Interpretable LLM–Based Detection of Loose Associations Using Synthetic Speech Data in Early Psychosis | Publicación