UniHDSA: A Unified Relation Prediction Approach for Hierarchical Document Structure Analysis
Jiawei Wang, Kai Hu, Qiang Huo
2025-03-27
Summary
This paper is about creating a new way for computers to understand the structure of documents, like reports or articles, by looking at how the different parts are related to each other.
What's the problem?
It's hard for computers to understand the layout and organization of documents, which makes it difficult for them to extract information or summarize the content.
What's the solution?
The researchers developed a new AI model that treats different parts of the document as being related in various ways, allowing it to understand the overall structure more effectively.
Why it matters?
This work matters because it can help computers better understand and process documents, which is important for tasks like information retrieval, document summarization, and knowledge extraction.
Abstract
Document structure analysis, aka document layout analysis, is crucial for understanding both the physical layout and logical structure of documents, serving information retrieval, document summarization, knowledge extraction, etc. Hierarchical Document Structure Analysis (HDSA) specifically aims to restore the hierarchical structure of documents created using authoring software with hierarchical schemas. Previous research has primarily followed two approaches: one focuses on tackling specific subtasks of HDSA in isolation, such as table detection or reading order prediction, while the other adopts a unified framework that uses multiple branches or modules, each designed to address a distinct task. In this work, we propose a unified relation prediction approach for HDSA, called UniHDSA, which treats various HDSA sub-tasks as relation prediction problems and consolidates relation prediction labels into a unified label space. This allows a single relation prediction module to handle multiple tasks simultaneously, whether at a page-level or document-level structure analysis. To validate the effectiveness of UniHDSA, we develop a multimodal end-to-end system based on Transformer architectures. Extensive experimental results demonstrate that our approach achieves state-of-the-art performance on a hierarchical document structure analysis benchmark, Comp-HRDoc, and competitive results on a large-scale document layout analysis dataset, DocLayNet, effectively illustrating the superiority of our method across all sub-tasks. The Comp-HRDoc benchmark and UniHDSA's configurations are publicly available at https://github.com/microsoft/CompHRDoc.