This paper presents the Lexicon-Grammar classification of Italian idioms that has been constructed on formal principles and, as such, can be exploited in information extraction. Among MWEs, idioms are those fixed constructions which are hard to automatically detect, given their syntactic flexibility and lexical variation. The syntactic properties of idioms have been formally represented and coded in binary matrixes according to the Lexicon-Grammar framework. The research takes into account idioms with ordinary verbs as well as support verb idiomatic constructions. The overall classification counts 7,000+ Italian idioms. In particular, two binary matrixes of two classes of idioms will be presented. The class C1 refers to the Verb + Object constructions, whereas the class EPC refers to the preposi-tional constructions with the support verb essere. Pre-constructed lexical resources facilitate idioms retrieval both in the case of "hybrid" and "knowledge-based" approaches to Natural Language Processing .