Facebook pages of leading parties and politicians in Germany, France, Poland and Hungary between 2009 and 2024

MORES is an innovative research project that addresses the challenges faced by liberal democracy, particularly concerning the role of moral emotions, values, and identities in politics. The research and innovation project aims to counter both the detachment of citizens from politics and the overemotionalisation of political debates, recognising the threats these phenomena pose to democratic principles. The project has developed customized, fine-tuned XLM-RoBERTa models that are capable of detecting emotions at the sentence level in political texts. Its training process combines manually labeled data, created through an expert coding system, with synthetic samples generated by OpenAI’s GPT-4-turbo. The current version of the model is available: https://huggingface.co/spaces/poltextlab/emotion_in_text To analyze these questions, data from Facebook pages of leading parties and politicians were downloaded by the CrowdTangle software. The databases contain individual Facebook posts broken down into sentences. For each sentence, a separate column indicates the emotional load estimated by the RoBERTA language models. The four datasets include data from four countries - Hungary, Poland, France and Germany - , between 1 January 2009 and 14 August 2024. The datasets are currently under embargo and can only be accessed via a request sent to Szabo.Gabriella@tk.hu. Access decisions will be made on a case-by-case basis by the owners of the datasets. The datasets will be made open access by 31 December 2027.