Dmesure: readability for FFL exercises

Title

Lexical and syntactic complexities: a difficulty model for automatic generation of language exercises in FFL.

Abstract

This research aims to develop a difficulty model, based on linguistic features, and able to assess automatically the level of texts or sentences, which can then be used in a system of automatic generation of language exercises. This model should ensures that the resulting exercices fit the learners' level.

More precisely, this research is about:

• classification of texts or sentences according to the CEFR scale (Common European Framework of Reference for Languages). These classified pieces of texts may be used for didactic purposes, such as automatic generation of language exercises, since their difficulty is controlled.
• exploration of the predictive capacity of various linguistics features for this classification task ;
• comparison of the performances of various classification techniques, such as logistic regression and data mining algorithms, on our dataset.