This work is framed in the vocabulary generated by large language models such as ChatGPT, and specifically, in the one specialized in Spanish as a foreign language. For this purpose, two large language models, ChatGPT and LlaMa, have been asked to generate written comprehension texts for students of Spanish level from A1 to B2. For this purpose, it is necessary to use a specific input that generates the closest response to our purpose. After the request, the samples have been analyzed following the references of the specific notions section of the Instituto Cervantes Curricular Plan. Likewise, a comparison has been made between both models, hoping to find differences between the answers provided. With this analysis, we wanted to check whether the language models can be useful when preparing materials and, in particular, whether the textual samples generated follow the standards of the Cervantes Institute's Curriculum Plan with respect to vocabulary.
Read full abstract