Using Images to Find Context-Independent Word Representations in Vector Space