A quantitative study on the morphology of Italian multiword expressions

Internal morphological rigidity of multiword expressions is a highly debated topic in linguistic studies, although observations are usually qualitative, and rarely supported by actual data specifically extracted with this aim and on this issue. We report on corpus-based experiments on Italian nominal MWEs, with the purpose of determining degrees of internal fixedness numerically. We observe that MWEs lie on a variability continuum (both across and within morphosyntactic configurations) and we suggest measures to quantify this scale. However, lexico-semantic and morphosyntactic factors do not contribute equally to the definition of MWE-hood and nominal expression rather seem to be biased towards the former. We conclude that, also in the perspective of extraction tasks, semantic generalisations must necessarily come into play.