AI Paper Atlas

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

A Joint Model for Definition Extraction with Syntactic Connection and Semantic Consistency

Amir Veyseh, Franck Dernoncourt, Dejing Dou, Thien Nguyen

AAAI 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Multilingual language representation pretraining · 36 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
jointly add sentence classification to sequence labeling (ablation)Ours · Textbook and ContractF1: 48.8 -> 50.6; 68.5 -> 71.7Improved
add GCN over the dependency tree (ablation)Ours · Textbook and ContractF1: 48.8 -> 50.6; 68.5 -> 71.7Improved
add dependency path prediction (ablation)Ours · Textbook and ContractF1: 48.1 -> 50.6; 68.2 -> 71.7Improved
add direct local semantic consistency (ablation)Ours · Textbook and ContractF1: 49.6 -> 50.6; 70.5 -> 71.7Improved
add indirect local semantic consistency (ablation)Ours · Textbook and ContractF1: 49.8 -> 50.6; 69.6 -> 71.7Improved
add global semantic consistency (ablation)Ours · Textbook and ContractF1: 49.0 -> 50.6; 69.1 -> 71.7Improved
remove DPP and increase GCN from 2 to 3 layers (ablation)Ours · Textbook and ContractF1: 50.6 -> 48.5; 71.7 -> 68.0Hurt
remove DPP and increase GCN from 2 to 4 layers (ablation)Ours · Textbook and ContractF1: 50.6 -> 48.0; 71.7 -> 67.3Hurt
use the joint model instead of CVTSequence labeling · WCL, W00, Textbook, and ContractF1: 82.5 -> 83.3; 58.9 -> 60.6; 47.3 -> 50.6; 68.2 -> 71.7Improved
use the joint model instead of SASequence classification · WCL, W00, Textbook, and ContractF1: 94.2 -> 99.5; 57.4 -> 67.2; 64.2 -> 70.3; 84.1 -> 91.7Improved
initialize embeddings with BERTOurs · WCL and W00 sequence labelingF1: 83.3 -> 85.3; 60.6 -> 66.9Improved
initialize embeddings with BERTOurs · Textbook sequence labelingF1: 50.6 -> 54.0Improved

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