Il riconoscimento automatico di locuzioni verbali con l’ausilio del software Taltac2

This paper concerns the use of a linguistic resource constituted by verbal locutions derived from the GRADIT, to be applied – via an algorithm implemented in the TaLTaC2 software – for analysing any text. Some of the quantitative characteristics of the resource and the construction of the local grammar of a locution are illustrated. This is done, following the logic of the linguistics of corpora, by applying the resource to a huge corpus of newspaper articles. The calculation of the frequency of these locutions at the level of lemmas provides knowledge of their use in the press, to be exploited for the analysis of specific corpora. Moreover, the key algorithm of automatic recognition is described, as well as the application of the resource to Obama’s speech at the University of Cairo.