Automatic prediction of text readability for international Chinese Language education
Shuqin Zhu, Man Zhang, Dongdong Guo · 2024
This paper studies the text readability influencing factors and its automatic assessment issues in international Chinese language education. First, a graded text corpus was constructed based on international Chinese language textbooks, followed by an analysis of the factors influencing text readability in four dimensions: domain features, surface features, structural features, and functional features. Each feature involves three levels: character, word, and sentence. Finally, a readability formula was constructed based on a multiple linear regression model, which was then subjected to readability analysis and validation. The results show that the goodness of fit R2 of the readability formula constructed in this study is high, in which key factors such as the number of Chinese character types, the proportion of band-1 words, the average number of parts of speech (POS) in sentences, the standard deviation of the number of POS in sentences, the proportion of band-7 words, the proportion of function words, the standard deviation of the number of Chinese character strokes, and the mean of Chinese character grade bands can significantly explain the grade bands of text readability.