Light verb constructions (LVCs), also known as compound verbs, require spacial treatment within a computational system. Recently, there has been some work on the automatic identification of LVCs, in resource-rich languages, such as English. Our goal is to adapt such existing techniques for the automatic treatment of LVCs in an under-resourced language, such as Persian. We focus on the most common subclass of Persian LVCs which are noun+verb constructions. LVCs are often formed semi-productively: Although a light verb occurs with a wide range of nouns, it tends to productively combine with certain semantic classes to form LVCs. We expand an existing measure of determining LVC acceptability (for English) to make explicit use of semantic classes of nouns (in Persian). We show that this new class-based acceptability measure outperforms the original measure, when applied to Persian candidate LVCs.
展开▼
