To design (semi-)supervised models for syntactic and semantic parsing, it is essential to have corpora annotated in deep syntax. The Labex EFL support led to a first annotation phase on the Sequoia corpus (Candito and Seddah, 2012), but we need to broaden this base to for machine learning. Based on the complete annotation guide built for French (Candito et al 2014, Perrier et al 2014), this operation aims to produce deep syntactic annotation (Candito et al, 2014, Perrier et al 2014) on the French Social Media Bank and the French QuestionBank corpus.