European Parliament Speeches 2002-2022

This dataset contains the 137,844 speeches given in the European Parliament between September 2002 and December 2022, covering five election periods. The speeches in the dataset are segmented into 474,545 paragraphs, with each line containing one paragraph. Each line in the dataset contains: speech_paragraph - The id of the speech and the number of the paragraphs. date - The date as YYYY-MM-DD text_en - The paragraph text hf_sentiment - Sentiment of the paragraph as {POSITIVE, NEGATIVE} determined by https://huggingface.co/siebert/sentiment-roberta-large-english tb_sentiment - Sentiment polarity of the paragraph as determined by https://textblob.readthedocs.io/en/dev/quickstart.html#sentiment-analysis. The polarity is mapped to {VERY NEGATIVE, POSITIVE, NEURTAL, NEGATIVE, VERY NEGATIVE} at thresholds +-0.2 and +-0.6. The speeches are extracted from the European Parliament's ``Verbatim Report of Proceedings'', which are released for each day at http://redmapl3.europarl.europa.eu/RedmapFront/media/reds_iPlCre_Sit/. The proceedings were extracted from the .xml files. All non-English speeches were translated to English using the deepl.com API.