A New Approach to Readability Study Based on Information Computing

This article demonstrates a new approach to measuring English readability applying techniques from natural language processing and information theory. We attempt to regard reading process as a process of information processing and transferring. And then we apply statistical language modeling technique and information theory to compute the information of text passage, based on which the readability measuring models are trained. The results of experiments show that this new approach holds a higher correlation value (0.849) than the traditional formula such as ARI formula (0.540) with the grade level of test data. And the new approach can take the word sequence, readerpsilas language characteristics and environment into account when measuring readability, which cannot be performed by the previous formulas.