This article reports the evaluation of the integration of data from a syntactic-semantic lexicon, the Lexicon-Grammar of French, into a syntactic parser. We show that by changing the set of labels for verbs and predicational nouns, we can improve the performance on French of a non-lexicalized probabilistic parser.