< previous page page_177 next page >

Page 177
7.1.2.
Other approaches have tried to overcome these difficulties by representing the items of a lexicon as predicates with n arguments. Bierwisch, for instance, represents father as "X parent of Y + Male X + (Animate Y + Adult X + Animate Y)" and kill as "Xs cause (Xd change to (Alive Xd) + Animate Xd)." This kind of representation not only takes into account the immediate semantic markers (in form of a dictionary), but also characterizes the item through the relations it can have, within the framework of a proposition, with other items. In this perspective single semantic items are viewed as already inserted in a possible co-text.
Generative semantics has improved the use of predicate calculus, but shifting from the representation of single terms to the logical structure of the propositions (McCawley, Lakoff, and others). Only Fillmore has tried, with his case grammar, to unify both interpretive-compositional and generative perspective. Fillmore remarks that the verbs ascend and lift are both motion verbs and are both used to describe a motion upward, but lift requires conceptually two objects (the one moving upward, the other causing the motion), whereas ascend is a one-argument predicate. This remark leads one to recognize that arguments, in natural languages, can be identified with roles (similar to the actants in Greimas' structural semantics); for any predicate there is an Agent, a Counteragent, an Object, a Result, an Instrument, a Source, a Goal, an Experiencer, and so on. This kind of analysis solves very well the problem of the classification of features, following a sort of logic of action. Moreover, it satisfies the encyclopedic requirement and transforms a purely classificatory representation into an operational schema: the composition of the meaning of a predicate tells us how to act in order to give rise to the denoted action or in order to isolate it within a context. To walk, for instance, should mean that there is a human agent, using ground as a counteragent, moving his body in order to displace it (as a result) from a spacial source to a spacial goal, by using legs as instrument, and so on.
However, some objections can be raised. (i) Whereas the roles can be recognized as a set of innate universals expressed by a fixed inventory of linguistic expressions, the linguistic features which fill in these roles are again potentially infinite (how many kinds of instrument can be foreseen?). (ii) The proposal of such a 'case grammar' seems to work apropos of predicates, but requires some additions as far as the representation of arguments is concerned. Using a knife as instrument, I can kill someone, but what about the semantic representation of knife? It seems that, more than a predicate argument structure, it could be useful in this case to employ such categories as who produces it, with what material, according to what formal rule and for what purpose. This kind of representation recalls the four Aristotelian causes (Efficient, Formal, Material, and Final); but the representation of an 'object' could also be

 
< previous page page_177 next page >