Outcomes of the CLARIN Work on Sentiment in Parliaments Now Available

CLARIN-IT is pleased to announce the publication of the following outcomes of the work on sentiment in parliaments carried out within the CLARIN ParlaMint project:
  • an XLM-R-large model additionally pre-trained od 1.7 billion tokens of ParlaMint and EuroParl data;
  • the ParlaSent dataset, consisting of 18,200 sentences manually annotated for sentiment, in 7 European languages;
  • the XLM-R-ParlaSent model, which is the XLM-R-Parla model fine-tuned on the ParlaSent dataset;
  • an arXiv paper describing the whole extensive annotation and experimentation work.