Online translation is a gift but it is also a problem. The translated text replaces the source text immediately, as though there were no process of interpretation to be gone through, no choices to be made, no responsibility to be taken, nothing to think about. This happens invisibly when material is gathered from different languages for an AI overview, or when news or other content is automatically translated to suit your settings; often it is also how translation is done using Large Language Model chatbots like ChatGPT, DeepSeek or Claude.
But Large Language Models in fact have the ability to do something much more interesting – or rather, to help you do it. They can open up the variety of language for you to investigate, and help you to see what is at stake in translating one way rather than another. The Translalia app is an experimental tool for exploring this potential. It asks you to think about the kind of language you are translating out of, and the kind of language you want to translate into (translations don’t have to be in standardised languages but can be in any dialect, idiom or kind of languaging). And it shows you that there are always different ways a text can be remade in whatever kind of language you choose to use. Translalia makes visible the complexity that is hidden by most online translation. It gives you time and space to engage in translation as a process, keeping your thoughts, feelings and imagination at the heart of it.
Translalia has emerged from the AI, Decoloniality and Creative Poetry Translation Project; it is currently being trialled with schools and other groups in Agrentina, India and the UK. You are free to try out Translalia if you are interested; but please note it is an experimental tool, not a product: it is not operated by Oxford University. Anything you enter into Translalia remains private to you and is covered by GDPR; you can easily delete all your material at any time. In its current form, Translalia works only with OpenAI’s Large Language Models – you can choose which one in the ‘Translation Model’ section of the app. We would love to get your feedback on Translalia: if you would like to leave some you can do so here.
Guidance notes for Translalia are here.
An introductory video is here.
Translalia is here.
Translalia is the creation of Matthew Reynolds and the AI, Decoloniality and Creative Poetry Translation Project (participants: Joseph Hankinson, Deepshikha Behera, Luciana Beroiz, Yingxin Chen, Karen Cresci, Annmarie Drury, Sarah Ekdawi, Vani Nautiyal, Mary Newman, Wen-Chin Ouyang, Rafaa Ragebi, María Eugenia Rigane, Xuemeng Zhang). The idea of making an app came from Lachlan Kermode. The process was supported by Dominik Lukes and Oxford University's AI and ML Competency Centre. The coding was done by Raj Trivedi.
