DocFusion: A Unified Framework for Document Parsing Tasks

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chai, Mingxu, Shen, Ziyu, Zhang, Chong, Zhang, Yue, Wang, Xiao, Dou, Shihan, Kang, Jihua, Zhang, Jiazheng, Zhang, Qi
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909619259965440
author Chai, Mingxu
Shen, Ziyu
Zhang, Chong
Zhang, Yue
Wang, Xiao
Dou, Shihan
Kang, Jihua
Zhang, Jiazheng
Zhang, Qi
author_facet Chai, Mingxu
Shen, Ziyu
Zhang, Chong
Zhang, Yue
Wang, Xiao
Dou, Shihan
Kang, Jihua
Zhang, Jiazheng
Zhang, Qi
contents Document parsing is essential for analyzing complex document structures and extracting fine-grained information, supporting numerous downstream applications. However, existing methods often require integrating multiple independent models to handle various parsing tasks, leading to high complexity and maintenance overhead. To address this, we propose DocFusion, a lightweight generative model with only 0.28B parameters. It unifies task representations and achieves collaborative training through an improved objective function. Experiments reveal and leverage the mutually beneficial interaction among recognition tasks, and integrating recognition data significantly enhances detection performance. The final results demonstrate that DocFusion achieves state-of-the-art (SOTA) performance across four key tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2412_12505
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle DocFusion: A Unified Framework for Document Parsing Tasks
Chai, Mingxu
Shen, Ziyu
Zhang, Chong
Zhang, Yue
Wang, Xiao
Dou, Shihan
Kang, Jihua
Zhang, Jiazheng
Zhang, Qi
Computation and Language
Document parsing is essential for analyzing complex document structures and extracting fine-grained information, supporting numerous downstream applications. However, existing methods often require integrating multiple independent models to handle various parsing tasks, leading to high complexity and maintenance overhead. To address this, we propose DocFusion, a lightweight generative model with only 0.28B parameters. It unifies task representations and achieves collaborative training through an improved objective function. Experiments reveal and leverage the mutually beneficial interaction among recognition tasks, and integrating recognition data significantly enhances detection performance. The final results demonstrate that DocFusion achieves state-of-the-art (SOTA) performance across four key tasks.
title DocFusion: A Unified Framework for Document Parsing Tasks
topic Computation and Language
url https://arxiv.org/abs/2412.12505