Readability Analysis of Kannada Language

摘要

This paper proposes a neural network based approach to predict a readability score for a given sentence in Kannada language without the use of any predefined word list. We have used textbooks of Grade-1 to Grade-10 as input to train the model, and were able to achieve a correlations well above all the state-of-the-art methods. We also show and prove the general behavior of readability scores of w.r.t the number of sentences considered in each class.