Efficient Greedy Boosting for Chinese Textbook Readability

Readability means how easy it is for learners to know the written text, and highly readable articles are easier to understand. Readability research is one of the important topics in the field of linguistics and psychology, and text readability analysis is the core of readability research. At present, the research on English readability has been relatively mature, but the research on Chinese readability is still relatively few. In this paper, a new algorithm called Fully Corrective Greedy Multi-classification Boosting(FCGM for short) is proposed to study the readability. Let FCGM and the four traditional machine learning methods which are SVM, logistic regression, Naive Bayes and random forest trained on the corpus we established. By comparing with multiple measurement indicators, it is proved that our method FCGM can achieve preferable results.