Answers are grounded in this paper and cite their sources.
Aiming at the difficulty of jointly addressing entity and event extraction, privacy risk, and communication efficiency in power-marketing text, this study proposes FedTrans-KIE, a federated Transformer framework for Edge IoT. The framework uses a shared Transformer encoder for joint entity recognition and event detection, applies client-side per-example gradient clipping with an L2 threshold of 1.0, performs bounded adaptive clipping of client updates, and adds Gaussian noise to the server-aggregated update before release. FedTrans-KIE achieved entity-recognition and event-detection F1 scores of 97.3% and 96.9%, respectively, with an average F1 of 97.1%. With a Gaussian noise multiplier of σ = 0.65, the membership-inference attack success rate decreased from 82.3% to 54.7%, while the average F1 remained 97.0%. These results provide empirical evidence of an improved privacy-utility trade-off. Because the exact cumulative privacy loss depends on the implemented sampling and adaptive-clipping procedure, no numerical ε guarantee is claimed for this setting. The model converged after 51 communication rounds, with a total communication volume of 27.1 GB, 60.5% and 21.7% lower than FedAvg-BERT and FedAvg-Transformer, respectively. Overall, FedTrans-KIE supports accurate and communication-efficient information extraction while reducing empirically observed privacy leakage; protection against an untrusted aggregation server requires additional secure-aggregation mechanisms.
No takes yet. Share an insight, caveat, or question.
Meng et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: