An approach to abstractive summarization for norwegian bokmål

Abstract

© Springer Nature Switzerland AG 2019. In this paper we present a new Norwegian language (Bokmaal) dataset for abstractive summarization task. And we also present results for this task for methods like classic TextRank and modern neural networks based ones, the best method, which uses named entity reparation technique, gets 24.3 ROUGE-1 and 0.02 extraction score.