Multi-task generative dialogue summarization learning framework with topic-based data augmentation

Jing Shan

doi:10.1016/j.csl.2026.101998

Multi-task generative dialogue summarization learning framework with topic-based data augmentation

2026-05-06

0

PRE

AI

DOI：10.1016/j.csl.2026.101998

Original

Original request for help

Save

Abstract

En 中文

• We introduce a multi-task learning dialogue summarization framework with topic-based data augmentation. We have also developed an open-source implementation of our framework, which will be publicly released along with the final version of this paper. • We incorporate external knowledge at both the entity and paragraph levels, enabling the model to better comprehend and summarize dialogues with enriched semantic understanding. • We conduct extensive experiments including comparative experiments, ablation studies, human assessments, and case studies on two real-world datasets. Through extensive experiments, we demonstrate that our method achieves new state-of-the-art results on both SAMSum and DialogSum datasets.

Keywords:

multi-task learning

dialogue summarization

topic-based data augmentation

entity-level knowledge

paragraph-level knowledge

Multi-task generative dialogue summarization learning framework with topic-based data augmentation

Abstract

Journal

Organization