Prototype-Regularized Federated Learning for Cross-Domain Aspect Sentiment Triplet Extraction
2026-04-10 • Computation and Language
Computation and Language
AI summaryⓘ
The authors focus on a task called Aspect Sentiment Triplet Extraction, which tries to find aspects, opinions, and feelings in sentences. They point out that previous methods worked on single datasets and couldn't share useful info across different types of data. To fix this, they created a system called PCD-SpanProto that lets different clients share summarized knowledge (called prototypes) without sharing raw data, protecting privacy. Their tests showed this approach works better and uses less communication than older methods.
Aspect Sentiment Triplet ExtractionSentiment AnalysisFederated LearningPrototype LearningCross-Domain LearningPrivacyContrastive RegularizationDomain HeterogeneityModel Aggregation
Authors
Zongming Cai, Jianhang Tang, Zhenyong Zhang, Jinghui Qin, Kebing Jin, Hankz Hankui Zhuo
Abstract
Aspect Sentiment Triplet Extraction (ASTE) aims to extract all sentiment triplets of aspect terms, opinion terms, and sentiment polarities from a sentence. Existing methods are typically trained on individual datasets in isolation, failing to jointly capture the common feature representations shared across domains. Moreover, data privacy constraints prevent centralized data aggregation. To address these challenges, we propose Prototype-based Cross-Domain Span Prototype extraction (PCD-SpanProto), a prototype-regularized federated learning framework to enable distributed clients to exchange class-level prototypes instead of full model parameters. Specifically, we design a weighted performance-aware aggregation strategy and a contrastive regularization module to improve the global prototype under domain heterogeneity and the promotion between intra-class compactness and inter-class separability across clients. Extensive experiments on four ASTE datasets demonstrate that our method outperforms baselines and reduces communication costs, validating the effectiveness of prototype-based cross-domain knowledge transfer.