In this paper we present some experiments on the building of French resources for the TTL POS tagger (Ion, 2007). TTL is a collection of interconnected text preprocessing modules (sentence splitter, tokenizer, tagger, le mmatizer and chunker) with resources for Romanian and English but with no resources available for French. We show how we develop the required POS tagging training corpus and that the average POS tagging accuracy for French exceeds 97% when TTL is trained on this corpus.