Intérêt et pièges de la Toile en tant que corpus pour la recherche en linguistique (sous l’angle de recherches sur la valence verbale)

It may seem a crazy bet or at least a challenge to exploit such a corpus as the Web for linguistic studies. We recognize the strange, nearly mysterious and indeed imperfect status of the Web corpus. This paper aims to expose, nevertheless, how it is possible to develop a not so complex questioning strategy to achieve new sophisticated syntactic and semantic descriptions of French verbs. The subject here especially treated is the verb alternations, including the lexical means. Even if we cannot easily appeal to statistical analysis to exploit the Web corpus, the Web is very useful to study the frequent variations, but gives also easier access to rare or new phrases, which reflect the advantages of such a mechanism of alternation: among others, it offers easy topicalisation, by the means of subjectivization, of rather many elements of a verbal sentence. This progress can also be precious for translation.