ILC Spring Seminar 09/07/2015

Despite Automatic Text Simplification (ATS) is an active research area across the international NLP community, with significant applicative outcomes for accessibility, very few studies have addressed ATS for the Italian language so far. In this talk, we present our current research in the field, which has led to the development of a first language-specific resource specifically conceived for the investigation of automatic and semi-automatic text simplification for Italian. We illustrate the theoretical underpinnings and the methodological approach adopted for the creation of such a resource, which has to be viewed as a “parallel monolingual corpus”. Some preliminary quantitative findings will be also presented, showing that the most frequent simplification operations retrieved from the corpus not only have a strong correlation with several linguistic complexity features automatically extracted from the parsed text, but also a different distribution according to the “strategy” pursued by the human expert who simplified the text. These data represent the starting point for the design of a flexible semi-automatic text simplification system, i.e. a system specialized for different categories of readers and textual domains.