Research map / Video / Video-language understanding
On Semantic Similarity in Video Retrieval
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use semantic-similarity evaluation instead of instance-based evaluation | video retrieval benchmarks · MSR-VTT, YouCook2, EPIC-KITCHENS | evaluation robustness: ~30 geometric mean for EPIC-KITCHENS, ~6.0 for MSR-VTT, ~5.0 for YouCook2 | Helped |
| Train CE with semantic proxy-based relevance instead of instance-only relevance (ablation) | CE on YouCook2 · YouCook2 | nDCG | Helped |
| Use a lower relevance threshold during semantic training (ablation) | CE trained with semantic proxies · YouCook2 | nDCG | Helped |
| Train with the same semantic proxy used for evaluation (ablation) | CE trained with semantic proxies · YouCook2 | nDCG | Helped |
| Vary the semantic training threshold: BoW peaks at T=0.2; PoS, Syn and MET peak around T=0.4 (ablation) | CE trained with semantic proxies · YouCook2 | nDCG: T=0.2; T=0.4 | Mixed |
| Use BERT or Word2Vec caption similarity as the semantic proxy (ablation) | semantic video retrieval evaluation · YouCook2, MSR-VTT, EPIC-KITCHENS | semantic relevance quality | Hurt |
| Use visual-feature similarity as the semantic proxy (ablation) | semantic video retrieval evaluation · YouCook2, MSR-VTT, EPIC-KITCHENS | semantic relevance quality | Hurt |
| Use METEOR relevance with Random retrieval | Random baseline · MSR-VTT, YouCook2 | nDCG | Mixed |
| Evaluate with semantic proxies rather than instance-based metrics | retrieval baselines · YouCook2, MSR-VTT, EPIC-KITCHENS | relative baseline performance | Mixed |
Closest papers by meaning
- Mitigating Semantic Collapse in Partially Relevant Video Retrieval (NeurIPS 2025 · similarity 0.90)
- Bridging Information Asymmetry in Text-video Retrieval: A Data-centric Approach (ICLR 2025 · similarity 0.89)
- Video-ColBERT: Contextualized Late Interaction for Text-to-Video Retrieval (CVPR 2025 · similarity 0.89)
- Text Proxy: Decomposing Retrieval from a 1-to-N Relationship into N 1-to-1 Relationships for Text-Video Retrieval (AAAI 2025 · similarity 0.88)
- Bridging the Semantic Granularity Gap Between Text and Frame Representations for Partially Relevant Video Retrieval (AAAI 2025 · similarity 0.87)
- ConViS-Bench: Estimating Video Similarity Through Semantic Concepts (NeurIPS 2025 · similarity 0.87)
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.