No abstract

Jacques SteinlinAlpage Cet article présente le repérage des connecteurs de discours dans le corpus « French Treebank » (FTB) déjà annoté pour la morpho-syntaxe. C'est la première étape dans l'annotation discursive complète de ce corpus. Il s'agit de projeter sur le corpus les éléments répertoriés dans LexConn, lexique des connecteurs du français, et de i ltrer les occurrences de ces éléments qui n'ont pas un emploi discursif mais par exemple un emploi d'adverbe de manière ou de préposition introduisant un complément sous-catégorisé. Plus de 10 000 connecteurs ont été identii és.Mots clés : connecteurs de discours, annotation discursive de corpus, grammaire et discours This paper presents the identii cation of discourse connectives in the "French Treebank" corpus (FTB) already annotated for morpho-syntax. This is the i rst step in the full discursive annotation of this corpus. The method consists in projecting on the corpus the items that are listed in LexConn, a lexicon of French connectives, and then i ltering the occurrences of these elements that do not have a discursive use, but for example are used as an adverb of manner or a preposition introducing a subcategorized complement. More than 10K connectives have been identii ed.

No abstract

No abstract

The paper presents an analysis for a systematic distinction between NPIs and FCIs cross-linguistically. It first relates to a range of analyses that treat NPIs and FCIs as indefinites (in the sense of Kamp (1981) and Heim (1982)) that obey a particular semantic/pragmatic constraint that distinguishes them from plain indefinites. For example, Jayez and Tovena (2005) analyse this constraint as an "equity" constraint which demands that all the entities in the referential domain of the indefinite are equally likely referents for this indefinite. The new contribution of this paper consists in proposing that NPIs and FCIs differ on the fact that this "equity" constraint has a different informational status whether it enters the semantic composition of NPIs or FCIs. This is supported by evidence based on the study of three ways of coercing the interpretation of this kind of items.