TRIE ++ : Towards End-to-End Information Extraction from Multiple Categories of Visually Rich Documents (Under Review)