The work presented here is developed within a project devoted to the acquisition of English prosody by French learners, using speech technology modifications, and knowledge about L1 and L2 prosody. Our goal is to provide learners with relevant feedback in an automatic manner by comparing the prosodic cues of their realizations to that of a model. We focus on the English lexical accent and propose methods to correct automatically the learner's realizations. We present our strategy and illustrate it through a concrete example.