image2vec learning word and image representations for reasoning published presentations and documents on DocSlides.