Research map / Image generation / GANs
An Interpretable Generative Adversarial Approach to Classification of Latent Entity Relations in Unstructured Sentences
Shiou Tian Hsu, Changsung Moon, Paul Jones, Nagiza Samatova
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 |
|---|---|---|---|
| use learned rationales instead of one proxy-attention rationale (ablation) | Encoder · SemEval-2010 Task 8 | macro-F1: 84.5 -> 89.5 | Improved |
| use learned rationales instead of two proxy rationales (ablation) | Encoder · SemEval-2010 Task 8 | macro-F1: 85.3 -> 89.5 | Improved |
| use learned rationales instead of three proxy rationales (ablation) | Encoder · SemEval-2010 Task 8 | macro-F1: 87.8 -> 89.5 | Improved |
| increase number of proxy rationales from one to three (ablation) | Proxy rationales + Encoder · SemEval-2010 Task 8 | macro-F1: 84.5 -> 87.8 | Improved |
| rationale-based model vs proxy attention model | Our model · SemEval-2010 Task 8 | macro-F1: 87.5 -> 89.5 | Improved |
| rationale-based model vs best prior (ATT-Pooling-CNN) | Our model · SemEval-2010 Task 8 | macro-F1: 88.0 -> 89.5 | Improved |
| rationale-based model vs engineered SVM | Our model · SemEval-2010 Task 8 | macro-F1: 82.2 -> 89.5 | Improved |
| rationale-based model vs best dependency-tree model (DRNNs) | Our model · SemEval-2010 Task 8 | macro-F1: 85.6 -> 89.5 | Improved |
| provide ground-truth rationale labels from the start of training | Our model · SemEval-2010 Task 8 | macro-F1 | Improved |
| provide ground-truth rationales to converged/converging models | Our model · SemEval-2010 Task 8 | macro-F1: 1-1.5% improvement | Mixed |
| combine Selector and Generator / train end-to-end (ablation) | Our model · SemEval-2010 Task 8 | macro-F1 | Hurt |
Closest papers by meaning
- SEE: Syntax-Aware Entity Embedding for Neural Relation Extraction (AAAI 2018 · similarity 0.88)
- Prototypical Representation Learning for Relation Extraction (ICLR 2021 · similarity 0.86)
- Joint Entity and Relation Extraction with a Hybrid Transformer and Reinforcement Learning Based Model (AAAI 2020 · similarity 0.86)
- Integrating Relation Constraints with Neural Relation Extractors (AAAI 2020 · similarity 0.86)
- A Generative Model of Words and Relationships from Multiple Sources (AAAI 2016 · similarity 0.86)
- Self-Attention Enhanced Selective Gate with Entity-Aware Embedding for Distantly Supervised Relation Extraction (AAAI 2020 · similarity 0.85)
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.
