Exploring Cross-Lingual Guidance in Abstractive Summarization

Abstract

Cross-lingual guidance (CLG) as an augmentation method is often applied in cross-lingual summarization (CLS) to improve its performance. In this paper, we empirically study how cross-lingual information of different quality benefits the encoding and decoding procedures for both cross-lingual and mono-lingual abstractive summarization. We specially propose a summarization model DualSum which can utilize CLG in both encoding and decoding, and construct a dataset BiRead with high-quality parallel bilingual document-summary pairs. The empirical experiments will show how CLS and MLS are influenced by CLG.

Tasks

Reproductions