Abstract
In this paper, alternation based model of the valency lexicon of Czech verbs, VALLEX, is described. Two types of alternations (changes in valency frames of verbs) are distinguished on the basis of used linguistic means: (i) grammaticalized alternations and (ii) lexicalized alternations.
Both grammaticalized and lexicalized alternations are either conversive, or non-conversive. While grammaticalized alternations relate different surface syntactic structures of a single lexical unit of a verb, lexicalized alternations relate separate lexical units.
For the purpose of the representation of alternations, we divide the lexicon into data and rule components. In the data part, each lexical unit is characterized by a single valency frame and by applicable alternations.
In the rule part, two types of rules are contained: (i) syntactic rules describing grammaticalized alternations and (ii) general rules determining changes in the linking of situational participants with valency complementations typical of lexic