The problem of learning language models from large text corpora has been widely studied within the computational linguistic community. However, little is known about the performance of these language models when applied to the computer vision domain. In this work, we compare representative models: a window-based model, a topic model, a distributional memory and a commonsense knowledge database,...