Text this: A Semantic Weight Adaptive Model Based on Visual Question Answering