unsupervised learning of visual sense models for polysemous published presentations and documents on DocSlides.