Intelligent Archive Management Based on Deep Learning Technology Driven by Artificial Intelligence

To improve the efficiency and intelligence level of an archive management system (AMS) in multimodal data processing, this study proposes and designs an intelligent AMS based on deep learning (DL). However, the traditional AMS faces problems, such as inaccurate data classification, low query efficiency, long response delay, and insufficient system expansibility, when dealing with massive multimodal data. To address these issues, mainstream DL models such as Decoding-Enhanced Bidirectional Encoder Representation from Transformers with Disentangled Attention (DeBERTa), Contrastive Language-Image Pretraining (CLIP), and Swin Transformer are compared with the proposed optimized models. The optimized model’s performance improvement in multimodal archive management tasks has been validated through multidimensional experimental assessments. In comprehensive performance comparison experiments, the optimized model demonstrates excellent performance across several key metrics, including resource consumption, response time, data processing throughput, query efficiency, and fault recovery capability. For instance, the optimized model’s response time in text processing tasks is 98.367 milliseconds (ms), significantly lower than the Swin Transformer’s 156.234 ms. Regarding audio processing tasks, the optimized model’s resource consumption is only 4.387 GigaByte (GB), markedly lower than DeBERTa’s 6.823 GB. Furthermore, in terms of user satisfaction, the proposed model scores as high as 9.238 in text processing, indicating an enhancement in the user experience. Through effectiveness evaluation experiments, this study further confirms the superiority of the optimized model in terms of accuracy, processing delay, self-learning ability, error rate, security assessment, and system scalability. Moreover, the optimized model achieves an accuracy of 94.23% in text processing, nearly 4% higher than DeBERTa, and reduces the error rate in audio processing to 3.78%, showing greater stability and reliability. Therefore, this study provides new solutions for AMS in the fields of multimodal data processing and intelligent management, especially in enhancing system performance, optimizing user experience, and strengthening system security and scalability.