首页|期刊导航|大数据|构建面向汉语二语学习者的手写作文语法纠错数据集:方法与应用

构建面向汉语二语学习者的手写作文语法纠错数据集:方法与应用OA

Constructing handwritten grammatical error correction dataset for CSL learners:methodology and application

中文摘要英文摘要

针对中文语法纠错研究以印刷体文本数据集为主、缺少手写文本专用数据集的现状,分析了构建手写语法纠错数据集面临的关键挑战.据此,构建了面向汉语二语学习者的手写作文语法纠错数据集,以真实手写作文为语料,使用优化光学字符识别处理涂改重写等噪声,经过文本识别、标准化标注、图文对齐与质量校验,形成细粒度多模态的标注数据,共包含8 726条有效样本,覆盖599类细粒度错误类型,可为手写场景语法纠错、错误诊断及智能汉语教学提供数据支撑.

Current research on Chinese grammatical error correction mainly relies on printed text datasets and lacks dedicated data for handwritten text.Accordingly,a handwriting Chinese grammatical error correction dataset for Chinese as a second language learners was constructed.Using authentic handwritten compositions as the corpus,an optimized optical character recognition system was employed to process noise such as overwriting and revisions.Through text recognition,standardized annotation,image-text alignment,and quality validation,fine-grained,multi-modal annotated data was generated.The dataset comprises 8 726 valid samples,covering 599 fine-grained error types,and can provide data support for grammatical error correction,error diagnosis,and intelligent Chinese language teaching in handwriting scenarios.

李春秋;张潇晓;梁远远;袁丹;兰韵诗;王晔

华东师范大学数据科学与工程学院,上海 200062华东师范大学数据科学与工程学院,上海 200062华东师范大学国际汉语文化学院,上海 200062华东师范大学国际汉语文化学院,上海 200062华东师范大学数据科学与工程学院,上海 200062华东师范大学数据科学与工程学院,上海 200062

信息技术与安全科学

中文语法纠错手写作文批改汉语二语数据集构建光学字符识别

Chinese grammatical error correctionhandwritten composition correctionChinese as a second languagedataset constructionoptical character recognition

《大数据》 2026 (5)

97-113,17

华东师范大学文理交叉跨学科培育项目(No.2025ECNU-WLJC007) The Interdisciplinary Cultivation Program for ArtsScience Integration,East China Normal University(No.2025ECNU-WLJC007)

10.11959/j.issn.2096-0271.BDR26164

评论