Federated Query Rewriting for Conversational AI: Privacy-Preserving, Cross-Channel Retrieval on Voice and Web
DOI:
https://doi.org/10.22399/ijcesen.4297Keywords:
Federated learning, Conversational AI, Retrieval-Grounded Query Rewriting, Differential Privacy , Vertex AI, Non-IID DataAbstract
We present a federated framework for retrieval-grounded query rewriting that serves a single Dialogflow CX agent across telephony and web. Local, adapter-based rewriters are trained per tenant and aggregated with secure aggregation; optional (ε, δ)-differential privacy bounds the privacy loss. A channel-consistency regularizer aligns voice↔web rewrites using only local statistics. On embedding retrieval with Vertex AI Matching Engine, our best federated configuration improves nDCG@10 by 6.8% and MRR by 5.2% over a centralized baseline while keeping P95 latency ≤ 280 ms. We describe a privacy-accounted pipeline with Firestore DLP memory, Apigee X controls, and BigQuery/Cloud Trace evaluation, showing that federated optimization can enhance accuracy and robustness under non-IID traffic without centralizing transcripts.
References
[1] Salvi, R. M., & Barman, P. K. (2025). Evolving architectures and long-horizon planning in multi-agent conversational AI: A decade in review. The American Journal of Interdisciplinary Innovations and Research, 7(7), 106–122. https://doi.org/10.37547/tajiir/Volume07Issue07-10
[2] A. Sarkar and L. Vajpayee, “Augmenting the FedProx Algorithm by Minimizing Convergence,” arXiv.org, 2024. Available: https://arxiv.org/abs/2406.00748
[3] Nji, F. N., Salvi, R. M., Tirumala, S., Wang, J., & Zheng, X. (2024). Evaluation of traditional and deep clustering algorithms for multivariate spatio-temporal data. Lawrence Livermore National Laboratory. https://doi.org/10.2172/2519314
[4] K. Joelle, S. Cabrera, B. Miguel, and D. Veloso, “Explainable Knowledge Synthesis in Organisations: A Graph RAG Framework for Internal Knowledge Management Internship Report Master in Modelling, Data Analysis and Decision Support Systems Supervised by,” 2025. Available: https://repositorio-aberto.up.pt/bitstream/10216/169509/2/742060.pdf
[5] Z. Wang et al., “Breaking Secure Aggregation: Label Leakage from Aggregated Gradients in Federated Learning,” IEEE INFOCOM 2024 - IEEE Conference on Computer Communications, pp. 151–160, May 2024, Available: https://doi.org/10.1109/infocom52122.2024.10621090.
[6] [6] C. Liu, N. Bastianello, W. Huo, Y. Shi, and K. H. Johansson, “A survey on secure decentralized optimization and learning,” arXiv.org, 2024. Available: https://arxiv.org/abs/2408.08628
[7] Salvi, R. M. (2025). Omnichannel conversational search: Maintaining context and consistency across voice and web interfaces. International Journal of Applied Mathematics, 38(8s), 1100–1114. https://doi.org/10.12732/ijam.v38i8s.630
[8] Henna Kokkonen et al., “Autonomy and Intelligence in the Computing Continuum: Challenges, Enablers, and Future Directions for Orchestration,” arXiv (Cornell University), May 2022, Available: https://doi.org/10.48550/arxiv.2205.01423.
[9] P. S N, D. K. J. B. Saini, N. Shelke, A. Pimpalkar, G. H. Kumar, and V. V, “Privacy-Preserving and Scalable Secure Aggregation for Federated Learning in Edge Computing,” 2025 Second International Conference on Cognitive Robotics and Intelligent Systems (ICC - ROBINS), pp. 182–188, Jun. 2025, Available: https://doi.org/10.1109/icc-robins64345.2025.11086126.
[10] E. Jo, D. A. Epstein, H. Jung, and Y.-H. Kim, “Understanding the Benefits and Challenges of Deploying Conversational AI Leveraging Large Language Models for Public Health Intervention,” Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, pp. 1–16, Apr. 2023, Available: https://doi.org/10.1145/3544548.3581503.
[11] M. Xu, L. Wen, J. Liao, H. Wu, K. Ye, and C. Xu, “Auto-scaling Approaches for Cloud-native Applications: A Survey and Taxonomy,” arXiv.org, 2025. https://arxiv.org/abs/2507.171.
[12] G. Kaushal, “Throughput vs Latency Graph | BrowserStack,” BrowserStack, Jul. 02, 2025. https://www.browserstack.com/guide/throughput-vs-latency
[13] R. Xu, N. Baracaldo, and J. Joshi, “Privacy-Preserving Machine Learning: Methods, Challenges and Directions,” arXiv:2108.04417 [cs], Sep. 2021, Available: https://arxiv.org/abs/2108.04417
[14] Mirco Planamente, Chiara Plizzari, Simone Alberto Peirone, B. Caputo, and A. Bottino, “Relative Norm Alignment for Tackling Domain Shift in Deep Multi-modal Classification,” International journal of computer vision, vol. 132, no. 7, pp. 2618–2638, Feb. 2024, Available: https://doi.org/10.1007/s11263-024-01998-9.
[15] A. Deshpande et al., “C-STS: Conditional Semantic Textual Similarity,” arXiv.org, 2023. Available: https://arxiv.org/abs/2305.15093
[16] F. Zhou and H. Chen, “Cross-Modal Translation and Alignment for Survival Analysis,” Thecvf.com, pp. 21485–21494, 2023, Accessed: Nov. 10, 2025. [Online]. Available: http://openaccess.thecvf.com/content/ICCV2023/html/Zhou_Cross-Modal_Translation_and_Alignment_for_Survival_Analysis_ICCV_2023_paper.html
[17] Faruque, O., Nji, F. N., Cham, M., Salvi, R. M., Zheng, X., & Wang, J. (2023). Deep spatiotemporal clustering: A temporal clustering approach for multi-dimensional climate data. In Machine Learning and Knowledge Discovery in Databases (ECML PKDD 2023, Applied Data Science Track), LNCS 14175 (pp. 76–91). Springer. https://doi.org/10.1007/978-3-031-43430-3_6
[18] M. I. Khalid, M. Ahmed, and J. Kim, “Enhancing Data Protection in Dynamic Consent Management Systems: Formalizing Privacy and Security Definitions with Differential Privacy, Decentralization, and Zero-Knowledge Proofs,” Sensors, vol. 23, no. 17, p. 7604, Jan. 2023, Available: https://doi.org/10.3390/s23177604.
[19] Faruque, O., Nji, F. N., Cham, M., Salvi, R. M., Zheng, X., & Wang, J. (2023). Deep spatiotemporal clustering: A temporal clustering approach for multi-dimensional climate data. In Machine Learning and Knowledge Discovery in Databases (ECML PKDD 2023, Applied Data Science Track), LNCS 14175 (pp. 76–91). Springer. https://doi.org/10.1007/978-3-031-43430-3_6
[20] A. Arora, “Comprehensive Cloud Security Strategies for Protecting Sensitive Data in Hybrid Cloud Environments,” SSRN Electronic Journal, Jan. 2025, Available: https://doi.org/10.2139/ssrn.5268180.
[21] R. M. Salvi, Spatio-temporal multivariate weather data clustering using DBSCAN and K-Medoids methods, M.S. thesis, Univ. Maryland, Baltimore County, 2023.
[22] C. Kaplan, “Inherent trade-offs in privacy-preserving machine learning,” Hal.science, Nov. 2024, Available: https://theses.hal.science/tel-04874804.
[23] F. Stathopoulou, A. Ferikoglou, M. Katsaragakis, D. Masouros, S. Xydis, and D. Soudris, “SynergAI: Edge-to-Cloud Synergy for Architecture-Driven High-Performance Orchestration for AI Inference,” arXiv.org, 2025. Available: https://arxiv.org/abs/2509.12252
[24] R. Zhang, L. Nie, C. Zhao, and Q. Chen, “Achieving Semantic Consistency Using BERT: Application of Pre-training Semantic Representations Model in Social Sciences Research,” SSRN Electronic Journal, Jan. 2025, Available: https://doi.org/10.2139/ssrn.5043698.
[25] Salvi, R., Islam, M. F., Chowdhury, S. H., Podell, J., Hu, P., Badjatia, N., & Chen, L. (2022). Nowcasting PSH-AM: Towards real-time assessment of paroxysmal sympathetic hyperactivity using continuous vital sign measurements in neurocritical units. Proceedings of the AMIA Annual Symposium (abstract). No DOI assigned. Index records: DBLP and UMBC Lab page.in *Proc. AMIA Annu. Symp.*, 2022.
[26] S. Das, S. R. Chowdhury, N. Chandran, D. Gupta, Satya Lokam, and R. Sharma, “Communication Efficient Secure and Private Multi-Party Deep Learning,” Proceedings on Privacy Enhancing Technologies, 2025. Available: https://petsymposium.org/popets/2025/popets-2025-0010.php
[27] Nji, F. N., Salvi, R. M., Tirumala, S., Wang, J., & Zheng, X. (2022). Evaluation of clustering algorithms for spatio-temporal multivariate weather data. Lawrence Livermore National Laboratory. https://doi.org/10.2172/1990001
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 International Journal of Computational and Experimental Science and Engineering
This work is licensed under a Creative Commons Attribution 4.0 International License.
