Technical Report: Image Captioning with Semantically Similar Images