A Corpus for Visual Question Answering Annotated with Frame Semantic Information

14168

Knowledge Triples