on to study how the author manifests in their writing through symptoms of dementia, analysing and comparing the corpora of Iris Murdoch, Agatha Christie and P. D. James (Le, Lancashire, Hirst & Jokel, 2011). Lancashire later expanded this analysis to include additional authors including Ross Macdonald, known for his hardboiled private eye novels, and children’s writer Enid Blyton (Lancashire 2015).

This presentation will provide a synopsis of Lancashire’s work in cognitive stylistics, with particular focus on the significance of Lancashire’s neurocognitive literary theory and how an author’s vocabulary reveals the creative mechanisms of the individual author.

(Paper 3) (Poetic) Bonds in Primal Testament: A Digital Tools-Based Analysis of Ian Lancashire’s Book-Length Poem

In 2016, Ian Lancashire published Primal Testament, a book-length poem about science, life, and, as perhaps all poems are about, language. One of the features of the poem is how it plays with the letters CGAT, the letters standardly used to represent the nucleobases that are the core constituents of the DNA molecule. Lancashire’s playful use of the CGAT letters within the poem invites, at the very least, a statistical analysis of the use of these and perhaps other letters throughout the poem. Lancashire’s playfulness with language can be extended to an analysis of his poem. Stéfan Sinclair (2003) advocates for a playfulness with the use of digital tools when approaching a text: “I am suggesting that play is an integral part of a humanist’s interpretive activities, and it should not be neglected in the design of our tools” (181). This paper will present the results of a playful, digital-tools-based investigation into the textual components of the poem. One can theorize that the lines of Lancashire’s poem represent base pairs in a DNA strand. A count of the frequencies of the occurrences of the letters CGAT at the start and at the end of the lines should possibly be higher than one might usually expect. A tabulation of the frequencies reveals that the letters A, T, and C do indeed occur more frequently at the start of the lines, while G has only a minimal increased presence. The letters G and T occur much less frequently than expected at the ends of the lines. These comparisons are based on Lancashire’s edition of T. S. Eliot’s The Waste Land (on Representative Poetry Online). While Lancashire seems to be downplaying the importance of guanine in his poem, the letter G does occur moderately more frequently than it does in a large sample corpus of English writing. The letter T is even more prominent in Lancashire’s poem, while the frequencies of appearance of the letters A and C are not much different than the large sample corpus. This paper will also examine the positioning of the letters within individual lines of the poem, theorizing about how the letters of the lines are rooted in the use of CGAT and held together by a poetic equivalent of electromagnetic bonds. Comparisons of data about Lancashire’s poem to similar data from other poems, in particular poems on RPO, will playfully suggest ties to and breaks from the models Lancashire is following, such as the poems of Milton, T. S. Eliot, and Yeats. Ultimately, the main goal of this paper is to draw attention to and honour the poetic creativity of Ian Lancashire using some digital text analysis tools with which he worked and which he appreciated.

(Works Cited ommitted due to length restrictions.)