Genre-oriented Readability Assessment: a Case Study

摘要

Whether and to what extent readability assessment is genre-dependent is an issue which has important consequences also in the design and development of educational applications. In this paper, we address this issue from an applicative point of view by investigating whether general purpose readability assessment tools can reliably be used for dealing with texts belonging to different genres. Different experiments have been carried out showing that classification-based approaches to readability assessment can achieve reliable results only by using genre-specific models. Since the construction of genre-specific models is a time consuming task, we proposed a new ranking method for readability assessment based on the notion of distance: we also showed that this method can be usefully exploited for automatically building genre-specific training corpora, thus creating the prerequisites for overcoming the inherent problems of classification-based readability assessment. All reported experiments have been carried out on Italian, a less-resourced language as far as readability assessment is concerned.