AXON-E: A Cognitive Audit Framework for Scalable Oversight of Massive Enforcement Video Archives
Hongguanghua  Gaomiaomiao  Yujinwen  Jiangziyan  Huzhoupeng

Shangrao Tobacco Company, Jiangxi Province Shangrao,334000；

Abstract: The comprehensive recording of administrative enforcement has created petabyte-scale video archives, yet leveraging this data for effective, closed-loop supervision remains a critical challenge for digital governance. Manual spot-checks are profoundly inadequate, rendering the vast majority of recordings “dormant data” that fails to provide its intended value for oversight. To address this, this paper introduces AXON-E (Atomized eXpertise Organization Network for Enforcement), a back-end cognitive audit framework. Its core methodology involves the deep deconstruction of unstructured regulatory documents (e.g., “Evaluation Standards for On-site Enforcement Audio-visual Records”) into a machine-readable knowledge base via Knowledge Atomization and a novel Eight-Dimensional Enforcement Knowledge Coordinate System. The framework employs a high-throughput, asynchronous pipeline that uses multimodal AI to parse recordings into structured “enforcement event streams” for automated compliance verification. Deployment evidence reveals that AXON-E can audit 100% of recordings, dramatically increasing the issue detection rate from less than 5% to over 85%. This provides a powerful engine for scalable accountability and data-driven governance, marking a paradigm shift from reactive spot-checks to proactive, comprehensive oversight.

Keywords: Cognitive Audit; Enforcement Supervision; Regulatory Technology (RegTech); Knowledge Atomization; Multimodal AI; Retrieval-Augmented Generation (RAG)

References

[1] Cheng, H., et al. (2024). “A Survey on Large Multimodal Models for Video Understanding.” arXiv preprint arXiv:2407.03159.

[2] Lewis, P., et al. (2020). “Retrieval-augmented generation for knowledge-intensive NLP tasks.” Advances in Neural Information Processing Systems, 33.

[3] Gao, Y., et al. (2024). “Retrieval-Augmented Generation for Large Language Models: A Survey.” Communications of the ACM, 67(6), 65-75.

[4] Yao, Z., et al. (2024). “RAG vs. Fine-tuning: A Comprehensive Survey on Knowledge Injection in LLMs.” arXiv preprint arXiv:2407.06993. (Discusses hybrid search and metadata filtering as key RAG strategies).

[5] Arslanian, H., & Fischer, F. (2022). The Future of Finance: The Impact of FinTech, AI, and Crypto on Financial Services. Palgrave Macmillan.

[6] Xi, Z., et al. (2023). “The Rise and Potential of Large Language Model Based Agents: A Survey.” arXiv preprint arXiv:2309.07864.

[7] Radford, A., et al. (2023). “Robust speech recognition via large-scale weak supervision.” Proceedings of the 40th International Conference on Machine Learning.

[8] Bredin, H., et al. (2020). “Pyannote.audio: neural building blocks for speaker diarization.” ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[9] Engin, Z., & van der Torre, L. (2024). “AI-supported governance: A new paradigm for public administration.” Data & Policy, 6, e12.

[10] Jobin, A., Ienca, M., & Vayena, E. (2023). “The global landscape of AI ethics guidelines.” Nature Machine Intelligence, 5(9), 974-985.