Improving Factual Error Correction for Abstractive Summarization via Data Distillation and Conditional-generation Cloze

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Li, Yiyang, Li, Lei, Hu, Dingxin, Hao, Xueyi, Litvak, Marina, Vanetik, Natalia, Zhou, Yanquan
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914677733195776
author Li, Yiyang
Li, Lei
Hu, Dingxin
Hao, Xueyi
Litvak, Marina
Vanetik, Natalia
Zhou, Yanquan
author_facet Li, Yiyang
Li, Lei
Hu, Dingxin
Hao, Xueyi
Litvak, Marina
Vanetik, Natalia
Zhou, Yanquan
contents Improving factual consistency in abstractive summarization has been a focus of current research. One promising approach is the post-editing method. However, previous works have yet to make sufficient use of factual factors in summaries and suffers from the negative effect of the training datasets. In this paper, we first propose a novel factual error correction model FactCloze based on a conditional-generation cloze task. FactCloze can construct the causality among factual factors while being able to determine whether the blank can be answered or not. Then, we propose a data distillation method to generate a more faithful summarization dataset SummDSC via multiple-dimensional evaluation. We experimentally validate the effectiveness of our approach, which leads to an improvement in multiple factual consistency metrics compared to baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2402_08581
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Improving Factual Error Correction for Abstractive Summarization via Data Distillation and Conditional-generation Cloze
Li, Yiyang
Li, Lei
Hu, Dingxin
Hao, Xueyi
Litvak, Marina
Vanetik, Natalia
Zhou, Yanquan
Computation and Language
Improving factual consistency in abstractive summarization has been a focus of current research. One promising approach is the post-editing method. However, previous works have yet to make sufficient use of factual factors in summaries and suffers from the negative effect of the training datasets. In this paper, we first propose a novel factual error correction model FactCloze based on a conditional-generation cloze task. FactCloze can construct the causality among factual factors while being able to determine whether the blank can be answered or not. Then, we propose a data distillation method to generate a more faithful summarization dataset SummDSC via multiple-dimensional evaluation. We experimentally validate the effectiveness of our approach, which leads to an improvement in multiple factual consistency metrics compared to baselines.
title Improving Factual Error Correction for Abstractive Summarization via Data Distillation and Conditional-generation Cloze
topic Computation and Language
url https://arxiv.org/abs/2402.08581