Exploring biases related to the use of large language models in a multilingual depression corpus | Publicación