Design and implementation of neural network for English reading difficulty classification
Reading text is one of the most important factors which influence English reading.Effective evaluation to the diffi-culty of English text can provide different reading materials to the learners with different reading abilities,help teachers to scientifi-cally select appropriate teaching resources,and provide scientific guidance for test proposers.In this paper,based on the neural dy-namics algorithm(NDA)method of convergent difference neural network(CDNN),the text data are preprocessed with feature selec-tion,weighting and sample normalization for training networks with different mapping functions,and then the output of the network is enhanced and generalized by voting rules for augmented generalization,thus realizing a reading difficulty classification method for English text combined by a voting convergent difference neural network(Voting-CDNN,V-CDNN),which improves the compu-tational efficiency and classification accuracy.The experimental results show that the highest and average values of classification accuracy of V-CDNN reach 98.81%and 95.45%,respectively,which further confirm that V-CDNN is a high-performance classifier in terms of computation time,average accuracy and maximum accuracy.
text classificationvoting ruleconvergent difference neural network
万方数据
维普
