My belief is that "Deep Network" systems will fail to produce commercially useful results on language just as symbolic systems failed in the face of visual recognition.