Graph-based Document Structure Analysis

TL;DR

This work defines graph-based Document Structure Analysis (gDSA) to capture holistic document structure by generating relational graphs among layout elements. It introduces GraphDoc, a large-scale dataset with 80K document images and 4.13M relation annotations across spatial and logical relations, enabling tasks like reading order and hierarchical structure analysis. The paper presents DRGG, an end-to-end Document Relation Graph Generator that jointly detects elements and predicts a relation graph, achieving strong results (e.g., mAP_g@0.5 of 57.6% and mR_g of 30.7% on GraphDoc) and demonstrating the benefit of integrating graph reasoning into document understanding. The work also provides a rule-based annotation pipeline, comprehensive evaluation metrics for gDSA, and extensive ablations, highlighting the potential of graph-based representations to improve real-world document processing tasks while noting current limitations and avenues for multimodal and multi-page extensions.

Abstract

When reading a document, glancing at the spatial layout of a document is an initial step to understand it roughly. Traditional document layout analysis (DLA) methods, however, offer only a superficial parsing of documents, focusing on basic instance detection and often failing to capture the nuanced spatial and logical relations between instances. These limitations hinder DLA-based models from achieving a gradually deeper comprehension akin to human reading. In this work, we propose a novel graph-based Document Structure Analysis (gDSA) task. This task requires that model not only detects document elements but also generates spatial and logical relations in form of a graph structure, allowing to understand documents in a holistic and intuitive manner. For this new task, we construct a relation graph-based document structure analysis dataset (GraphDoc) with 80K document images and 4.13M relation annotations, enabling training models to complete multiple tasks like reading order, hierarchical structures analysis, and complex inter-element relation inference. Furthermore, a document relation graph generator (DRGG) is proposed to address the gDSA task, which achieves performance with 57.6% at mAP$_g$@0.5 for a strong benchmark baseline on this novel task and dataset. We hope this graphical representation of document structure can mark an innovative advancement in document structure analysis and understanding. The new dataset and code will be made publicly available at https://yufanchen96.github.io/projects/GraphDoc.

Figures (8)

• Figure 1: GraphDoc Dataset Overview. Figure \ref{['fig1-a']} illustrates the threefold considerations, including (i) the inclusion of spatial and logical relations, (ii) support for multiple relations between layouts pairs, (iii) and the integration of non-textual elements. Figure \ref{['fig1-b']} demonstrates the state-of-the-art performance of our proposed method, showcasing $\text{mAP}$ results for the Document Layout Analysis (DLA) task, as well as $mR_g$ and $mAP_g$ results for the graph-based Document Structure Analysis (gDSA) task on the GraphDoc dataset.
• Figure 2: Overview of the GraphDoc Dataset's Task, which illustrates both DLA and gDSA tasks of GraphDoc are based on image analysis.
• Figure 3: Logical Relationship in GraphDoc Dataset. There are 4 instinct types of relations. The relational graph effectively filters out extraneous connections that might appear in other types of diagrams, providing a clearer representation of the actual relationships.
• Figure 4: Relation statistics on the GraphDoc dataset. The chord diagram on the left illustrates the distribution of relationships among various layouts. The heatmap on the right visualizes the intensity of relations based on layouts (deeper color means higher intensity). Below the heatmap, a detailed image presents the case of Reference relations for Picture.
• Figure 5: Proposed Document Relation Graph Generator (DRGG) for Document Layout Analysis and Document Structure Analysis. The key of our model is illustrated in the Relation Head, which is responsible for predicting relations between layout elements. The remaining parts are the standard encoder-decoder architecture used for object detection.