Parallel Attention Network with Sequence Matching for Video Grounding