GTLR: Graph-Based Transformer with Language Reconstruction for Video Paragraph Grounding | IEEE Conference Publication | IEEE Xplore