핵심 개념
SocialPET improves few-shot stance detection by incorporating social network structure into pattern generation.
초록
Stance detection in social media is challenging due to limited labeled data.
SocialPET leverages language models and social network structure for improved performance.
Outperforms baseline models in identifying instances of the 'against' class.
Experimentation on P-Stance and Multi-target datasets demonstrates the efficacy of SocialPET.
SocialPET introduces socially informed knowledge into the pattern generation process.
Results show consistent improvement over baseline models, particularly in the 'Against' class.
Analysis reveals variations in performance across different targets based on social network structure.
Jaccard scores indicate differences in social network overlap between supporters and opponents of each target.
SocialPET shows promise in enhancing stance detection through social network insights.
통계
SocialPET는 기존 모델 PET보다 성능을 향상시키며 'Against' 클래스에서 특히 뛰어난 성과를 보입니다.
SocialPET는 소셜 네트워크 구조를 활용하여 패턴 생성에 사회적 지식을 통합합니다.
인용구
"Our work advances research in few-shot stance detection by introducing SocialPET."
"SocialPET proves the effectiveness of leveraging language models and social network structure."