This paper presents the conceptual design and early implementation of a three-year, community-led digital heritage initiative to locate, preserve, and ethically steward a dispersed corpus of Uyghur-language ethnographic materials, many collected by or under the direction of Professor Rahile Dawut, a leading Uyghur ethnographer now serving a life sentence as a political prisoner in China.

These recordings, photographs, notes, and manuscripts are not merely “data”: they are living traces of a vibrant cultural world—music, storytelling, ritual practice, and local knowledge—now subject to systematic repression. The collection arrives to the archive already shaped by displacement, risk, and fragmentary custody: cassettes, laptops, hard drives, and paper held across multiple countries, often in private hands. Designing an archive at a distance is not a temporary constraint but a methodological condition: the repository must be built for diaspora stewardship under surveillance, while resisting familiar digital-humanities impulses to render cultural materials open and universally reusable.

Anchored at the University of British Columbia through the Collaborative Digital Heritage Studio and the university library, the project is co-designed with Uyghur Elders, cultural practitioners, and scholars in the diaspora—especially in Istanbul, home to one of the largest Uyghur communities outside China. The archive is being developed as a secure, multilingual and multi-script digital repository in Uyghur, English, and Chinese, with support for Arabic, Latin, Cyrillic, Chagatai, and Old Uyghur scripts. Its aims are to safeguard fragile materials against loss, and to support cultural continuity by enabling community-led education, language work, and revitalisation—quiet forms of political resistance in the current climate.

Aligned with the CSDH/SCHN theme of the “untranslatable,” this paper argues that untranslatability in DH is not only lexical; it is structural. Every act of digitisation and description is an act of modelling—through metadata schemas, controlled vocabularies, and search and retrieval assumptions—that risks flattening culturally specific relations into ostensibly neutral categories. Multi-script Uyghur materials make this unavoidable: orthographic variation, competing romanisations, contested place-names, and genre terms that carry social histories do not resolve neatly into one-to-one equivalences. Following Barbara Cassin’s provocation, we treat translation as iterative and contingent. In practice, this means documenting translation decisions, preserving ambiguity where it is meaningful, and designing for plural and contested relations rather than forcing a single authoritative label.

Methodologically, the paper draws on two decades of building “archives at a distance” through the Digital Himalaya Project and the World Oral Literature Project, where questions of custodianship, representational authority, and long-term sustainability have repeatedly converged with the pragmatics of digitisation, metadata, and platform design. Empirically, I focus on two main issues: (1) the politics of metadata (who names, who classifies, and which epistemologies become legible), and (2) access and reuse in an era of AI (tiered permissions, community review, and protections against extractive republication and LLM training). Our hope is that the archive will offer a transferable model for endangered-language and cultural heritage projects working under conditions of authoritarianism and forced migration: an infrastructure that is both searchable and durable, and yet at times strategically untranslatable.

The “untranslatable” in DH is frequently approached as a technical challenge or a problem of material limitation. This paper argues that such framings are insufficient. Instead, it contends that untranslatability in DH is systemic and epistemic, produced by the “coloniality of knowledge” embedded in the field’s foundational epistemic cultures, methods, and disciplinary norms (Quijano, 2000; Mignolo, 2011). Drawing on decolonial theory and the concept of “epistemicide” (Santos, 2014), the paper argues that untranslatability emerges from historical conditions that have marginalized, subordinated, or erased certain ways of knowing, conditions that continue to shape contemporary digital knowledge production.

Rather than arising solely from the inherent qualities of cultural objects or from technological inadequacy, untranslatability is shown here to be conditioned by epistemological assumptions that determine what qualifies as data, which cultural forms are rendered legible, and which methods are institutionally validated and funded. DH, shaped in its early formation by computational linguistics and text-centered literary analysis, has canonized technologies and practices optimized for textual corpora, statistical modeling, and linguistic abstraction. These epistemic commitments privilege writing as the primary vehicle of knowledge and, in doing so, structure the limits of what can be translated into digital form.

To develop this argument, the paper draws a comparative analogy with Translation Studies, a field that has long grappled with the limits of text-centered theories when applied beyond interlingual translation. As Translation Studies scholars have shown, theoretical frameworks developed for text-to-text translation often falter when extended to intersemiotic, performative, or multimodal practices. A similar dynamic operates in DH: methods designed for textual analysis are frequently extended to oral traditions, Indigenous knowledges, situated linguistic practices, and relational epistemologies, producing forms of untranslatability that are epistemic rather than technical in nature.

Building on this parallel, the paper reinterprets Lawrence Venuti’s concept of “foreignization” (Venuti, 1995) as a model for epistemic resistance within DH. While foreignization in translation resists cultural domestication by preserving linguistic difference and exerting what Venuti terms “resistancy” against dominant target-culture values, this paper proposes the concept of “epistemic foreignization” in DH to describe practices that deliberately preserve untranslatable residues rather than assimilating them into fluent, extractive, or normalized digital forms. In this framework, untranslatability functions not as a failure of representation but as an epistemic boundary that exposes the limits of dominant DH methodologies.

Situating this discussion within the Canadian context, the paper engages debates in Indigenous DH and ethical knowledge stewardship. It argues that the geopolitics of knowledge production often positions Indigenous and non-Western epistemologies as local, derivative, or supplemental rather than as foundational sources of theory. In this context, untranslatability can be understood as a form of epistemic refusal, a resistance to abstraction, extraction, and epistemic violence enacted through digital modeling.

Ultimately, this paper reframes untranslatability as a necessary site of critique in DH. By foregrounding epistemology and the historical processes of epistemicide that shape technological development, it calls for a decolonial reorientation of DH that recognizes that not all forms of knowledge should, or can, be translated into digital form.