In digital spaces, language and identity are always already translated through technical infrastructures that reduce, re-map, and re-represent cultural and social categories. While this is often read as a form of mediation, it can be studied as a form of translation, insofar as both the form and the content undergo a shift in the digital publics. This paper positions the enduring misapplication of the term “Indian” to describe diverse South Asian communities within Canada as an instance of the untranslatable in digital racialisation, a form of reduction that both reflects and reproduces epistemic injustice. Drawing on key frameworks from epistemic injustice scholarship (Fricker 2007; Dotson 2014; Mills 1997), the paper investigates how online systems and discourse practices encode, circulate, and sustain a flattened racial category that resists meaningful differentiation even when such nuance is essential to self-identification and ethical representation.

Be it social media or large language models, “marketed as “artificial intelligence”,” as mentioned in the CFP for the conference, digital platforms do not only translate identity terms. Instead, they model them as per the representational affordances of databases, machine learning classifiers, and content tagging regimes. In this process, culturally specific and historically situated markers of nationality, ethnicity, language, and religion are often collapsed into simplified categories that algorithms can process, index, or monetize. The misused label “Indian,” deployed as a catch-all for people of Pakistani, Bangladeshi, Nepali, and sometimes West Asian origin in Canadian online discourse, exemplifies a wider problem of digital translation: the impossibility of one-to-one equivalence between lived identities and their rendered digital forms.

From an epistemic injustice perspective, this digital racialisation produces two intertwined harms. First, it fosters hermeneutical injustice, where the structural absence of adequate interpretive resources, both in culture and in computational regimes, prevents accurate representation of how South Asian subjects understand and name themselves. The label “Indian” thereby becomes not only misapplied but unrepresentative, epistemically occluding the specificities of distinct diasporic trajectories. Second, the phenomenon enacts testimonial injustice, as the self-descriptions and counter-narratives put forward by those affected are frequently dismissed, ignored, or overwritten in online and algorithmically mediated contexts. What is at stake here is not merely a terminological disagreement but an ethical and epistemological problem about who gets to define the terms of recognition in digital spaces, and how technologies shape those terms of recognition.

This paper situates the mislabelling of South Asians within broader discussions about the limits (of translation) in digital humanities work. It underscores how digital infrastructures mediate meaning in ways that are not neutral– they embody specific epistemic assumptions about racial categories, autonomy of interpretation, and the legitimacy of certain knowers over others. By interrogating how the category “Indian” has become both algorithmically reinforced and socially circulated within Canadian digital environments, this paper attends to the untranslatable as a critical concept, one that reveals where digital systems fail to account for cultural specificity, and where they produce harms that extend beyond the digital into embodied social realities.

This project proposes a pilot study that explores how computational sentiment analysis struggles when it is applied to literary texts that move across languages. Using Simin Daneshvar’s short story "The Accident" as an example, the study asks what happens when a Persian narrative and its published English translation are processed through sentiment analysis tools that classify text by emotional polarity. The guiding hypothesis is that Persian prose often conveys feeling through cultural and relational registers such as shame, ta’arof, fatalism, and domestic irony. These forms of affect do not always map cleanly onto English and can be flattened, softened, or redirected into different emotional categories when translated.

Although these shifts may appear subtle at the level of individual phrases, they can collectively alter the emotional trajectory of a narrative when read by a computational system. For example, expressions of moral pressure or humiliation in the Persian original may appear as neutral or even positive in English, while certain kinds of irony or social critique become bluntly negative. This paper does not present final results but instead outlines a method for aligning parallel passages, running comparative sentiment analysis, and reflecting on the interpretive challenges that emerge from multilingual literary corpora.

By foregrounding translation and cultural specificity, the project raises broader questions about what becomes untranslatable when literary feeling meets computational models. It suggests that sentiment analysis assumes a universal and directly legible emotional vocabulary, which may not hold across languages or cultural contexts. The goal of this pilot is not only to test a tool but also to reflect on how computational methods shape what becomes visible or invisible in world literature.

In the current scenario, digital mapping has been increasingly making a successive way to visualize, analyze and interpret spatial dimensions of cultural, historical, and literary data. Within literary food studies, such practices often focus on what can be named, localized and categorised — ingredients, recipes, markets or historical routes. Yet in everyday food practices, especially in a city like Kolkata, the meaning often resides elsewhere: in taste remembered rather than consumed, in smells that recollect childhood memories, festivals or seasonal change. This paper focuses on how such sensory and emotional dimensions of food remain “untranslatable” within digital realms, not due to technological limitation, but because it is experiential, embodied and culturally specific.

By theorizing Barbara Cassin’s idea of the “untranslatable”, it projects not just as a barrier but as something that demands continuous negotiation — the paper argues that taste and smell operate as forms of cultural knowledge that resist direct digitization. Food such as ilish machh, panta bhat and fuchka — these food items are deeply embedded in personal and collective memory in Kolkata. They are deeply tied to anticipation, comfort, belonging and their meanings shift across seasons, life stages and social contexts. By translating these dishes simply as data points or culinary categories, the proximity to overlook the emotional attachment through which they are known or remembered gets pretty close and hence blurs the physical attributes.

The paper aims to present a digital food mapping project that approaches sensory knowledge indirectly, by treating untranslatability as a methodological condition rather than a technical issue. Instead of attempting to encode taste and smell, the map layers place-based food data with short literary excerpts, oral memory narratives and affective descriptions gathered from Bengali and Indian English texts as well as everyday recollections. Though these narrative and emotional annotations do not aim to recreate sensory experience, rather they produce gestures towards it, allowing individuals to encounter food through stories, associations and spatial proximity.

As far as the theoretical framework is concerned, the paper brings together Cassin’s work on untranslatability, memory studies and sensory anthropology to frame food as embodied knowledge that exceeds both language and computation. The concept of humanistic visualization by Johanna Drucker further emphasizes interpretation, partiality and reflective modelling over claims of completeness or objectivity. From this perspective, the absence of direct sensory translation is not a failure but an invitation to show how meaning travels through narrative, emotion and place.

Moreover, the paper reflects on the responsibilities involved in digitally representing lived food cultures — particularly paying attention to bilingualism, cultural specificity and ethical knowledge representation. By focusing on food cultures in Kolkata, the paper demonstrates how digital mapping can remain analytically rigorous while still honouring what cannot be fully rendered, translated or visualised. In doing so, it proposes a model for sensory digital humanities that values emotional attachment and memory as central, rather than peripheral forms of knowledge.