Document Understanding
Document understanding aims to enable computers to comprehend the content and structure of documents, including text, images, and layouts, to extract key information and answer questions. Current research focuses on improving the efficiency and accuracy of multimodal large language models (MLLMs) for this task, often employing techniques like knowledge distillation, synthetic data generation, and efficient visual processing to handle high-resolution and long-context documents. These advancements are significant because they improve information retrieval, automate document processing tasks, and address privacy concerns through techniques like machine unlearning, ultimately impacting various fields from healthcare to finance.
Papers
GlobalDoc: A Cross-Modal Vision-Language Framework for Real-World Document Image Retrieval and Classification
Souhail Bakkali, Sanket Biswas, Zuheng Ming, Mickaël Coustaty, Marçal Rusiñol, Oriol Ramos Terrades, Josep Lladós
Long-Range Transformer Architectures for Document Understanding
Thibault Douzon, Stefan Duffner, Christophe Garcia, Jérémy Espinas
Visually-Situated Natural Language Understanding with Contrastive Reading Model and Frozen Large Language Models
Geewook Kim, Hodong Lee, Daehee Kim, Haeji Jung, Sanghee Park, Yoonsik Kim, Sangdoo Yun, Taeho Kil, Bado Lee, Seunghyun Park
AWESOME: GPU Memory-constrained Long Document Summarization using Memory Mechanism and Global Salient Content
Shuyang Cao, Lu Wang