About the projects
Writing across Languages: Computational Methods for the Study of Multilingual Literary Texts
A range of recent Anglophone novels are located at the edge of English, fostering connections to other languages such as French, Hindi, Italian, Kiswahili, Mandarin or Spanish. The project seeks to introduce students in the Master’s degree programmes Comparative Studies and Literary Translation to the potentials of approaching multilingual narratives by employing computer-based digital humanities methods derived from corpus linguistics and natural language processing. Availing itself of HHU’s various e-learning services and tools, the project provides a hands-on experience of analysing and translating passages of selected post-monolingual Anglophone novels in an immersive online environment. We will make use of interactive video tutorials, complementary quizzes and an academic blog.
Three interactive video tutorials will be created to acquaint students with computational literary studies in general and the usefulness of ANNIS, an open-source corpus linguistics software, for analysing and translating multilingual literature in particular. An instance of ANNIS has already been hosted and launched at a HHU server: annis.slam.phil.hhu.de. Going beyond available ANNIS tutorials, interactive content with H5P will be developed to examine specific literary examples, stressing that digital tools not only facilitate the analysis and translation of literary texts but also reinforce linguistic (and other) power hierarchies. Proceeding from this critical position, the videos will illustrate how ANNIS can be used to (1) convert passages of selected literary texts into linguistics formats to enable annotations on distinct linguistic levels and, in effect, the creation of a unique dataset that conforms to copyright laws; (2) scrutinise this dataset for different types of multilingual narration; (3) transform this dataset into a parallel corpus to study translational phenomena. To assess the effectiveness of the interactive video tutorials and prepare students‘ independent study of literary examples, three complementary quizzes will be developed that allow students to remember, apply and critically reflect upon the video content. Furthermore, and attesting to the project’s experimental and collaborative design, students will discuss their experiences with the application of corpus linguistics software in literary and translation studies contexts in an academic blog.
In the future, the vision is to develop a follow-up project that moves beyond textual analysis and creating empirical data on how readers of varying degrees of bilingualism handle and understand post-monolingual Anglophone novels.
AI Literacy for Machine Translation: Critical Reflection of Biases and Parallel Corpora for Multilingual Literary Texts
Our project seeks to familiarise students with the various pitfalls of employing digital tools to translate multilingual literatures. Offering hands-on experiences of the social, political and cultural implications of machine translation (MT), it fosters students’ AI literacy by engaging them in 1) computer-based translation practices that encourage critical reflection on the manifold biases of those LLMs that inform translation software such as ChatGPT, DeepL or Google Translate and b) the creation of parallel corpora that set a new benchmark for digital translation in multilingual literary environments.
AI will likely continue to bring about large-scale societal changes, including the transformation of research and teaching in the humanities or the translation industry, in the years to come. Approaching the numerous risks of AI from the combined perspectives of literary, comparative and translation studies as well as the digital humanities, our project imparts technical as well as conceptual knowledge about translation software, develops critical viewpoints on its social, political and cultural ramifications and devises approaches that serve to de-bias computational translation practices.
To prove how multilingual fiction can serve to evaluate and pluralize MT, our project builds on and further develops the literary corpus that was created within our first ELFF project. It thus assumes that a range of contemporary Anglophone novels yields the potential to develop a critical awareness about the power hierarchies that structure digital translation and to diversify MT. Accordingly, our project pursues two central objectives:
Firstly, it realises the necessity to engage students in the practical application of translation tools and the critical discussion of AI-generated contents. Apart from understanding the AI ‘black box’, experimenting with different software helps students recognise how computer-based approaches to translation reinforce a monolingual literary norm and thus undermine the epistemic potential of the novels. Drawing on recent approaches in postcolonial and multilingual digital humanities, we believe that the experience of failure provides a necessary step towards a kind of AI literacy which conceptualises the ‘digital’ not as immaterial but as an economic system generative of multi-layered power imbalances.
Secondly, our project realises the potential to involve students in transforming the digital archive and, in effect, creating knowledge diversity. Seeking to refashion digital translation as a responsible and sustainable practice, we support our students in the creation of parallel corpora. Based on a broad selection of multilingual passages from the novels and their human translations, parallel corpora advance the development of multilingual language models, setting a new benchmark for MT in the context of multilingual fiction.