Research map / Vision-language and multimodal / Human and embodied modeling
Good News, Everyone! Context Driven Entity-Aware Captioning for News Images
Ali Furkan Biten, Lluis Gomez, Marcal Rusinol, Dimosthenis Karatzas
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 |
|---|---|---|---|
| Add article context with sentence-level average encoding | Show Attend Tell captioning pipeline · GoodNews | CIDEr: 12.15% -> 12.75% | Improved |
| Add named-entity insertion rather than omit entities (ablation) | Show Attend Tell · GoodNews | CIDEr: 1.67% -> 10.03% | Improved |
| Use context-based entity insertion instead of random insertion (ablation) | Show Attend Tell · GoodNews | CIDEr: 10.03% -> 12.15% | Improved |
| Use attention-guided entity insertion instead of context-based insertion (ablation) | Ours (Avg.) · GoodNews | exact-match entity precision: 8.17 -> 9.19 | Improved |
| Use attention-guided entity insertion instead of context-based insertion (ablation) | Ours (Avg.) · GoodNews | CIDEr: 12.75% -> 12.70% | No clear effect |
| Use simple average rather than weighted-average article encoding (ablation) | Ours + CtxIns · GoodNews template captions | CIDEr: 17.132% -> 17.577% | Improved |
| Use simple average rather than TBB article encoding (ablation) | Ours + CtxIns · GoodNews template captions | CIDEr: 16.624% -> 17.577% | Improved |
| Use attention-guided entity insertion rather than context-based insertion (ablation) | Ours (Avg.) · GoodNews | partial-match entity recall: 17.88 -> 19.48 | Improved |
| Use average article encoding rather than weighted average (ablation) | Ours + CtxIns · GoodNews | CIDEr: 12.37% -> 12.75% | Improved |
| Use average article encoding rather than TBB (ablation) | Ours + CtxIns · GoodNews | CIDEr: 12.70% -> 12.75% | No clear effect |
| Use visual and textual context rather than visual-only captioning | Show Attend Tell + CtxIns · GoodNews | CIDEr: 12.15% -> 12.75% | Improved |
| Use the proposed visual-plus-textual model rather than Show Attend Tell + CtxIns | GoodNews captioning · GoodNews human evaluation | preference rate: 53% | Improved |
Closest papers by meaning
- Transform and Tell: Entity-Aware News Image Captioning (CVPR 2020 · similarity 0.93)
- Show, Interpret and Tell: Entity-Aware Contextualised Image Captioning in Wikipedia (AAAI 2023 · similarity 0.90)
- Knowledge Completes the Vision: A Multimodal Entity-aware Retrieval-Augmented Generation Framework for News Image Captioning (AAAI 2026 · similarity 0.90)
- Panoptic Captioning: An Equivalence Bridge for Image and Text (NeurIPS 2025 · similarity 0.86)
- CAPability: A Comprehensive Visual Caption Benchmark for Evaluating Both Correctness and Thoroughness (NeurIPS 2025 · similarity 0.86)
- BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs (CVPR 2025 · similarity 0.86)
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.
