TDATR: Improving End-to-End Table Recognition via Table Detail-Aware Learning and Cell-Level Visual Alignment

2026-03-24

AI Summary

This work addresses the limitations of existing table recognition methods, which often decouple structural and content modeling, and suffer from data scarcity in low-resource settings due to their reliance on large annotated datasets. To overcome these challenges, the authors propose a “Perceive-and-Fuse” framework that jointly models table structure and content through detail-aware perception learning. The approach integrates a structure-guided cell localization module and a cell-level visual alignment mechanism, enabling efficient end-to-end recognition. Formulated within a language modeling paradigm, the method performs multi-task learning to implicitly fuse fine-grained table details and directly generates HTML output without requiring dataset-specific fine-tuning. Evaluated on seven benchmark datasets, the model achieves state-of-the-art or highly competitive performance, significantly outperforming current end-to-end approaches while enhancing robustness and interpretability in low-resource scenarios.

Technology Category

• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Machine Learning: Structured Learning
• Computer Vision: Representation Learning for Vision

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Tables are pervasive in diverse documents, making table recognition (TR) a fundamental task in document analysis. Existing modular TR pipelines separately model table structure and content, leading to suboptimal integration and complex workflows. End-to-end approaches rely heavily on large-scale TR data and struggle in data-constrained scenarios. To address these issues, we propose TDATR (Table Detail-Aware Table Recognition) improves end-to-end TR through table detail-aware learning and cell-level visual alignment. TDATR adopts a ``perceive-then-fuse'' strategy. The model first performs table detail-aware learning to jointly perceive table structure and content through multiple structure understanding and content recognition tasks designed under a language modeling paradigm. These tasks can naturally leverage document data from diverse scenarios to enhance model robustness. The model then integrates implicit table details to generate structured HTML outputs, enabling more efficient TR modeling when trained with limited data. Furthermore, we design a structure-guided cell localization module integrated into the end-to-end TR framework, which efficiently locates cell and strengthens vision-language alignment. It enhances the interpretability and accuracy of TR. We achieve state-of-the-art or highly competitive performance on seven benchmarks without dataset-specific fine-tuning.

Problem

• table recognition
• modular pipeline
• end-to-end learning
• data-constrained scenario
• structure-content integration

Innovation

• table recognition
• end-to-end learning
• cell-level alignment
• structure-aware modeling
• language modeling paradigm