As digital government services, smart offices, digital archives, and artificial intelligence become increasingly integrated, document intelligence is being applied ever more widely in scenarios such as layout parsing, content recognition, information extraction, intelligent question answering, and multimodal understanding. High-precision, highly robust document parsing capability is the core foundation for digitizing, structuring, and intelligently utilizing massive volumes of paper and electronic documents.
However, real-world document data commonly suffers from complex layouts, blurred handwriting, occlusion and distortion, multi-source heterogeneity, mixed languages, and interleaved handwritten and printed text. These issues lead to sparse effective features and abundant interference, making it difficult for traditional text processing methods and general-purpose vision models to handle complex document tasks. In particularly challenging scenarios—such as ancient books and historical documents, handwritten receipts, formulas and tables, and blurred scans—existing algorithms are prone to parsing deviations, missed information detection, structural misjudgment, and failures in semantic understanding. This severely limits the industrial adoption of document intelligence technologies and raises the bar for building high-precision, strongly generalizable, and highly reliable document vision and multimodal learning systems.
In recent years, multimodal large models, few-shot learning, robust feature learning, and intelligent evaluation techniques have advanced rapidly, offering a brand-new technical paradigm for complex document parsing. Against this backdrop, research into the fundamental theories, core algorithms, and engineering applications of document intelligence for real-world complex scenarios carries significant academic value and broad prospects for practical deployment.
To support the development of the digital content industry and smart government services, to systematically review frontier research progress in document intelligence in China, and to promote algorithmic innovation, technological iteration, and deep integration of industry, academia, and research, China DocAI 2026 is hereby convened. The conference will take place in Beijing from December 25 to 27, 2026. Submissions are warmly invited, and outstanding papers will be selected for publication in the “Intelligent Document” special issue of the Journal of Image and Graphics.
