GASG: Minimal Syntax, Maximal Lexicon and Prolog

0. A new non-transformational generative grammar, which can be classified as an extreme categorial grammar [18] (or construction grammar [15]), is proposed in this paper, called Generative/Generalized Argument Structure Grammar (GASG). A fundamental theoretical problem of (dynamic) semantics (concerning compositionality; section 1) and the metatheoretical problem of ''Perfect Syntax" in current generative literature (section 3) have led us to conclude that a grammar like GASG is worth elaborating. The relevance of GASG here lies in the fact that both its morphosyntactic representational level and its (discourse) semantic level are Prolog-like, which promises immense computational advantages (section 2).
1. The (discourse) semantic starting-point concerns the principle of compositionality, attributed to Frege, which is a way (the strictest way) of articulating the systematic correspondence between syntax and semantics [19]: ''the meaning of a complex expression is a function of the meanings of its parts and of the syntactic rules by which they are combined." Groenendijk and Stokhof [9],[10] argue that there are methodological, philosophical, empirical, computational and practical reasons ''to be interested in trying to keep to compositionality."
In the extension of model-theoretic semantics from the sentential to the discourse level, however, such theories have emerged (e.g. [11, 12, 13, 14]) that are expressly non-compositional. Groenendijk and Stokhof [9-10] attribute their non-compositional nature just to their most attractive component: the simple and straightforward (Prolog-like) discourse representation structures (DRSs), so they suggest an anti-representationalist move in dynamic semantics.
We argue, nevertheless, that a dynamic semantics based upon DRS-like representations need not necessarily be thought to be non-compositional. Compositionality of a semantic system is always to be interpreted as ''compositionality according to some kind of syntax." What Groenendijk and Stokhof [9] prove is nothing else but the fact that Kamp's semantics is not compositional according to the traditional predicate-logical syntax. Nor is it questionable that this semantics is not compositional according to syntaxes of usual Montague grammars [8] and traditional phrase structure grammars [19] either.
As compositionality is a matter of the relationship between syntax and semantics, non-compositionality of a theory can be fixed in at least two ways: by changing the semantics or by changing the syntax. We argue that what is required is an appropriate (morpho-)syntax, relative to which DRS-like representations are compositional, in the strictest sense. The attractive discourse representation structures themselves inspired me first to raise the possibility of such a syntax [1], and to begin to work out its details (in an isomorphic way to DRSs). The version to be presented here has been called Generative Argument Structure Grammar (GASG), because of the distinguished role of a comprehensive (morpho-) lexical characterization of argument structures [2],[3].
The basic hypothesis is that the morpholexical characterization of (fully inflected) words can be made so rich that each word can determine its potential environments in grammatical sentences; so once all these (lexically predicted) environmental conditions have been satisfied in a ''numeration" of words [6], [7], such a sentence structure is given that determines a phonetic realization, on the one hand, and a (DRS-like) semantic representation, on the other. This idea can be understood as a total generalization of that of categorial grammars and modern PSGs [4: 81]: ''...grammars can be simplified considerably if categories incorporate information about the categories with which they combine." In GASG, each inflected word has its ''own category" (within a lexical inheritance network, of course). Movement can be dispensed with because the scope of a word's environmental conditions may range up to the boundaries of a whole sentence. Moreover, phrase structures can also be dispensed with because the satisfaction of environmental conditions manifests itself as feature unification, accessible to semantics, so grammatical connections need not be realized as constituents.
2. The task of the Prolog program (e.g. [17]) in GASG where each inflected word is assigned a complex morpholexical description concerning the potential environment of the given word in possible grammatical sentences is nothing else but checking the satisfaction of these (unistratal) descriptions (serving as conditions) in proposed sentences (simultaneously). Among the factors to be checked are case, person, number, gender, and other agreement and categorial features of potential environmental words as well as their relative position in the sequence of words.
We argue that the homogeneous treatment of these factors peculiar to GASG promises an effective implementation of computer-aided translation, for instance (especially between languages of ''remote" types).
3. Whereas in the late eighties the possibility of a grammar based on such a rich morpholexical structure like the one sketched above seemed incredible to generative syntacticians, nowadays Chomsky himself [6], [7], [16] declares the following principle of lexical inclusiveness: ''outputs consist of nothing beyond properties of items of the lexicon (lexical features); in other words, ... the interface levels consist of nothing more than arrangements of lexical features." ''Phrase structure theory can be eliminated entirely," ''... movement is driven by morphological requirements," and ''economy conditions exclude 'extra moves' [7] (c.f. ''Perfect Syntax")."
GASG can be regarded as a straightforward implementation of such central ideas of current generative linguistics as Perfect Syntax, the condition of lexical inclusiveness, the elimination of phrase structure grammar, and the idea of a morphology-driven grammar [5],[6],[7]. The cornerstone of my reasoning is that these ideas are not necessarily to be associated with the practice of a redundant ''syntactic encoding" of morphological and intonational information. We regard a movement triggered by morphological checking requirements as an instance of this ''syntactic encoding," and argue for the retainment of these morphological (and other non-syntactic) ''dependencies" on a more abstract level, immediately accessible to semantics.
From a metatheoretical perspective, the elaboration of a grammar like GASG is an attempt to answer the most important question (of different branches) of generative linguistics in the late nineties: Where is the endpoint of the permanent four-decade tendency of transportation of the descriptive power of grammar from syntax to the lexicon? Is the extreme answer, i.e. syntax is nothing else but feature unification, possible? Note that this conception of a ''minimal syntax and maximal lexicon" would provide the conceptually simplest structure for the ''computational system for human language" [6] so we are interested in not giving up elaborating a system like this, or rather, we are interested in trying to construct it.
4. Finally, we would like to repeat that GASG is an attempt to realize the simplest possible grammar: the Perfect Grammar with a maximal lexicon and an entirely minimalized morphosyntax. It might be that a grammar like this cannot exist at all; but even the least chance must not be missed because of the (meta-) theoretical advantages as well as because of the computational advantages that the double Prolog-like representation guarantees.
References:
1. Alberti, G. (1990), 'Quantifiers and the Chronology of Unification.' Paper read at the 3rd Symposium on Logic and Language, Révfülöp, Hungary. Unpublished.

2. Alberti, G. (1996), 'Generative Argument Structure Grammar: A Strictly Compositional Syntax for DRS-Type Representations.' To appear in Acta Linguistica Hungarica.

3. Alberti, G. (1998), 'GASG: Minimal Syntax, Maximal Lexicon.' To appear in Working Papers in the Theory of Grammar. Theoretical Linguistics Programme, Budapest University (ELTE) and Research Institute for Linguistics, Hungarian Academy of Sciences.

4. Borsley, R. D. (1996), Modern Phrase Structure Grammar. Blackwell, Oxford & Cambridge.

5. Brody, M. (1995), `Towards Perfect Syntax'. Working Papers in the Theory of Grammar, Vol. 2., No. 4. Theoretical Linguistics Programme, Budapest University (ELTE) and Research Institute for Linguistics, Hungarian Academy of Sciences.

6. Chomsky, N. (1995a), 'Bare Phrase Structure.' In G. Webelhuth (ed.), Government and Binding Theory and the Minimalist Program. Blackwell, Oxford & Cambridge, pp. 383-439.
7. Chomsky, N. (1995b), 'Categories and Transformations.' In N. Chomsky (ed.), The Minimalist Program. MIT Press, Cambridge, Mass.

8. Dowty, D.R., R.E. Wall, and S. Peters (1981), Introduction to Montague Semantics. Reidel, Dordrecht.

9. Groenendijk, J., and M. Stokhof (1989), 'Dynamic Predicate Logic. Towards a compositional, non-representational semantics of discourse.´ ITLI, Amsterdam.

10. Groenendijk, J., and M. Stokhof (1990), 'Dynamic Montague Grammar.' In L. Kálmán and L. Pólos (eds.), Papers from the 2nd Symposium on Logic and Language. Akadémiai, Budapest, pp. 3-48.

11. Heim, I. (1982), The semantics of definite and indefinite noun phrases. Diss. U. Mass., Amherst.

12. Heim, I. (1983), 'File change semantics and the familiarity theory of definiteness´. In Bä uerle, R., Schwarze, Ch. and Stechow, A. von (eds.) Meaning, use and interpretation of language. Berlin, De Gruyter.

13. Kamp, H. (1981), 'A theory of truth and semantic representation´. In Groenendijk, J., Janssen, T. and Stokhof, M. (eds.). Formal methods in the study of language. Amsterdam: Mathematical Centre.

14. Kamp, H., and U. Reyle (1993), From Discourse to Logic Vol. I. Kluwer Academic Publ.

15. Kálmán, L., and Rádai G. (1996), 'Discontinuous Constituents in Hungarian from a Constructionist Perspective.' To appear in Approaches to Hungarian 6. (ed. C. de Groot and I. Kenesei)

16. Marantz, A. (1995), 'The Minimalist Program.' In G. Webelhuth (ed.), Government and Binding Theory and the Minimalist Program. Blackwell, Oxford & Cambridge, pp. 349-382.

17. Márkusz, Zs. (1988), Prologban programozni könnyû [It is Easy to Program in Prolog]. Novotrade, Budapest.

18. Oehrle, R. T., Bach, E., and Wheeler, D. (1988, ed.), Categorial Grammars and Natural Language Structures. Reidel, Dordrecht.

19. Partee, B. H., Meulen, G. B. ter, and Wall, R. P. (1990), Mathematical Methods in Linguistics. Kluwer Academic Publ.