Multi-granularity Text Representation and Transformer-Based Fusion Method for Visual Question Answering | IEEE Conference Publication | IEEE Xplore