"Dreaddit"

"Stress is increasingly observable on social media platforms and interactions, yet most research focuses on short-form content like Twitter or specific modalities, like speech. The Dreaddit dataset, introduced by Columbia University researchers, addresses this by targeting lengthy, multi-domain narratives. Comprising over 190,000 Reddit posts across five categories\u2014Abuse, Social, Anxiety, PTSD, and Financial\u2014it includes a high-quality subset of 3,500 human-annotated segments for binary stress classification. Unlike simple keyword matching, Dreaddit enables models to capture nuanced, complex emotional expressions in long-form text. It serves as a vital benchmark for training deep learning systems like StressAI to achieve superior contextual and semantic understanding."