Handwritten answer sheet dataset enables better noisy document parsing
HANS: A Handwritten Answer Sheet Dataset for Noisy Hybrid Document Parsing
Artificial IntelligenceComputer Vision and Pattern Recognition
Summary
Grading handwritten student answers is hard because sheets can have messy writing, mix math and text, and include corrections or crossed-out parts. The authors created HANS, a special collection of real student answer sheets that show these challenges and provide detailed labels to help computers learn to understand them better. They also designed a new computer method called NA-GOT that reduces noise and improves reading accuracy on these tricky sheets. Their tests show current tools struggle with HANS, but NA-GOT works more accurately and steadily.
What this means in practice
- •For edtech product teams: Incorporate improved parsing algorithms to automatically grade messy handwritten student answer sheets mixing text and math expressions with noise.$Commercial implications: Enables smarter, more accurate automated grading software sold to educational institutions by improving recognition of real-world messy student answers.
- •For document digitization teams: Develop more robust recognition tools that handle hybrid handwritten documents with corrections and complex layouts for archival and processing.
Authors
Xiazhen Wu, Wansong Qin, Yangbin Zheng, Liangda Fang, Zhan Li, Xiujie Huang, Liushen Zhou, Quanlong Guan
Abstract
Intelligent grading and automated scoring technologies constitute critical infrastructure for smart education. However, existing document parsing and handwriting recognition benchmarks are predominantly designed for well-structured printed documents or isolated mathematical expressions, lacking datasets that capture the complex characteristics inherent to student answer sheets, including multi-line derivation processes, heterogeneous mixtures of text and mathematical formulae, and noise artifacts such as strikethroughs. To address this gap, we introduce HANS, the first dataset explicitly constructed for real-world educational scenarios, encompassing mathematical expressions, natural language text, hand-drawn tables, and diverse noise patterns including corrections and deletions, accompanied by fine-grained annotations that establish a reliable foundation for robust recognition research. Building upon HANS, we propose NA-GOT, an end-to-end framework that achieves two-stage noise suppression through a lightweight noise suppression module operating at the feature level, complemented by a noiseaware attention mechanism incorporated into the decoding stage. Experimental results demonstrate that HANS poses substantial challenges to existing methods, while NA-GOT achieves significant improvements in both accuracy and stability for answer process recognition. The dataset will be made publicly available upon publication.