AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Mining Entity Synonyms with Efficient Neural Set Generation

Jiaming Shen, Ruiliang Lyu, Xiang Ren, Michelle Vanni, Brian Sadler, Jiawei Han

AAAI 2019 · Label-efficient and robust learning › Recognition and long-tail · approach: Multilingual language representation pretraining · 29 citations (OpenAlex)

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 changedWhereMeasuredEffect
use holistic set-instance scoring rather than the strongest reported baselineSynSetMine · WikiARI: 44.78 ($\pm$0.28) -> 56.43 ($\pm$1.31)Improved
use holistic set-instance scoring rather than the strongest reported baselineSynSetMine · NYTARI: 43.92 ($\pm$0.90) -> 44.91 ($\pm$2.16)Improved
use holistic set-instance scoring rather than the strongest reported baselineSynSetMine · PubMedARI: 58.91 ($\pm$0.08) -> 74.33 ($\pm$0.66)Improved
add the Embedding Transformer (ablation)SynSetMine set scorer · WikiARI: 46.48 -> 56.43Improved
add the Embedding Transformer (ablation)SynSetMine set scorer · NYTARI: 39.86 -> 44.91Improved
add the Post Transformer (ablation)SynSetMine set scorer · WikiARI: 1.50 -> 56.43Improved
add the Post Transformer (ablation)SynSetMine set scorer · NYTARI: 0.82 -> 44.91Improved
use Both-250-500 hidden layer sizes instead of Both-100-200 (ablation)SynSetMine set scorer · WikiARI: 49.38 -> 57.34Improved
use Both-250-500 hidden layer sizes instead of Both-100-200 (ablation)SynSetMine set scorer · NYTARI: 37.64 -> 48.89Improved
use Both-300-600 hidden layer sizes instead of Both-250-500 (ablation)SynSetMine set scorer · WikiARI: 57.34 -> 56.26Hurt
use Both-300-600 hidden layer sizes instead of Both-250-500 (ablation)SynSetMine set scorer · NYTARI: 48.89 -> 46.65Hurt
use mixed negative sampling instead of share-token sampling (ablation)SynSetMine training pairs · Wikiclustering performanceImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy