Collocations as an index for distinguishing text genres

This paper aims to incorporate collocations as an index to distinguish text genres: our main hypothesis is that collocations, as well as other linguistic features, are potentially suitable to identify genres. Thus, this is mostly an exploratory study, aimed at verifying this hypothesis and at taking a deeper look into register variation across different genres in Italian with computational and statistical methods. Furthermore, in a broader perspective, this study might give significant contributions in other fields, such as automatic genre identification [Santini 2004], measure of text cohesion [Louwerse et al. 2004] or text readability, where the detection of collocations as a marker of genres can increase the accuracy of computational tools devoted to these tasks.