With the rapid development of information technology, traditional library information management systems are no longer able to meet the needs of modern users for efficient, convenient, and personalized services. We aim to design and implement an artificial intelligence based library information management system to improve the management efficiency and service quality of the library. The system adopts a layered architecture design, including front-end user interface layer, back-end business logic layer, and data storage layer, to ensure high performance, high availability, and high scalability of the system. The functional modules cover user management, book management, intelligent retrieval, and data analysis, to meet the diverse needs of users. By integrating artificial intelligence technologies such as natural language processing, machine learning, and image recognition, the system has achieved functions such as semantic retrieval, personalized recommendation, and multimodal retrieval, significantly enhancing the user experience. During the development process, the agile development process is followed and technology stacks such as Python, Django, MySQL, etc. are used to ensure efficient development and stable operation of the system. The test results show that the system response time can be maintained within 2 seconds under high concurrency conditions, and user satisfaction reaches 85% . This study provides new perspectives and methods for research in the field of library information management, and provides strong technical support for the digital transformation and intelligent upgrading of libraries, which has important theoretical and practical significance.

The diversity and complexity of video content make video label classification difficult. Different videos may have similar features but belong to different categories, or videos of the same category may have significant differences in presentation. To effectively improve the accuracy of video label classification results, a video label classification algorithm combining CNN(Convolutional Neural Network) and rotated forest is proposed. Classify film and video tags into two stages. In the first stage, the rotation forest algorithm is used to segment the sample set of film and television video labels. Through feature transformation, each subset of samples is transformed into a completely new feature space, and multiple new sample subsets with significant differences are obtained. The AdaBoost algorithm is used to iterate multiple times in the sample set and construct multiple AdaBoost classifiers.The probability averaging method is introduced to fuse the classification results and obtain preliminary label classification results. In the second stage, the film and television video features captured by the quaternion Gabor filtering convolution algorithm and the preliminary classification results of the labels obtained in the first stage are used as inputs for the CNN. L1 regularization is introduced in the fully connected layer to constrain the complexity of the model and prevent overfitting. The film and television video label classification is completed through multiple rounds of iterative training. The test results show that the proposed algorithm has good performance in film and television video label classification and can effectively meet the personalized needs of users.

To address the limitations of existing course evaluation models-specifically their insufficient sensitivity to cross-sentence context and inadequate extraction of semantic importance-this paper proposes a multi-granularity semantic analysis-based model that more accurately captures students' true intentions within textual feedback and supports downstream tasks such as sentiment classification and knowledge extraction. The model integrates pre-trained models and deep neural networks to extract word-level, sentence-level, and part-of-speech-level text feature vectors for semantic analysis and processing. Using course evaluation texts as an example, we conduct experiments and analyses. The model employs both precise and fuzzy matching evaluation methods and incorporates Dropout and the ReLU ( Rectified Linear Unit) activation function to enhance its generalization capability. In the experiments, we improved the model's classification performance by adopting various text preprocessing strategies, including stopwords removal and key term selection. The results indicate that the proposed model excels in sentiment analysis for course evaluations, achieving an accuracy of 92. 53% ,particularly when dealing with ambiguous sentiment boundaries. For course evaluations, the proposed semantic analysis model effectively captures detailed feedback from students, providing an efficient automated evaluation tool for the education sector and optimizing teaching quality.