2008 Volume 25 Issue 3 Pages 272-281
This paper proposes a model of predicting EFL learners’ sentence reading time based on learners’ reading ability and linguistic features of a sentence. The reading ability here means the learners’ TOEIC (Test of English for International Communications) scores, and the linguistic features indicate lexical, syntactic and discourse complexities. We use natural language processing technology to automatically extract the linguistic features from the sentences, and construct the model by using the Support Vector Machines as a learning mechanism to combine the learners’ and the linguistic features. An experiment has shown that the proposed model predicts reading time with an error rate of 18.8%, which is lower than the error rates marked by other models.
