Describir: Improving Human Action Recognition in Videos with CNN– sLSTM and Soft Attention Mechanism