中文文本适读性评估特征体系研究——基于语言认知的可解释性视角  

A Study on the Feature System for Readability Assessment of Chinese Texts:From a Language Cognition-based Interpretability Perspective

作　　者：杨丽姣[1] 徐会丹 宋培彦 YANG Lijiao;XU Huidan;SONG Peiyan

机构地区：[1]北京师范大学国际中文教育学院,北京100875 [2]清华大学人文学院,北京100084 [3]天津师范大学管理学院,天津300387

出　　处：《语言文字应用》2025年第1期100-113,共14页Applied Linguistics

基　　金：国家社科基金一般项目“基于文本适读性智能评估方法的汉语儿童阅读资源体系构建研究”(23BYY198)支持。

摘　　要：在国际中文教育数字化应用快速发展的背景下,阅读材料量化分析与智能评估的需求日益增长。本研究基于《国际中文教育中文水平等级标准》的等级框架,融合语料库方法与自然语言处理技术,从语言认知的可解释性出发,设计并构建了一种适用于国际中文教育领域的文本适读性评估特征体系,系统剖析了多维度语言特征与学习者词汇语义认知、文本理解之间的内在联系,为中文文本适读性评估的理论与实践提供了新视角。为进一步探究特征体系的科学性与实用价值,本研究构建了大规模基础资源库,验证了词语抽象度、构式密度等特征在文本适读性评估中的关键作用,发现汉语文化词等特征的适用性及局限性。在此基础上,本课题团队研发了中文文本适读性智能评估系统(ACTR)。该系统提升了国际中文教育领域文本自动定级的精细度与准确性,有助于中文阅读资源的高质量评估、建设与优化。Amid the rapid development of digital applications in the field of International Chinese Education,the demand for quantitative analysis and intelligent evaluation of reading materials has been steadily increasing.This study,grounded in the proficiency level framework of the International Chinese Language Education Standards,integrates corpus-based methods with natural language processing techniques to design and construct a comprehensive Chinese text readability assessment model.By exploring the intrinsic relationships between multidimensional linguistic features and learners'cognitive processing and semantic understanding,this study provides new theoretical and practical perspectives for the evaluation of text readability in Chinese.To further validate the scientific rigor and practical value of the assessment framework,a largescale foundational resource database was developed.Empirical analyses confirmed the critical roles of key linguistic features,such as word abstraction and construction density,in readability assessment.Additionally,the applicability and limitations of specific features,such as Chinese cultural terms,were systematically examined.Building upon this foundation,the research team developed the Assessment of Chinese Text Readability(ACTR)system.This intelligent evaluation system demonstrated high prediction accuracy in text grading across datasets with diverse stylistic features,offering robust quantitative analysis and intelligent support for the evaluation,development,and optimization of high-quality Chinese reading resources.

关 键 词：文本适读性 特征体系 认知可解释性 语料库方法 自然语言处理技术 

分 类 号：H195[语言文字—汉语]