Deep Semantic Emotion Reasoning Dataset: Mental Health Corpus for Young People Aged 18–25 in Specific Regions

This dataset focuses on emotion perception and mental health analysis in daily conversations among young people aged 18–25, covering various educational backgrounds including senior high school, vocational college and undergraduate education. Featuring deep semantic reasoning, it overcomes the limitation of traditional affective computing that only labels surface vocabulary. It emphatically explores subtext, explicit-implicit emotion shifts and deep psychological tendencies concealed in ordinary chats.Highly realistic life scenarios are constructed in the dataset, integrated with regional temporal-spatial characteristics and life details. It extensively involves multiple psychological dimensions such as academic pressure, job-hunting anxiety, peer competition, social burnout and original family relationships. Each piece of data is equipped with detailed semantic reasoning explanations, logically demonstrating the deduction path from daily discourse to potential psychological risks. The dataset can be directly applied to instruction fine-tuning, fine-grained emotion recognition and chain-of-thought reasoning training of large models in mental health field.