Lexicon-grammar

The theoretical basis of Lexicon-grammar is Zellig Harris' distributionalism,[2] and in particular the notion of transformation in the sense of Zellig Harris. In fact, Maurice Gross was a student of Zellig Harris. The conventions for the presentation of grammatical information are intended to be as simple and transparent as possible. This concern comes from Zellig Harris, whose theory is oriented towards the directly observable "surface"; it is a difference from Generative grammar, which normally uses abstract structures such as deep structures.

Lexicon-grammatical method is inspired by experimental science. It emphasizes the collection of facts, and thus the confrontation with the reality of language use, from a quantitative and a qualitative point of view.[3]

Quantitatively: a lexicon-grammar includes a program of systematic description of the lexicon. This involves large-scale work, which can be carried out by teams and not by individual specialists. The exclusive search for general rules of syntax, independent of the lexical material they handle, is denounced as a dead end. This is different from Generative grammar generative grammar, which values the notion of generalization.

Qualitatively: methodological precautions are applied to ensure good reproducibility of the observations, and in particular to guard against the risks associated with constructed examples. One of these precautions is to take as a minimum unit of meaning the basic sentence. Indeed, a word acquires a precise meaning only in a context; moreover, by inserting a word in a sentence, one has the advantage of manipulating a sequence that may be judged acceptable or unacceptable. It is at this price that syntaxico-semantic properties could be considered as defined with sufficient precision to make sense to compare them with the whole lexicon. These precautions have evolved in line with needs and the appearance of new technical means. Thus, from the beginning of the 1990s, the contributors of the lexicon-grammar have been able to use more and more easily the use of attested examples in corpora. This new precaution has simply been added to the previous ones, making the lexicon-grammatical method a method that belongs to both introspective linguistics and corpus linguistics, much as advocated by Fillmore. The American projects FrameNet and VerbNet show a relative convergence towards objectives close to those of Lexicon-grammar.

Lexicon-grammar also requires formalization. The results of the description must be sufficiently formal to allow for :

- a verification by comparison with the reality of language use;

- an application to the automatic processing of languages, and more particularly to deep linguistic processing, in particular through the development of syntax analysers by computer scientists.