SEAHORSE is a dataset for multilingual, multifaceted summarization evaluation. It consists of 96K summaries with human
ratings along 6 quality dimensions: comprehensibility, repetition, grammar, attribution, main idea(s), and conciseness,
covering 6 languages, 9 systems and 4 datasets.
