No. You’re trying to measure whether the machine is accurately translating his thoughts into characters. He may make errors in spelling or memory. Success is whether the machine spits out the characters he intended, even if they weren’t the ones he was supposed to send.