Główne pojęcia
Introducing negative seed entities enhances ultra-fine-grained semantic comprehension in Entity Set Expansion.
Streszczenie
UltraWiki introduces negative seed entities to address the challenges of representing ultra-fine-grained semantic classes. The dataset includes 50,973 entities and 394,097 sentences across 236 ultra-fine-grained semantic classes. Two frameworks, RetExpan and GenExpan, are proposed to evaluate large language models for Ultra-ESE. Strategies like contrastive learning and retrieval augmentation enhance model performance.
Statystyki
UltraWiki encompasses 50,973 entities and 394,097 sentences.
The dataset includes 236 ultra-fine-grained semantic classes.
Each query is represented with 3-5 positive and negative seed entities.
Cytaty
"Negative seed entities eliminate the semantic ambiguity by contrast between positive and negative attributes."
"To assess model performance in Ultra-ESE and facilitate further research, we constructed UltraWiki."
"Extensive experiments confirm the effectiveness of our proposed strategies."