SEMANTIC CLASSIFICATION OF VERBS IN THE CROATIAN VERB VALENCY DATABASE

U radu se razmatra opravdanost semantičke razdiobe glagola kao polazišta u njihovoj sintaktičkoj obradbi. Pri podjeli glagola na semantičke skupine pojedini autori polaze od sintaktičkih kriterija (npr. Levin 1993, Dorr 1997, Korhonen i Briscoe 2004, Mikelić Preradović 2010), ističući snažnu povezanost značenjskih obilježja glagola i njihove sintaktičke strukture. S pomoću različitih sintaktičkih alternacija određuje se pripadaju li glagoli istoj semantičkoj skupini. S druge strane, dio jezikoslovaca smatra da je opravdanije držati se isključivo semantičkih kriterija i značenjskih odnosa među glagolima koji pripadaju istomu semantičkom polju (npr. Fellbaum 1998, Šojat 2012). Pri odlučivanju o načinu obradbe glagola u Bazi hrvatskih glagolskih valencija pošlo se od podjele glagola na semantičke skupine zbog spomenute povezanosti dviju razina te pretpostavke da neizvorni govornik hrvatskoga može predvidjeti sintaktički obrazac glagola unutar određene semantičke skupine ako zna sintaktički obrazac prototipnih glagola te semantičke skupine. Na osnovi kriterija čestotnosti i zastupljenosti u priručnicima relevantnim za ovladavanje hrvatskim kao stranim jezikom do razine B1 odabrano je 900 glagola. Ti su glagoli prvotno raspoređeni u semantičke skupine prema značenju koje se u rječnicima navodi kao prvo, no to prvo značenje služi samo za osnovnu, polazišnu podjelu s obzirom na to da se pojedini glagoli po svojim ostalim značenjima mogu naći i u kojoj drugoj skupini.In this paper the justification for the use of semantic classification as a starting point in the syntactic description of verbs is discussed. Some authors use syntactic criteria as the basis of their classification of verbs into semantic groups (e.g. Levin 1993, Dorr 1997, Korhonen and Briscoe 2004, Mikelić Preradović 2010), emphasizing the strong relationship between the verbs’ semantic features and their syntactic structure. They determine whether the verbs belong to the same semantic group by using various syntactic alternations. However, some linguists consider that the application of exclusively semantic criteria and the analysis of semantic relationships between verbs belonging to the same semantic field is a more justified approach (e.g. Fellbaum 1998, Šojat 2012). In the Croatian Verb Valency Database (Birtić and Nahod 2015; Birtić and Runjaić 2015), verb descriptions were created by classifying the verbs into semantic groups based on the aforementioned correlation between the two levels and the assumption that the non-native speakers of Croatian can predict syntactic patterns for most verbs within the semantic group if they know the syntactic patterns of the prototypical verbs within the same semantic group. On the basis of the criteria of frequency and occurrence of verbs in handbooks relevant for learning Croatian as a foreign language, 900 verbs that are necessary to master the Croatian language at B1 level have been selected. These verbs were divided into semantic groups according to the first meaning cited in the dictionaries, but for some verbs the criterion according to which verbs are put into a certain semantic group based on the first meaning has been brought into question. More precisely, it is not clear whether the lexicographers selected the oldest or the most common meaning as the first meaning listed for a particular verb. However, the first meaning was used only for the basic division into 34 semantic groups, because most of the verbs are polysemous and belong to other groups as well. Consequently, the introduction of new semantic groups is inevitable. The long-term goal of this research is to define prototypical syntactic patterns for each semantic group, i.e. to find common syntactic patterns within a semantic group