Research map / Video / Video-language understanding
A Graph-Based Framework to Bridge Movies and Synopses
Yu Xiong, Qingqiu Huang, Lingfeng Guo, Hang Zhou, Bolei Zhou, Dahua Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace VSE averaging with the event flow module (ordered shot-sentence alignment) | appearance-node retrieval · MSA (cross-movie) | Recall@1: 4.49 -> 6.80 | Improved |
| replace VSE averaging with the event flow module | appearance-node retrieval · MSA (within-movie) | Recall@1: 21.36 -> 27.67 | Improved |
| add character interaction module graph matching on top of EFM | EFM with appr.+action+cast · MSA (cross-movie) | Recall@1: 21.12 -> 24.15 | Improved |
| add character interaction module graph matching on top of EFM | EFM with appr.+action+cast · MSA (within-movie) | Recall@1: 30.58 -> 31.92 | Mixed |
| add cast (character) nodes to VSE | VSE · MSA (cross-movie) | Recall@1: 5.34 -> 19.05 | Improved |
| add action nodes to VSE | VSE · MSA (cross-movie) | Recall@1: 4.49 -> 5.34 | Improved |
| add subtitle features to the shot representation (ablation) | VSE appr. · MSA (cross-movie) | Recall@10: 19.66 -> 24.51 | Improved |
| limit CIM rescoring to the top 15 candidates instead of 40 (ablation) | EFM+CIM · MSA (cross-movie) | Recall@1: 23.91 -> 24.15 | Improved |
| rerank top 40 instead of 80 candidates (ablation) | EFM+CIM · MSA (cross-movie) | Recall@10: 63.71 -> 62.86 | No clear effect |
| rerank with CIM at all (N=80) versus the previous stage (ablation) | EFM · MSA (cross-movie) | Recall@1: 21.12 -> 23.42 | Improved |
| use propagation degree J=2 rather than J=1 in graph pruning (ablation) | EFM+CIM · MSA (cross-movie) | Recall@1: 23.30 -> 24.15 | Improved |
| increase propagation degree from J=2 to J=3 (ablation) | EFM+CIM · MSA (cross-movie) | Recall@1: 24.15 -> 24.03 | No clear effect |
Closest papers by meaning
- MovieNet: A Holistic Dataset for Movie Understanding (ECCV 2020 · similarity 0.87)
- MovieBench: A Hierarchical Movie Level Dataset for Long Video Generation (CVPR 2025 · similarity 0.84)
- SeriesBench: A Benchmark for Narrative-Driven Drama Series Understanding (CVPR 2025 · similarity 0.83)
- Visual Objectification in Films: Towards a New AI Task for Video Interpretation (CVPR 2024 · similarity 0.82)
- Evaluating Multimodal Narrative Understanding of Popular Hollywood Films (NeurIPS 2026 · similarity 0.82)
- Unveiling the Value of Motion for Cinematic Camera Trajectories (NeurIPS 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
