BriefGPT.xyz
Aug, 2022
基于质心预训练的多文档摘要
Multi-Document Summarization with Centroid-Based Pretraining
HTML
PDF
Ratish Puduppully, Mark Steedman
TL;DR
本文提出了一种简单的预训练目标：选择每个文档簇的基于ROUGE的中心点作为摘要，以用于多文件摘要的预训练。通过多个MDS数据集的零-shot和完全监督实验，我们证明了我们的Centrum模型比最先进的模型更好或具有可比性。
Abstract
In
multi-document summarization
(MDS), the input is a cluster of documents, and the output is the cluster summary. In this paper, we focus on
pretraining objectives
for MDS. Specifically, we introduce a simple pr
→