Aug 2026· Advanced Electromagnetics· Vol 15, pp. 9418-9425· 0 citations· 19 references
TL;DR
A full-chain architecture of “perceptionunderstanding-reasoning-decision-making-interaction” is constructed, covering heterogeneous data ingestion, cross-modal alignment, semantic interoperability, conversational interaction, workflow automation, security control, and efficiency evaluation.
Abstract
The deep integration of the digital economy and artificial intelligence has accelerated the transition of accounting systems from informatization and automation toward intelligence and multimodal processing. Traditional accounting systems are mainly designed for structured data, and therefore show limited ability to parse unstructured documents, images, audio, and video, as well as insufficient integration between business and financial data. This study designs a multimodal intelligent accounting system based on large language models, computer vision, speech recognition, knowledge graphs, intelligent agents, and data governance theory. A full-chain architecture of “perceptionunderstanding-reasoning-decision-making-interaction” is constructed, covering heterogeneous data ingestion, cross-modal alignment, semantic interoperability, conversational interaction, workflow automation, security control, and efficiency evaluation. The system treats OCR, ASR, and RPA as downstream tool plugins coordinated by a domain-specific LLM agent. It also considers the engineering requirements of secure wireless data transmission, electromagnetic-compatible terminal environments, and reliable networked accounting infrastructures. The proposed architecture provides theoretical and practical support for improving accounting efficiency, financial risk control, decision support, and enterprise digital transformation.
With the rapid development of the digital economy and artificial intelligence, the formalization of accounting language has become a crucial factor in promoting intelligent accounting for textile enterprises, enabling more precise financial tracking of raw material costs and textile manufacturing overheads. This paper systematically studies the formalization of accounting language and explores the inherent limitations and practical challenges faced by intelligent systems in understanding accounting language. The inherent ambiguity of accounting language, the complexity of professional judgment, and the differences in cross-cultural contexts hinder intelligent systems from understanding it. Based on the specific characteristics of accounting practices in traditional manufacturing industries such as textiles, this paper proposes establishing a multi-layered formalization system, a rule- and data-driven hybrid intelligent model, and a human-machine collaborative understanding mechanism to improve the understanding of accounting language by intelligent systems, thereby promoting the transformation and upgrading of accounting informatization towards intelligent accounting.
Inspection and testing underpin product safety and regulatory compliance across industries such as manufacturing, healthcare and food and beverage. However, conventional engineering test processes that mainly depend on manual, offline actions, failed to adapt to next generation, automated, digital and environmental-sensitive testing, resulting in very low productivity and high cost for most firms. The automation and digitalization of inspection and testing processes have become a research hotspot in both academia and industry. The natural language processing and computer interpretation of test results have been a focus of AI research as well. Nonetheless, due to the lack of real data, the verification and simulation of real inspection and testing environments are still difficult for researchers. As a solution, we manage to develop a multimodal cognitive test system that fuses the textual regulatory documents and instrument data, through a series of modules and processing flows. Our test system is based on a so-called multimodal cognitive agent, which includes large language model, vision module, knowledge graph and retrieval-augmented generation. We introduce the design, development and application of our test system, which used for rubber heater in a glass factory, and layout the future challenges for the exploration of multimodal agent technology in the test engineering. The project is supported by the Guangxi Key Research and Development Program. This paper was partially presented at the IEEE 2023 International Conference on Intelligent Commerce (ICIC).
Zhiruo Li, Wenlin You, Xinrong Zhang et al.· Frontiers in Computing and I...· 0 citations
Manual monitoring has inherent shortcomings, including low efficiency, inability to achieve real-time data acquisition, and relatively high error rates that prevent 24-hour uninterrupted observation. To solve the above problems, this paper constructs a four-layer intelligent monitoring system on the basis of Internet of Things (IoT), which includes perception layer, transmission layer, processing layer and application layer. The system is designed in accordance with internationally recognized IoT architectural standards, which forms a closed-loop process covering signal acquisition, data transmission, and data analysis. This paper details the key functions, composition, technical principles, and coordination mechanisms of each layer. This architecture integrates edge computing and cloud computing, builds a multi-protocol transmission mechanism, and addresses the limitations of traditional monitoring systems, which rely on overly simplistic data processing models and suffer from prolonged response times. Practical application demonstrates that the four-layer framework offers robust real-time responsiveness and scalability, and can greatly improve monitoring efficiency, demonstrating the practical value of the proposed system.
Mayifei Wu· Applied and Computational En...· 0 citations
The findings advocate for the integration of AI-powered pipelines within ERP systems as a transformative approach to enable scalable, intelligent, and high-fidelity data processing, essential for next- generation enterprise software resilience and performance.
Yuvaraj Kavala· International Journal of Com...· 0 citations
With the rapid advancement of intelligent sensing networks, edge computing, and Electromagnetic Waves, Antennas and Propagation technologies, efficient multimodal information fusion has become a fundamental requirement for data-driven monitoring and decision support in complex cyber–physical systems. To address the challenges of heterogeneous data inconsistency, spatiotemporal misalignment, and limited semantic interaction, this study proposes an intelligent evaluation framework integrating multi-source big data through Digital Twin–Knowledge Graph (DT-KG) fusion and a Multimodal Spatiotemporal Alignment Transformer Network (MSAT-Net). The proposed architecture combines video, audio, and textual information using heterogeneous feature encoding, cross-modal relative position encoding, adaptive gating fusion, and multi-head self-attention to construct unified semantic representations and interpretable quantitative evaluation models. Edge computing and differential privacy mechanisms are incorporated to enable real-time data acquisition while ensuring secure information processing and privacy preservation. Experimental results demonstrate that the proposed system achieves an evaluation accuracy of 92.3%, a Pearson correlation coefficient of 0.887 with ECERS-3 expert assessment, and significant improvements in robustness, inference efficiency, and long-term quality monitoring capability. Beyond preschool education, the proposed multimodal fusion framework provides an effective methodology for distributed sensing, semantic information propagation, adaptive data fusion, and communication-oriented intelligent monitoring, offering valuable engineering references for applications in Electromagnetic Waves, Antennas and Propagation.
Y. Yang, J. Chen, C. Bai· Advanced Electromagnetics· 0 citations
The rapid evolution of intelligent power systems and wireless communication infrastructures has accelerated the demand for efficient natural language interaction with large-scale power marketing databases. To address the limitations of conventional database querying, this study proposes a Retrieval-Augmented Generation (RAG)-enhanced Natural Language-to-SQL (NL2SQL) architecture that integrates domain knowledge retrieval with automated SQL generation. The proposed framework combines intention understanding, hybrid knowledge retrieval, schema-aware query generation, and compliance verification into a unified workflow for intelligent power marketing services. By incorporating database schemas and business rules through RAG, the system effectively reduces semantic ambiguity and improves SQL generation reliability in complex domain-specific scenarios. Case studies involving electricity bill inquiries and abnormal power consumption analysis demonstrate that the architecture achieves over 92% query accuracy while reducing response time to less than one minute. The proposed framework provides an efficient solution for intelligent data interaction in power systems and offers methodological support for communication-assisted smart energy services, where reliable information transmission and large-scale data access are essential. Its architecture also presents a transferable paradigm for knowledge-driven query systems deployed in digitally connected industrial environments.
L. Peng, W. Guo, Z. X. Shen et al.· Advanced Electromagnetics· 0 citations