Improving Few-Shot Named Entity Recognition with Causal Interventions

Improving Few-Shot Named Entity Recognition with Causal Interventions

Few-shot Named Entity Recognition (NER) systems are designed to identify new categories of entities with a limited number of labeled examples. A major challenge encountered by these systems is overfitting, particularly pronounced in comparison to tasks with ample samples. This overfitting predominan...

Full description

Bibliographic Details
Main Authors: Zhen Yang, Yongbin Liu, Chunping Ouyang, Shu Zhao, Chi Zhu
Format: Article
Language:English
Published: Tsinghua University Press 2024-12-01
Series:Big Data Mining and Analytics
Subjects:
few-shot named entity recognition (ner)
causal inference
prototypical network
Online Access:https://www.sciopen.com/article/10.26599/BDMA.2024.9020052

Internet

https://www.sciopen.com/article/10.26599/BDMA.2024.9020052

Similar Items