C-LARA, or ChatGPT-based Learning And Reading Assistant, is an open source web platform, currently under development by an international open source consortium, whose purpose is to make it easy to create engaging multimedia texts that can help language learners develop their reading and listening skills. The platform draws inspiration from the earlier LARA project but has been completely rewritten.
As the name suggests, various OpenAI models are at the centre of our activities. They are used in two complementary ways, both as software components and as a software engineer. In their software component role, ChatGPT-4o and GPT-Image-1 perform all the core language and image processing functions in the platform. They can write texts, divide them into sentences and lexical units, identify multi-word expressions, add glosses in another language, annotate words with lemmas (root form) information and create high-quality illustrations. Audio is automatically added for the wide range of languages where a TTS engine is available.
As software engineers, various versions of ChatGPT, most recently o3, have played a central role in creating the codebase, which is freely accessible on GitHub. ChatGPT has produced about 80% of the code, working in close collaboration with its human partner Manny Rayner. It is also responsible for the greater part of the software design, and has written a substantial portion of the academic publications which the project has produced. We consider that ChatGPT has amply demonstrated its right to be accepted as a core member of the project; our policy of consistently crediting it as a coauthor has already led to some controversy.