摘要：The paper deals with valence theory and its application to Croatian verbs. In the first part of the paper we briefly present the theoretical postulates upon which our description of verb valency is based. Thereby we mainly rely on frame semantics and its assumption that valency can be stated and described for the whole semantic frame, i. e. the same semantic valency is shared by all the members of the frame. We extend this assumption to semantic fields in the Croatian WordNet and treat them as frames with internal structure in terms of semantic relations, such as synonymy and hyponymy. We use this hypothesis as a theoretical basis in the construction of verb frames. We want to include verb frames in the Croatian WordNet in order to broaden its applicability in various NLP tasks, such as machine or machine aided translation, rule based parsing, data mining etc. In the second part of the paper we discuss some of the existing works dealing with Croatian verbs and valency description. In this discussion we focuss on the descriptive inventory for the so called syntactic valency of Croatian verbs. Particular attention is paid to some controversis in discussed models, namely to the so called infinitive and verbative complement. Although we do not provide a definitive solution regarding these issues, by giving the counterexamples we try to exemplify the causes of possible misinterpretaions. In the final part of the paper we present our own inventory for the morphosyntactic description and annotation of verbal complements used in this phase of the of verb frame construction.
