基于自然语言处理的图书阅读难度自动分级研究

Grading the reading difficulty degree based on natural language processing

图书阅读难度自动分级系统能够帮助儿童读者快速找到适合自己认知水平的图书。文章基于图书句子难度和字难度两个维度,建立了一个图书难度分级模型,并开发出对应的图书阅读难度自动分级系统。利用该系统对常见的儿童图书进行了测试,初步实验表明:从图书中随机选取字数达到2500-3000字时,图书阅读难度分级算法测试结果误差较小,综合使用字难度和句子难度的分级算法比单纯字难度分级算法和单纯句子难度分级算法的效果更好。该系统目前仅限于白话文图书应用。

The automatic grading system of reading difficulty degree can help children find books that are suitable for their own cognitive level. Based on the two dimensions of sentence difficulty and word difficulty, this paper establishes a hierarchical model of reading difficulty degree, and develops an automatic grading system of reading difficulty degree. The common children＇s books are tested by this system, preliminary experiments show that when the randomly selected books with words up to 2500-3000, the grading algorithm of reading difficulty degree results in less errors, and the grading algorithm using both sentence difficulty and word difficulty has a better result than that only using sentence difficulty or word difficulty. The system is currently limited to the books of using vernacular.