Research map / Adversarial robustness and security / Attacks on text and graphs
Improving Hypernymy Prediction via Taxonomy Enhanced Adversarial Learning
Read the paperCheck 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 |
|---|---|---|---|
| add adversarial training with Microsoft Concept Graph (AS-TEAL vs S-TEAL) (ablation) | S-TEAL · BLESS supervised | accuracy: 0.95 -> 0.96 | Helped |
| add adversarial training with taxonomy (AS-TEAL vs S-TEAL) (ablation) | S-TEAL · ENTAILMENT | accuracy: 0.87 -> 0.91 | Helped |
| add adversarial training with taxonomy (AS-TEAL vs S-TEAL) (ablation) | S-TEAL · ANIMAL | accuracy: 0.89 -> 0.92 | Helped |
| add adversarial training with taxonomy (AS-TEAL vs S-TEAL) (ablation) | S-TEAL · PLANT | accuracy: 0.93 -> 0.94 | Helped |
| add adversarial training with taxonomy (AS-TEAL vs S-TEAL) (ablation) | S-TEAL · VEHICLE | accuracy: 0.91 -> 0.93 | Helped |
| add supervised classifier on projections (S-TEAL vs U-TEAL) (ablation) | U-TEAL · Chinese FD | F1: 0.65 -> 0.69 | Helped |
| add supervised classifier on projections (S-TEAL vs U-TEAL) (ablation) | U-TEAL · Chinese BK | F1: 0.83 -> 0.83 | No clear effect |
| number of hidden layers >= 4 or hidden units far from embedding dimension | U-TEAL · BLESS | accuracy | Hurt |
| use unsupervised U-TEAL vs strongest prior (Roller et al. 2018) (ablation) | Roller et al. (2018) · BLESS | accuracy: 0.96 -> 0.96 | No clear effect |
| use unsupervised U-TEAL vs Nguyen et al. (2017) (ablation) | Nguyen et al. (2017) · WBLESS | accuracy: 0.87 -> 0.88 | Helped |
| U-TEAL vs best prior GLE method FR | FR (Vulic et al. 2017) · HyperLex nouns | Spearman rho: 0.283 -> 0.463 | Helped |
| S-TEAL vs Nguyen et al. (2017) without taxonomy | Nguyen et al. (2017) · ENTAILMENT | accuracy: 0.91 -> 0.87 | Hurt |
Closest papers by meaning
- On the Transitivity of Hypernym-Hyponym Relations in Data-Driven Lexical Taxonomies (AAAI 2017 · similarity 0.86)
- ExTaSem! Extending, Taxonomizing and Semantifying Domain Terminologies (AAAI 2016 · similarity 0.86)
- Hypernym Detection Using Strict Partial Order Networks (AAAI 2020 · similarity 0.85)
- GANTEE: Generative Adversarial Network for Taxonomy Enterance Evaluation (AAAI 2023 · similarity 0.84)
- Towards Robust Knowledge Unlearning: An Adversarial Framework for Assessing and Improving Unlearning Robustness in Large Language Models (AAAI 2025 · similarity 0.82)
- Language Models as Hierarchy Encoders (NeurIPS 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.