Pour une grammaire du discours. Elaboration d'une méthode ; exemples d'application

摘要

CONSIDERING A GRAMMAR OF DISCOURSE. THE DEVELOPMENT OF A METHOD AND SOME EXAMPLES OF APPLICATION This text sets out to form a statisticaly based grammar of discourse integrating lexical, syntactic and enunciative facts. With reference to the General Resolution of the 1976 CFDT congress and with the help of the factorial analysis of correspondances (where each sentence is coded according to different selected criteria), the following features have been studied: the order, the syntactic structure, the grammatical category of the noun as superficial subject, the relationship between the noun and the category of the main verb, the enunciative structure, the active or passive form, the affirmative or negative form and the modal structure. Such a grammar has two dimensions. It is both a "grammar of frequencies " and a "grammar of sequences". The "grammar of frequencies " considers the text as a homogeneous whole and makes a global study of the syntactic, lexical and enunciative frequencies and the correlations between them throughout the whole corpus. The corpus, homogeneized in such a way, is an abstract projection of the real corpus but the very concept of a probabilistic grammar leads necessarily to this kind of abstraction. Hence the need for a "grammar of sequences", which studies the distribution of preceeding rules within the sequence of sentences and therefore in the linear unfolding of the text.