Predicting readability of texts for Italian L2 students: a preliminary study

Text selection and comparability for L2 students to read and comprehend are central concerns both for teaching and assessment purposes.Compared to subjective selection. quantitative approaches provide more objective information, analysing texts at language and discourse level (Khalifa & Weir, 2009). Readability formulae such as the Flesch Reading Ease, the Flesch-Kincaid Grade Level and, for Italian, the GulpEase index (Lucisano and Piemontese, 1988), do not fully addressed the issue of text complexity. A new readability formula called Coh-Metrix was proposed (Crossley, Geenfield, & McNamara 2008), which takes into account a wider set of language and discourse features. A similar approach was proposed to assess readability of Italian texts through a tool called READ-IT (Dell’Orletta, Montemagni, & Venturi 2011). While READ-IT was tested on newspaper texts randomly selected, this contribution focuses on the development of a similar computational tool applied on texts specifically selected in the context of assessing Italian as L2. Two text corpora have been collected from the CELI (Certificates of Italian Language) item bank at B2 and C2 level. Statistical differences in the occurrence of a set of linguistic and discursive features have been analysed according to four different categories: length features, lexical features, morpho-syntactic features, and discursive features.