Automatic Grading Of Text Difficulty Of Extracurricular Reading Materials Based On Primary-School English Textbooks

Extracurricular English reading plays an important role in the development of students’ reading literacy and the implementation of the English Subject Core Competency.In recent years,graded reading has become a new trend in English teaching.It emphasizes that students of different ages should be provided with reading materials of different difficulty.However,the grading standards of reading materials in China are a little subjective,and teachers lack scientific methods in choosing reading materials for students.Therefore,this study aims to use quantitative and automatic methods to design a tool that can evaluate the difficulty level of English extracurricular reading materials,so as to grade extracurricular reading materials objectively and then help teachers choose students’ English reading materials scientifically.Automatic grading of text difficulty refers to the automatic judgment of the difficulty level of a text based on its features by computer.At present,the relevant research in China is slightly inadequate.First,researchers tend to use foreign readability formulas to evaluate the difficulty of our middle school or college English textbooks.Secondly,there is a lack of automatic English text difficulty grading tool for primary school currently.Therefore,this study will construct an automatic grading model of English text difficulty which is suitable for our primary school by taking reasonably graded English textbooks as the corpus and grades as the grading method.This study mainly focuses on the following questions:(1)What are the main text features that will affect the difficulty level of primary-school English textbooks?(2)What modeling method could be adopted in order to achieve automatic grading of English text difficulty?(3)Based on this model,how do teachers choose English extracurricular reading materials for students?In order to solve the problems above,this study mainly carried out the following work.The first is to construct a corpus of primary-school English textbooks.After collecting 5sets of primary-school English textbooks from grade 3 to grade 6,we successfully constructed this corpus through a series of processes,namely text extraction,manual proofreading,text normalization and so on.The second is to do quantitative analysis on features which will affect the difficulty of primary-school English textbooks.From 3 dimensions of vocabulary,sentence and discourse,this study explored the changes of 13 features in the textbooks with grade changes.This work could help teachers fully grasp the difficulty of the current primary-school English textbooks.The third is to construct the automatic grading model of English text difficulty.Firstly,eight main text features were extracted by calculating the correlation between text features and difficulty level.They were respectively total numbers of words,different words,the proportion of the new words,average word frequency,average word length,total numbers of sentences,average sentence length and the proportion of conjunction.Then,linear regression and neural network were used to construct automatic grading models.It was proved that the two models could be both used to predict and grade the difficulty of English texts.The fourth is to design a Graded Reading List of Primary-School English Extracurricular Reading Materials.Based on the model,85 books were evaluated.Then,according to their difficulty levels,this study classified them into four grades and then formed a graded reading list which would be convenient for teachers to choose reading materials initially.This work fully embodies the significance and value of this research.In conclusion,this study provides a theoretical basis for the establishment of standards for grading reading materials in the field of education and teaching.And it also provides a scientific method for teachers to select extracurricular reading materials.