Chapter 9 Understanding Multilingualism in Digital Humanities Infrastructures

Multilingualism Design in DH Infrastructures

A critical infrastructural perspective on multilingualism in the DH is fundamental, given the rapid spread of language technologies and research infrastructures dedicated to language resources in and around the DH community in the last few years. While many of these technical developments are welcome (and in some cases essential) for filling huge gaps in multilingual support, they have sometimes overshadowed the need for a more balanced (critical-social) evaluation of the relationship between DH, multilingualism, and infrastructure.

The field of critical infrastructure studies has expanded in recent years as a site for exploring global culture from the perspective of sociotechnical infrastructures, and this has increasingly been applied to study research infrastructures in the humanities and social sciences (and culture-based knowledge infrastructures more broadly) (Critical Infrastructure Studies.org).10 Alan Liu appraises the potential role of DH in treating infrastructure as an object of study. Noting the “convergence between infrastructure and culture for humanistic critique,” he argues that DH is “uniquely placed to interpret and critique culture at the level of infrastructure” (Liu, 2–4), and here I propose that this infrastructural critique urgently needs to be extended to the multilingual plane.

Whereas critical studies have devoted considerable attention to diversity in digital infrastructures in recent years (McPherson; Noble; Ricaurte), critical engagement with geocultural/multilingual diversity in infrastructure studies has been relatively scarce up to now. Work by authors such as P. P. Sneha has provided a welcome corrective to Global North–slanted studies on infrastructural diversity (Sneha), while David Wrisley (in exploring scholarly reactions in the Arab world to open scholarship models) has called for us to move beyond a narrow focus on European/North American models of knowledge creation and to engage more fully with “global infrastructural difference” (Wrisley).

Part of the overall challenge here is that in countries where digital infrastructure is designed and developed, multilingualism is roundly marginalized in public and academic discourse, and scholarly or professional fields addressing digital multilingualism are spread across radically diverse research networks with weak mutual engagement. The recent and considerable increase in attention to multilingualism and language diversity has led DH to contest the assumed normative nature of monolingualism, especially in countries with influential major digital industries such as the United States, where monolingual ideology is pervasive. Nevertheless, multilingual DH research in an anglophone context has attended to monolingual (and especially anglophone) bias in DH research infrastructures on largely practical grounds.

What different models exist for approaching multilingual challenges in DH research infrastructures, and what do they tell us about how the field views multilingualism? The next sections explore multilingual DH from four perspectives: (1) as access to language resources, (2) as literacy and ideation, (3) as translation, and (4) as tactical response.

Multilingualism and Language Resources

Common Language Resources and Technology Infrastructure (CLARIN) describes itself as the “research infrastructure for language as social and cultural data”11 and is one of the best-known infrastructural responses to language diversity in digital scholarship. It primarily consists of datasets, tools, and services developed by corpus and computational linguists to support research in the humanities and social sciences and is based on the model of a “single sign-on online environment” to facilitate the use of “digital language resources and tools from all over Europe and beyond” (CLARIN, “About CLARIN”).12

Rather than creating new resources centrally, CLARIN integrates access to resources generated either by national constituent organizations or elsewhere. Like other European research infrastructures, CLARIN operates in a highly multilingual space (the European research community) with a strong community identity, bolstered by regional policies and strategies that firmly promote multilingualism. While the general operating language (what Andreas Witt calls the “meta language”) is English, each national participating consortium makes its own choice regarding operating language (the German and Croatian consortia use both English and their respective national languages, for example; CLARIN, “CLARIN Participating Consortia”).13 In creating a networked federation of language data, service, or knowledge expertise centers, its choices with regard to language support are to some extent influenced by a combination of national funding priorities and national/regional community research interests and priorities.

Despite having a clear European bias—in the top ten languages listed at its Virtual Language Observatory,14 the only non-European ones are Japanese (eighth), Chinese (ninth), and, depending on our classification, Turkish (tenth)—CLARIN is open to all languages, and the CLARIN model has been an important reference point for research infrastructures beyond Europe, such as the South African Centre for Digital Language Resources (SADiLAR).15 Although there is no formal strategic directive regarding language choice, CLARIN does have knowledge centers for individual languages, language families, or other language groupings, as well as Knowledge Centres for linguistic diversity and language documentation.16 In addition, its Spanish K-Centre aims to offer language resources in four co-official languages of Spain: Castilian (Spanish), Catalan, Basque, and Galician, which points to growing awareness of the specific needs of lower-resourced languages.

The scale and variety of resources that CLARIN makes available clearly benefits language diversity in digital research practices in the humanities and social sciences, and it provides a valuable model for multilingual resource exchange and collaboration between language communities in DH.

Multilingual Literacy and Ideation

“Language is . . . always plural, always a place of difference,” argues Polezzi in her introduction to the concept of “language indifference” (Polezzi). Normalized monolingualism tends to render multilingual work invisible or to treat it as something narrowly technical or culture-free, and the challenge for multilingual DH has been to expand debate and action beyond the very valuable work being carried out in linguistics and language technologies. Brandão and I, in the previously mentioned publication, explore frameworks that DH might use to foster “a languages-centric agenda for DH” (Spence and Brandão), while Aliz Horváth emphasizes “sensitivity to multilingualism as an overarching concept,” which moreover needs to move DH infrastructure ideation beyond a Western perspective (Horváth, 1). Here, I wish to expand on how various DH infrastructural initiatives have fostered multilingual literacy and ideation.

Projects have attended to a broad range of multilingual challenges in DH in recent years, including historical and multilingual optical character recognition (OCR; Smith and Cordell), right-to-left language scripts,17 and multilingual NLP. The Multilingual DH network launched by Quinn Dombrowski in 2019 has served to aggregate non-English language resources (NLP and OCR resources in more than twenty-five historic and modern languages or language families), and it is also one of many efforts to disseminate methods-based DH resources across languages.18 Similarly, the New Languages for NLP workshops in 2021/2022 both trained researchers to create data and language models for low-resource (Quechua or Kannada) and historic languages (Classical Arabic or Old Chinese) and drew attention to the pitfalls of generic approaches (such as browser support limitations or the dangers of making assumptions about character/word/sentence boundaries when processing text in different languages). In publishing annotated linguistic datasets in an open repository, the project will also be one of a number of multilingual DH projects making a broader contribution to greater linguistic diversity in NLP.19

One major hurdle facing digital research is how to support NLS languages. At present, there are a number of barriers to carrying out DH research on NLS language content, including OCR quality, discovery in general information systems, and interface usability (Lee and Wagner). DH researchers working with East Asian and Middle Eastern languages and scripts have been among the most active in this area, and a series of workshops and projects have contributed to critical reviews of DH infrastructure through the NLS workshops organized by Cosima Wagner and Martin Lee in 2019 (Asef and Wagner), the German DH association’s Special Interest Group Arbeitsgruppe Multilingual DH,20 and the Disrupting Digital Knowledge Infrastructures (DDKI) collective. The DDKI’s plan “to formulate a set of guidelines for universities, libraries, relevant organizations, and developers to consider towards a more language-inclusive digital environment, sensitive to the needs of scholars working with non-Latin scripts” will make an important contribution here (Horváth, 7). More recently, this collective published “Six User Personas for the Multilingual DH Community,” which has helped lay the groundwork for further collective work (Horváth et al.).

Much attention in DH is currently focused on other globally dominant languages, such as Spanish, Arabic, and Chinese. There is far less consideration at present of endangered, minority, or heritage languages, and there is also a need for greater focus in DH on methods and resources operating in spoken forms of language, which is so crucial to languages with no (or little) textual tradition.

Whereas these projects draw attention to different language-mediated mechanics, projects such as Programming Historian (PH) and OpenMethods (OM) spotlight the challenges of modeling multilingual workflows. PH currently offers open-access, peer-reviewed tutorials in digital methods in four languages (English, Spanish, French, and Portuguese) within a geoculturally diverse editorial structure. In the PH multilingual workflow model, none of the operating languages are privileged (original creation can occur in any language), and this also attends to cultural/intralingual variation in the terms used (e.g., avoiding geographic dominance of any Spanish-speaking country/region in both the editorial board composition and policies). Translated content (and, where relevant, examples/data) are adapted to different linguistic and cultural contexts—and each language version has its own International Standard Serial Number (ISSN) and each PH lesson its own digital object identifier (DOI), which foregrounds linguistic or translation labor.

While PH offers a model for multilingual editing/resource creation, the OM platform presents a framework for multilingual research dissemination. OM functions both as a metablog for republishing micropublications, which present DH tools and methods, and as a forum for critical reflection on their use. One of the core aims of OM is to provide greater visibility to the application of DH methods and tools in different linguistic contexts, and its editorial team currently curates content in eleven languages. Each post is still available in its original language, but it is preceded by a short introduction in English, as part of an attempt to improve discovery of non-English content in DH research. The platform has recognized the barriers that this kind of multilingual approach has to overcome—such as lesser availability of non-English content, the slower workflow for non-English content due to fewer available editors who speak those languages, the preference for some language communities to speak English in a global setting, and the continued centralization of English as a lingua franca—but it does at least provide a provisional counterbalance to monolingual distortions within DH scholarly communications. While it was initially developed with European languages in mind, the platform has expanded to cover non-European languages.21

While, taken together, these initiatives still do not represent a holistic vision or strategy for multilingualism, they do characterize an attempt to build and strengthen multilingualism within the fabric of DH at various levels. In a study that I carried out as part of wider research into attitudes toward multilingual DH, one interviewee connected to the Digital Research Infrastructure for the Arts and Humanities (DARIAH)22 highlighted the importance of attending to multilingualism in research infrastructures as a “vital condition for rich and healthy humanities research.”23 This interviewee traced the implications for multilingual DH at various stages of the research ecosystem through various DARIAH collaborations, from its work with Europeana and other cultural heritage organizations (to facilitate access to multilingual source materials), to the OPERAS-P project (which examines and facilitates the conditions for open-access multilingual publishing), to its involvement in the GoTriple discovery service (which connects research data through multilingual vocabularies). DARIAH represents a loose federation model, and one can see how this approach could easily be threaded into a coherent strategy to address multilingual issues throughout the digital scholarly research life cycle.

Multilingual DH as Translation

“Cultural and Linguistic Variety—Transnational RIs,” a section in the 2011 European Science Foundation report “Research Infrastructures in the Digital Humanities,” provides an early (and rare) treatment of multilingualism and intercultural exchange within literature on digital research infrastructures. Recognizing that “theoretically informed, comparative and transnational research gains from access to large datasets” of diverse linguistic and cultural origin, the report highlights the need for “sensitivity to the need for many-level translations (translation taken in the broad sense of the word)” in infrastructure design (Research Infrastructures in the Digital Humanities, 32–34). What precisely that “broad sense” entails in practice is left to our imagination, but the report does signal the need not only to translate in the linguistic sense, but also to translate between different ontologies, taxonomies, and the different meaning-making cultures that they embody. As Wrisley remarks, “Crossing borders into different knowledge cultures can be confusing business, indeed” (Wrisley), and here translation skills and intercultural literacy play a key role.

How has the concept of translation been conceived within DH? In her examination of the “plurality of DH approaches: what I call a DH ecology of knowledges” (Ortega, 180), Élika Ortega builds on Mary Louise Pratt’s term “zones of contact” to argue for the importance of “translation work” in defining the connections among different geolinguistic communities. Recognizing DH as “a horizontal exchange of referents and nonfixed positions in regard to one another” rather than a “unitary” model, she describes projects such as RedHD in Translation and DH Whisperers, each being a tactical intervention to disrupt anglophone monolingualism in scholarly events and publication venues (Ortega, 180–83). She understands these initiatives to be translational not only in a linguistic sense, but also in the sense of “movement,” which renders DH accessible and receptive to nonhegemonic cultural and linguistic communities.

Earlier, we examined how PH provides a possible countermodel for multilingual workflows within DH. Here, I would like to turn to how it models the concept of “translation” within digital research infrastructure. Jennifer Isasi and Antonio Rojas Castro ground their analysis of the Spanish-language edition of the PH platform in translation theory (specifically contrasting the notion of “equivalence” and Skopos theory, which privileges a translation’s purpose) in their discussion of a translation strategy for DH pedagogical materials that facilitates multilingual understanding while also acknowledging multicultural diversity. They present three translation strategies that have been used on the project, which are respectively labeled linguistic (simple translation of equivalent terms, without adapting tutorial design or sample data), expressive (tutorial is partially adapted to a hispanophone audience, but there may still be English-language dependencies in some aspects, such as the tools used), and substantial (major changes in both the tutorial design and the materials used).

While we might wish to explore further the labels and divisions used, at the substantial end of the spectrum, according to their analysis, the material may have been reworked with references and examples designed specifically for the Spanish-language context and may involve reworking entire sections of the original text. This approach embodies the “broader sense” of translation mentioned earlier and helps the reader to better understand/digest the content (in this case, a tutorial) in their own personal linguistic and cultural context. As the authors point out, some tutorials present specific translation challenges due to linguistic differences (which lead to different results in different languages), as well as a lack of appropriate content or tool support for a given language. Theirs is a strategy that recognizes the differences in how cultures operate in digital spaces and what digital traces they leave behind, which in turn determines the effectiveness of a digital method in a particular language (Isasi and Rojas Castro).

Multilingual DH as Tactical Response

In their analysis of “the many languages of digital infrastructures,” Sneha and Sengupta signal the urgency in engaging with “knowledge and infrastructural gaps in order to make the web more multilingual, accessible and safe, particularly for marginalized and non-dominant communities.” Partly based on analysis of the State of the Internet’s Languages report, which they contributed to, they highlight the “content and participation gap,” based on numerous social and technical barriers that inhibit members of non-dominant language community from active participation in knowledge spaces (Sneha and Sengupta). Moving to a DH perspective, Eduard Arriaga cites the work of the historian-anthropologist Michel-Rolph Trouillot as Trouillot assesses the “silences” in DH “objects, tools and infrastructures” produced by design decisions in “architecture and engineering,”24 which we are only starting to engage with critically (Arriaga, 544).

Launched in 2012, GO::DH, a special interest group within the Alliance of Digital Humanities Organizations (ADHO), has played a key role in addressing language diversity in DH as part of its wider mission to promote collaboration among DH researchers across the world. Strongly focused around agile/pop-up projects, advocacy, and community-building, GO::DH has embodied a multilingual and Global South–facing ethos, creating both the Translation Commons and the Translation Toolkit initiatives to foster multilingual collaboration, as well as the DH Whisperers drive to encourage informal multilingual interpretation and interaction at DH conferences. Influential well beyond its limited means, GO::DH has enjoyed success partly due to the expansive, open, low-regulation, and community-driven approach that it has adopted. GO::DH has often embodied an “activist” ethic, a “tactical” response to mainstream DH infrastructures (with their historic anglophone/monolingual and Global North bias). This kind of approach speaks to David Berry’s concept of “tactical infrastructures”—adapted from the concept of tactical media—as “counter-infrastructures,” which, in contrast to the often “instrumental” perspective that DH has taken on infrastructure, enables “new modes of knowing and thinking, assembling and acting” (Berry). On a broader level, we might situate this within a wider concept of “multilingual DH as community,” where a heterogenous block of stakeholders including language activists, modern languages/area studies, and most important of all, DH research communities (operating in their own languages) has become significantly more active and visible within DH in recent years.

The numerous multilingual DH initiatives that have sprouted are starting to address multilingual research challenges in advanced humanities and social sciences research, but they do not generally engage extensively with stakeholder language communities (as opposed to research communities operating within a language). In his study of projects addressing the needs of Afro-Latinx and Afro–Latin American digital culture, Eduard Arriaga assesses the role of infrastructure in consolidating the hegemony of text and explores the representational erasure of marginal communities along racial and ethnic lines, which has often occurred in digital archives and publications. In describing their capacity for “representation and self-representation,” reparation and reclassification, which emphasizes infrastructure as social process, he introduces the concept of “expanded digital infrastructure” as a counterinfrastructure consisting of minimal computing, hybrid technologies, and the “use of digital dynamics based on the humanisation of the historically dehumanized person” (Arriaga, 548, 546). Crucially, the concept of expanded digital infrastructure necessarily integrates the agency and perspectives of both academic and marginal communities, and while not specifically multilingual in nature, the principles that Arriaga proposes are highly relevant for marginal linguistic communities.

How should multilingual action in DH be resourced? Some infrastructural initiatives promoting multilingualism described here enjoy medium- to long-term funding (e.g., European initiatives like CLARIN), but in most cases, they have so far relied on the vision and dedication of a small but committed group of people working on a largely voluntary basis. The social aspect of this multilingual infrastructure is driven by an increasingly coherent policy and research agenda, but it faces complex credit and workload issues, not least because multilingualism typically operates in liminal spaces in the academy. The recent growth of formalized multilingual DH collectives such as the ADHO and DARIAH community groups perhaps also points to a growing realization that (notably set against the long autumn of academic Twitter and convulsions in the wider social mediasphere) some aspects of the tactical approach have limitations, and that multilingual strategy needs to be embedded more deeply into DH infrastructures through policy action, such that the benefits (and costs) of acting multilingually are structurally integrated into their underlying fabric.

DH multilingual initiatives may have a combination of motivations that can influence their success: acting on specific digital research needs (diversity in NLP), a research interest on particular languages, the desire to promote research in particular languages, or a wider commitment to geolinguistic diversity at an activist or policy level (as made by GO::DH). Some projects may be shorter term in principle (to provoke discussions around multilingual exchange at research conferences, as in the DH Whisperers project), whereas other initiatives may have longer-term aspirations (such as multilingual editions of PH).

It is difficult to generalize about such a wide spectrum of multilingual moves in DH, but it is likely that infrastructural initiatives with a well-defined focus (whether that consists of fixed outcomes, such as a set of PH tutorials, or an infrastructure to foster DH research in a particular language) are more likely to endure over time than initiatives with looser commitment to global diversity in general.

There are clearly limits to what a volunteer community can support in an already overcommitted research space, and the challenge going forward will be how DH can operationalize multilingual transformation on a greater scale. In particular, there is a pressing need for research contrasting the incentives and barriers for individual language (DH) communities, including low-resourced ones, because they are the natural drivers of multilingual transformation in digital scholarship. What we know already, though, is that the linguistic hurdles in non-Anglophone DH labor systems are complex. Beyond the common pressure to publish in English (or other dominant language) for so-called global impact, a given DH language community needs to expend extra energy toward articulating its own research identity and culture within wider digital platforms and standards with a strong anglophone imprint. In my interview study, more than one respondent highlighted the obstacles in getting DH research in their language recognized because (in a situation analogous to what happens in linguistics) English-language research is often treated as the “gold standard” for advanced computational studies, due to its access to larger datasets, more developed tools, and more extensive literature.

Tactical approaches to digital multilingualism in DH mirror broader attempts to disrupt digital monolingualism in wider digital studies and practice. DH has much to offer, but also to learn, from wider community-driven and participatory approaches to the challenge. The changing terms of participation by language communities in digital knowledge production are enabling us to question how infrastructures shape and are shaped by language power dynamics, and in particular anglophone digital cultural dominance. It makes us attend to the dangers of infrastructural discourse in DH, which can easily facilitate, even if unconsciously, a kind of monolingual and “center-dominant” platformization where the (here-linguistic) margins passively adopt anglophone/major language epistemological designs. A tactical multilingual response would see DH become a disruptor, rather than a magnifier, of anglophone cultural hegemony and digital monolingualism in an expanded multilingual view of infrastructure.

In his analysis of global knowledge dynamics through social technology, Thomas Petzold notes that these dynamics are “not a matter of scale only” and “we need to examine the underlying cultural dynamics more closely” (64). If “the digital universe is still to be turned, from a global inequality amplifier with pockets of knowledge, into a truly global network of knowledge,” rather than simply adding new features in multiple languages to an anglophone root infrastructure, it should instead be understood as “a problem-solving network that identifies very precise actions to tackle or resolve any specific issue” (Petzold, 136). This chapter has examined how such “problem-solving networks” within DH have understood and addressed multilingual challenges, whether addressing access to language resources, ideating DH’s own multilingual strategies, enabling translation between languages and cultures, or using multilingualism as a tactical response to infrastructural needs.

David Gramling has charted how the “race to multilingualize trade logistics infrastructure has been profoundly rearranging both public discourses about multilingualism and the pathways of global content diffusion by way of multilingual technologies” (5); a charge that DH (and digital studies more broadly) have just begun to contemplate. This obfuscation, or appropriation for commercial interests, of a languages-focused agenda in digital research can easily lead to marginalization of multilingual expertise—in fields such as modern languages or area studies—privileging instead an agenda driven largely (if not solely) by machine translation and technolinguistic solutionism. Linguistic diversity in digital knowledge production is a useful proxy for knowledge creation, broadly speaking, and DH research infrastructures are thus representative of the ongoing conflict in knowledge infrastructures more generally between geolinguistic diversity and the dynamics of dominant language consolidation. Critical multilingual infrastructural studies play an important role in reorienting digital practice toward greater epistemic diversity and cultivating richer intralingual and geocultural dynamics in digital scholarship.

Notes

1. 1. See http://www.meta-net.eu/whitepapers/overview, http://www.dldp.eu/sites/default/files/documents/DLDP_Digital-Language-Survival-Kit.pdf, https://internetlanguages.org/media/pdf-summary/EN-STIL-SummaryReport.pdf, https://zenodo.org/record/5743283.

2. 2. See resources at https://internetlanguages.org/en/resources/.

3. 3. See https://european-language-equality.eu/.

4. 4. See Pawlicka-Deger for discussion on how this relates to relational infrastructure.

5. 5. See https://www.humanisti.ca, https://tthub.io/, http://www.globaloutlookdh.org/executive-board/.

6. 6. See https://arounddh.org/.

7. 7. See https://go-dh.github.io/translation-toolkit/.

8. 8. See https://go-dh.github.io/mincomp/.

9. 9. See https://programminghistorian.org, https://openmethods.dariah.eu.

10. 10. See https://cistudies.org.

11. 11. See https://www.clarin.eu/.

12. 12. See https://www.clarin.eu/content/about-clarin.

13. 13. See https://www.clarin.eu/content/participating-consortia.

14. 14. See https://www.clarin.eu/content/virtual-language-observatory-vlo.

15. 15. See https://www.sadilar.org/.

16. 16. See https://www.clarin.eu/content/knowledge-centres.

17. 17. See https://dhsi.org/course-archive/.

18. 18. See http://multilingualdh.org/.

19. 19. See https://newnlp.princeton.edu/.

20. 20. See https://dig-hum.de/ag-multilingual-dh.

21. 21. See also Horváth. On the epistemological implications of technical standards and the dangers of normative “global” DH practices “eliminating local variants of research,” see Priani Saisó.

22. 22. See https://www.dariah.eu/.

23. 23. The interviews form part of ongoing research into multilingual attitudes within DH that started in 2022. The interviewees (twenty in the first stage) were selected on the basis of their involvement in DH infrastructure design and/or involvement in multilingual DH research.

24. 24. Throughout the chapter, translations from Arriaga’s article have been done by the author.

Bibliography