Research map / Vision-language and multimodal / Visual question answering
Improving Event Causality Recognition with Multiple Background Knowledge Sources Using Multi-Column Convolutional Neural Networks
Canasai Kruengkrai, Kentaro Torisawa, Chikara Hashimoto, Julien Kloetzer, Jong-Hoon Oh, Masahiro Tanaka
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 single-column CNN with multi-column CNN (Base) | CNN-SENT · Event causality test set | average precision: 43.76 -> 49.34 | Improved |
| add all background knowledge sources (BP+WH+CL) (ablation) | MCNN Base · Event causality test set | average precision: 49.34 -> 55.13 | Improved |
| add short binary patterns (BP) (ablation) | MCNN Base · Event causality test set | average precision: 49.34 -> 54.32 | Improved |
| add why-QA answers (WH) (ablation) | MCNN Base · Event causality test set | average precision: 49.34 -> 52.03 | Improved |
| add clue-term sentences (CL) (ablation) | MCNN Base · Event causality test set | average precision: 49.34 -> 52.27 | Improved |
| add why-QA answer patterns (WH) as SVM features | Hashimoto14 SVM · Event causality test set | average precision: 47.32 -> 43.41 | Hurt |
| add clue-term sentence patterns (CL) as SVM features | Hashimoto14 SVM · Event causality test set | average precision: 47.32 -> 40.11 | Hurt |
| add short binary patterns (BP) as SVM features | Hashimoto14 SVM · Event causality test set | average precision: 47.32 -> 47.39 | No clear effect |
| remove BP-overlapping short patterns from WH and CL inputs (ablation) | Base+BP+WH+CL · Event causality test set | average precision: 55.13 -> 55.42 | Improved |
| limit patterns to 15 for SVM baseline (ablation) | Hashimoto14+BP+WH · Event causality test set | average precision: 47.52 -> 47.88 | No clear effect |
| use Wikipedia-trained instead of in-domain word embeddings (ablation) | Base+BP+WH+CL · Event causality test set | average precision: 55.13 -> 54.84 | No clear effect |
| use random instead of pre-trained word embeddings (ablation) | Base+BP+WH+CL · Event causality test set | average precision: 55.13 -> 48.48 | Hurt |
Closest papers by meaning
- MECD: Unlocking Multi-Event Causal Discovery in Video Reasoning (NeurIPS 2024 · similarity 0.84)
- A Joint Framework with Heterogeneous-Relation-Aware Graph and Multi-Channel Label Enhancing Strategy for Event Causality Extraction (AAAI 2024 · similarity 0.83)
- Integrating Background Knowledge for Scalable Causal Discovery (NeurIPS 2026 · similarity 0.83)
- Mastering Context-to-Label Representation Transformation for Event Causality Identification with Diffusion Models (AAAI 2024 · similarity 0.82)
- Two Causally Related Needles in a Video Haystack (NeurIPS 2025 · similarity 0.82)
- Multimodal Event Causality Reasoning with Scene Graph Enhanced Interaction Network (AAAI 2024 · 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.
