| Resumen |
Hope speech includes messages that express optimism, encouragement, or the desire for a better future. This paper provides an overview of the PolyHope shared task at IberLEF 2025. This task aims to detect hope speech in social media posts written in English and Spanish. The PolyHope task is divided into two parts. The first part is binary classification, where systems decide if a post expresses hope or not. The second part is a more detailed classification into five categories: Generalized Hope, Realistic Hope, Unrealistic Hope, Sarcasm, and Not Hope. To support this task, a new dataset called PolyHope V2 was created. A key feature of the dataset is that it includes sarcastic hope as a separate category. This allows models to better detect when hopeful language is used in a sarcastic or misleading way. Several research teams joined the task and tested different models, including traditional machine learning, deep learning, and transformer-based systems. The paper presents findings from the shared task, including best-performing methods and common challenges faced by teams such as imbalanced data, language mixing, and cultural differences in how people express emotions. |
