核心概念
Legal keyphrase generation dataset EUROPA is introduced, highlighting the need for multilingual datasets in the legal domain.
統計
"Our corpus is composed of 17,833 judgments, in 16 languages on average and spanning cases from 1957 to 2023."
"mBART50-8k model outperforms mBART50 with an increase in maximum input context length."
"mBART50-8k model shows improvement across all metrics."
引用
"Keyphrases can reduce the workload of legal experts by allowing them to get the gist of lengthy documents."
"mBART models succeed in outperforming mT5 models."