Acceso usuarios
Qian Li, Sven Tarp 
, Antoni Nomdedeu Rull
This paper reports on an ongoing research project aimed at developing a new type of Spanish learner’s grammar, different from those found in textbooks, grammar books and dictionaries. The new grammar, designed to be displayed in digital writing assistants, will explain problems that occur in written learner texts. The paper first describes the main features and functionalities of this grammar and how it will be presented to Spanish learners. It then discusses the development of a methodology for categorising relevant error types, using a unique combination of existing grammars, dictionaries and ChatGPT, all of it supervised by lexicographers with experience in language teaching. Based on this categorisation, the paper explains how the chatbot is prompted to write explanations of the different error types, which it does very well in fruitful interaction with the human lexicographers. The methodology is described in detail with several examples. Finally, the paper explains how the original Spanish explanations are machine translated into English and Chinese, and provides examples of the final result in each language. Throughout the paper, the complex relationship between generative AI and humans is discussed, and it is concluded that a successful result like the one achieved requires both the ability to handle the chatbot properly and the knowledge of the topic being dealt with.
