Intelligent library management system driven by large model

Abstract

The invention belongs to the technical field of library management systems, and particularly relates to a large-model-driven intelligent library management system comprising a semantic analysis module used for inputting book resource data into a large language model for analysis and extracting key semantic knowledge; the clustering module is used for performing clustering division on the semantic vector by using a quadtree segmentation algorithm according to the semantic vector converted by the large language model; the RDF data generation module is used for generating triple RDF data, and creating URI (Uniform Resource Identifier) association links among books through a similarity matching method; the RDF editor module is used for providing a user interaction interface. Mapping the descriptive labels of the vector clusters segmented by the quadtree with the triple RDF data attributes; generating a URI (Uniform Resource Identifier) association link based on a BIBFRAME system according to vector cluster similarity matching of different books; and the knowledge relevance of the whole library collection books is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of library management systems, and in particular relates to a large model-driven smart library management system. Background Art

[0002] With the advent of the digital age, libraries face the dual challenges of managing massive information resources and providing efficient knowledge services. Traditional library management systems have significant limitations in handling complex semantic associations and knowledge organization, making it difficult to meet modern users' needs for precise knowledge discovery and contextual retrieval. In recent years, large language models, such as Deepseek and other domestic open-source large-scale models, have achieved breakthroughs in natural language processing, providing technological possibilities for the innovation of library management systems. Furthermore, the BIBFRAME system, a next-generation bibliographic framework application designed to replace the MARC format, is more suitable for the semantic web environment.

[0003] However, current library management systems mostly use metadata standards such as MARC21 and Dublin Core to describe library collections. These descriptions are usually isolated, and it is difficult to discover the knowledge connections between book resources.

[0004] On the other hand, traditional classification methods often have difficulty in accurately classifying similar knowledge semantics, ignore the multidimensional attributes of knowledge, and a large number of potential knowledge associations are ignored, which seriously limits users' association browsing and retrieval learning capabilities. Summary of the Invention

[0005] In order to solve the technical problems that traditional library management systems have difficulty in discovering knowledge connections between book resources and accurately classifying similar knowledge semantics, the present invention provides a large model-driven intelligent library management system.

[0006] The purpose of the present invention can be achieved through the following technical solutions: The big model-driven smart library management system is characterized by: including a data analysis layer, a BIBFRAME system management layer, and a smart service platform layer; the data analysis layer includes a selection module, a semantic analysis module, and a clustering module; The acquisition and selection module is used to collect various types of book resource data and perform data format conversion and cleaning; The semantic analysis module is used to input book resource data into the large language model for analysis and extraction of key semantic knowledge, and convert the extracted key semantic knowledge into semantic vectors; The clustering module is used to cluster the semantic vectors converted from the large language model using the quadtree segmentation algorithm; The smart service platform layer is located above the BIBFRAME system management layer and is used to provide comprehensive reading services to end readers. The comprehensive reading services include book search, personalized recommendations, loan management, and multi-terminal synchronization: The smart service platform layer also includes a reader interaction module, an intelligent recommendation module, and a cross-system interconnection module: Reader interaction module: provides a unified UI for web, mobile, and self-service terminals; Intelligent recommendation module: calls the vector clusters and URI links generated by the present invention and outputs recommendation results in real time, including similar topics, related citations and hot lists; Cross-system interconnection module: Complying with ISOILL and NCIP protocols, it enables interlibrary loan and single sign-on with other libraries and electronic resource platforms; The BIBFRAME system management layer includes the RDF data generation module and the RDF editor module; The RDF data generation module is used to generate triple RDF data, and then create URI association links between books through similarity matching method; The RDF editor module is used to provide a user interactive interface that enables book catalogers to create, edit, verify, and correct triple RDF data, and then create and modify URI association links between books.

[0007] Preferably, the key semantic knowledge includes book titles, content summaries and application examples.

[0008] Preferably, the large language model is a Deepseek model.

[0009] Preferably, the specific implementation of process 3) includes: 31) Convert the encoded text into input format and feed it into the large language model for forward propagation calculation. The input format includes word ID, attention mask and sentence type identifier; 32) The large language model outputs a tensor of shape [batch_size, sequence_length, hidden_size]. The last four layers of output are extracted and the weighted average is calculated to generate a semantic vector. 33) PCA dimensionality reduction method is used to initially reduce the output 768-dimensional semantic vector to 64 or 32 dimensions.

[0010] Preferably, the specific process of the clustering module includes: a) Use the t-SNE dimensionality reduction method to reduce the dimension of all semantic vectors and map them into a unified two-dimensional space; b) Use the coordinate center point of the two-dimensional space as the root node of the quadtree to construct a quadtree, and then recursively split the two-dimensional space into four quadrants. Each quadrant can be further split; c) setting a threshold condition for triggering further splitting and a maximum depth to limit the splitting; the threshold condition includes the distance (such as covariance) between semantic vectors in the quadrant space exceeding a certain threshold or the number of vectors in the quadrant space exceeding a certain threshold; d) After the quadtree split is completed, each quadrant space represents a leaf node, and each leaf node represents a vector cluster of a classification category; e) The word embedding-based method automatically extracts the keywords that best represent the vector cluster as the descriptive label for subsequent mapping.

[0011] Preferably, the maximum depth is 6-8 layers.

[0012] Preferably, the specific process of the RDF data generation module includes: Map the descriptive label of each book's vector cluster to the corresponding triple RDF data attribute of the book in the BIBFRAME system; Perform similarity matching based on the triple RDF data attributes mapped to each book; Based on the matching results, URI interconnection links between books are created in the BIBFRAME system.

[0013] Preferably, the triple RDF data attributes include book title, content summary and application examples; the triple RDF data attributes correspond to the key semantic knowledge extracted by the large language model; Preferably, the specific process of the RDF editor module includes: Load the book resource data to be edited from the RDF data generation module and provide a user interaction interface; The cataloger selects the triple RDF data attribute of the corresponding book on the user interface for editing, including modifying the descriptive label or correcting the URI associated link object; Perform SHACL or OWL rule verification on this edit, identify, merge, and split related link objects with the same URI, and maintain the consistency of URI related links.

[0014] Preferably, the intelligent recommendation module triggers a recommendation when the minimum similarity is greater than or equal to 0.75.

[0015] Preferably, the cross-system interconnection module is compatible with ISO10160 / 10161 (ILL) and NCIP2.0 protocols.

[0016] Preferably, the reader interaction module supports unified identity authentication for the Web, Android / iOS clients, and self-service borrowing and returning terminals.

[0017] Beneficial effects of the present invention: 1. Leveraging the semantic understanding and analysis capabilities of large language models (such as Deepseek or other domestic open source large models), it effectively captures fine-grained semantic associations in book topics, abstracts, and application examples, with higher accuracy than traditional natural language processing (NLP) paradigms or TF IDF methods.

[0018] 2. By mapping high-dimensional semantic vectors to a fixed low-dimensional space and combining it with a quadtree segmentation algorithm, we achieve multi-level segmentation and clustering optimization of book resources. This clustering method breaks through the rigid limitations of traditional classification methods and can adaptively discover implicit connections between books, forming an interconnected knowledge graph network.

[0019] 3. Map the descriptive labels of the vector clusters after quadtree segmentation to the triple RDF data attributes; generate URI association links based on the BIBFRAME system based on the similarity matching of vector clusters of different books; and improve the knowledge relevance of the entire collection of books.

[0020] 4. The present invention transforms complex semantic associations into an intuitive and personalized reading experience through the intelligent service platform layer, significantly improving readers' knowledge discovery efficiency and collection utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a structural framework diagram of the smart library management system driven by the large model of the present invention; Figure 2 This is a flowchart of the steps of the clustering module in the smart library management system driven by the large model of the present invention; Figure 3 This is a flowchart of the steps of the RDF data generation module in the large model-driven smart library management system of the present invention. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0024] See also Figure 1-Figure 3 As shown in the figure, the big model-driven smart library management system includes a data analysis layer, a BIBFRAME system management layer, and a smart library service platform; the data analysis layer includes a selection module, a semantic analysis module, and a clustering module; (1) Overall process of the smart service platform: a) The system receives user search requests or reading records; b) Call the RDF triple and vector database to perform SPARQL+ANN similarity search; c) Return personalized recommendation results and synchronize them to the user's multiple terminals.

[0025] (2) Intelligent recommendation sub-process: 1) Read user profile (interest topics, borrowing history, reading time); 2) Filter the borrowable status of collections by ranking by cosine similarity of vector clusters; 3) Combine the candidate sets of "topic similarity", "citation association", and "new book release" and output the TopN through Softmax normalization; 4) The results are packaged in XML or JSON-LD format and pushed to the front end via REST API.

[0026] The acquisition and selection module is used to collect various types of book resource data and perform data format conversion and cleaning; During the implementation process, we collected data from various book resource formats and sources, including print books and e-books. We developed various format converters to convert data from various sources (such as MARC21 and UNIMARC) into XML or JSON formats, which serve as internal standards. These formats include fields such as book title, abstract, keywords, application examples, and author. Furthermore, we implemented data cleansing processes based on XML or JSON formats, including regular or rule-based deduplication (such as ISBN deduplication and title similarity detection), missing field completion (using external book metadata APIs such as Open Library), and standardized date and language formats.

[0027] The semantic analysis module is used to input book resource data into the large language model for analysis and extraction of key semantic knowledge, and convert the extracted key semantic knowledge into semantic vectors; the key semantic knowledge includes book titles, content summaries and application examples; The clustering module is used to cluster the semantic vectors converted from the large language model using the Quadtree segmentation algorithm. The BIBFRAME system management layer includes the RDF data generation module and the RDF editor module; The RDF data generation module is used to generate triple RDF data, and then create URI association links between books through similarity matching method; The RDF editor module is used to provide a user interactive interface that enables book catalogers to create, edit, verify, and correct triple RDF data, and then create and modify URI association links between books.

[0028] Specifically, the smart library management system of the present invention adopts a large model-driven design concept and is divided into two core layers: the data analysis layer and the BIBFRAME system management layer. The data analysis layer draws on the idea of ​​using large language models for text semantic vectorization (which has been implemented in large language models).

[0029] The BIBFRAME (Bibliographic Framework) system, developed and released by the U.S. Library of Congress in 2012, is based on Linked Data technology. It emphasizes linking data within the context of the Semantic Web, using resource description models like RDF to provide information and linking book resources via URIs. It aims to develop bibliographic description standards into a Linked Data model.

[0030] The present invention designs a solution that combines a large language model with the BIBFRAME system, providing a modular process of "data collection-semantic analysis-clustering classification-RDF generation-human-computer interaction" to achieve intelligent management of book resources.

[0031] Furthermore, the specific process of the semantic analysis module includes: 1) Input book resource data into a large language model for analysis and extraction of key semantic knowledge; 2) Use the Tokenizer tool provided with the large language model to encode the extracted key semantic knowledge; 3) The encoded text is passed through the Deepseek model’s forward computation to obtain a semantic vector of fixed dimension (e.g., 64 or 32 dimensions); Furthermore, the large language model is a Deepseek model or other domestically produced open-source large-scale model; preferably, other domestically produced open-source large-scale models. Specifically, unlike the traditional Deepseek model, which can only model context from left to right or right to left, other domestically produced open-source large-scale models simultaneously "look" at the words to the left and right of a word in the sequence at all levels, thereby more accurately understanding the semantics of a word or sentence and being more conducive to summarizing and refining the keywords and thematic essence of a book.

[0032] In the specific implementation process, a large language model (Deepseek or other domestic open source large models) is used to analyze and extract key semantic knowledge about book resources. This key semantic knowledge is structured or semi-structured, for example: Book title: includes book title, title, subtitle, etc.

[0033] Content summary: includes chapter information, keywords, theme concepts, etc.

[0034] Application examples: include application areas, industries or examples in the text.

[0035] Then, use the Tokenizer class tool (such as BertTokenizerFast or BertTokenizer) in the large language model that is specifically responsible for converting arbitrary text into a discrete TokenID sequence acceptable to the model; call it to complete the series of processing steps from word segmentation → mapping → packaging. Among them, word segmentation (tokenization) is to perform preliminary splitting, process spaces, punctuation, case normalization, etc., and cut the original text into "preliminary tokens (pre-tokens)."

[0036] Mapping (convert_tokens_to_ids): Assigns a unique integer ID to each sub-word (token); including special tags such as [CLS], [SEP], [MASK], etc.

[0037] Packing (build_inputs_with_special_tokens): Automatically add [CLS] and [SEP] at both ends of the sequence; handle padding ([PAD]), truncation, and attention mask.

[0038] After the above word segmentation, mapping, and packaging, a sequence of integer token IDs is output for use in the Deepseek model's forward computation. This forward computation uses a weighted average of the outputs of the last four Deepseek layers to generate semantic vectors, improving the robustness of the semantic representation. Using vector compression techniques, the high-dimensional original vectors (768 dimensions) of the large language model are reduced to 64- or 32-dimensional vectors through PCA for efficient clustering. A vector database (such as FAISS or Milvus) is then established to support efficient similarity matching computations.

[0039] Furthermore, the specific implementation of process 3) includes: 31) Convert the encoded text into input format and feed it into the large language model for forward propagation calculation. The input format includes word ID, attention mask and sentence type identifier; 32) The large language model outputs a tensor of shape [batch_size, sequence_length, hidden_size]. The last four layers of output are extracted and the weighted average is calculated to generate a semantic vector. 33) PCA dimensionality reduction method is used to initially reduce the output 768-dimensional semantic vector to 64 or 32 dimensions.

[0040] During implementation, the encoded text is converted into an input format acceptable to the Deepseek model, including an input ID, an attention mask, and a tag type ID. These encoded tensors are then fed into a pre-trained Deepseek model for forward propagation. The model processes the input layer by layer, capturing the text's deep semantic features through a self-attention mechanism and a feedforward neural network. For long text (such as abstracts), the system employs a segmented processing strategy, with each segment constrained to the Deepseek maximum sequence length (typically 128K tokens). Each segment is independently passed through the Deepseek model, and the system then integrates the representations of these segments to ensure that important semantic information is not lost.

[0041] The output of the Deepseek model is typically a high-dimensional vector, which requires feature extraction and dimensionality reduction to obtain the specified low-dimensional semantic vector. For single-label text (such as the [CLS] label), the system directly extracts the [CLS] label vector of the last layer of Deepseek as the representation of the entire sentence. For multi-label text, the system extracts the output of the last four layers of Deepseek and calculates a weighted average to enhance the richness of the semantic representation. The weight of each label can be determined by the attention score in the pre-training phase, or a uniform weight can be used. To improve the quality of semantic representation, the system implements a pooling operation selection mechanism that can dynamically select average pooling or maximum pooling strategies based on different types of text content.

[0042] After obtaining the semantic vector representation of the Deepseek model, the system needs to convert it into a semantic vector of a specified dimension in advance: the PCA dimensionality reduction method can be used to perform preliminary dimensionality reduction on the 768-dimensional semantic vector, usually to 64 or 32 dimensions, retaining about 95% of the variance information.

[0043] In a more preferred embodiment, after the dimensionality reduction is completed, the system will perform L2 normalization on the semantic vectors to ensure that the Euclidean length of all vectors is 1, which is convenient for subsequent similarity calculations.

[0044] Furthermore, the specific process of the clustering module includes: a) Use the t-SNE dimensionality reduction method to reduce the dimension of all semantic vectors and map them into a unified two-dimensional space; b) Use the coordinate center point of the two-dimensional space as the root node of the quadtree to construct a quadtree, and then recursively split the two-dimensional space into four quadrants. Each quadrant can be further split; c) Setting a threshold condition for triggering further splitting, and setting a maximum depth or minimum space size to limit splitting; the threshold condition includes the distance (such as covariance) between semantic vectors in the quadrant space exceeding a certain threshold or the number of vectors in the quadrant space exceeding a certain threshold; d) After the quadtree split is completed, each quadrant space represents a leaf node, and each leaf node represents a vector cluster of a classification category; e) Word embedding-based methods (such as ZhKeyBERT) automatically extract the keywords that best represent the vector cluster as the descriptive label for subsequent mapping; In the specific implementation process, the high-dimensional semantic vector is first mapped to a two-dimensional space using the t-SNE or UMAP nonlinear dimensionality reduction algorithm. The t-SNE algorithm is preferably used because it supports efficient mapping of new data without recalculating all the data.

[0045] Secondly, a vectorized two-dimensional coordinate system is established to ensure the unique positioning of semantic vectors in two-dimensional space. Simultaneously, based on the distribution characteristics of the two-dimensional vectors, the optimal position of the initial root node is dynamically calculated (which may not be the exact coordinate center). Threshold conditions for leaf node splitting are set: for example, when the number of vectors within a node exceeds 50-200 or the maximum distance between vectors within a node exceeds a preset threshold, a split is triggered. This prevents efficiency issues caused by severe imbalance in the tree structure.

[0046] In a specific example, the maximum depth may be set to 6-8 layers or the minimum quadrant space size may be set to no less than 1 / 512 of the total space.

[0047] After the quadtree split is completed, ZhKeyBERT and TF-IDF technology are combined to extract 6-8 most representative keyword groups from the book text in each leaf node as the descriptive label; the label of the main leaf node represents a broader topic, and the sub-leaf node represents a more specialized sub-topic, thus achieving hierarchical descriptive labeling.

[0048] Furthermore, the specific process of the RDF data generation module includes: Map the descriptive label of each book's vector cluster to the corresponding triple RDF data attribute of the book in the BIBFRAME system; Perform similarity matching based on the triple RDF data attributes mapped to each book; Based on the matching results, create URI interconnection links between books in the BIBFRAME system; Furthermore, the triple RDF data attributes include book title, content summary and application examples; the triple RDF data attributes correspond to the key semantic knowledge extracted by the large language model; In the specific implementation process, the RDF data generation module establishes a mapping rule base of the BIBFRAME system, for example: Map the bf:Topic attribute to its corresponding book title: including the book title, title, subtitle, etc.

[0049] Map the bf:work attribute to its corresponding content summary: including chapter information, keywords, theme concepts, etc.

[0050] Map the bf:instance attribute to its corresponding application instance: including the application field, industry, or instance in the text.

[0051] This automatically generates triple-based RDF data attributes and adopts the Turtle (TTL) format as the RDF storage and exchange standard. Based on the vector cluster similarity association mechanism for different books, a minimum similarity threshold is set at 0.75-0.9. Unique URI identifiers are generated for each book's associated links, following a unified naming convention. Multi-level URI association link types are established, including "topic similarity," "keyword similarity," and "reference citation" link types.

[0052] Furthermore, the specific process of the RDF editor module includes: Load the book resource data to be edited from the RDF data generation module and provide a user interaction interface; The cataloger selects the triple RDF data attribute of the corresponding book on the user interface for editing, including modifying the descriptive label or correcting the URI associated link object; Perform SHACL or OWL rule verification on this edit, identify, merge, and split related link objects with the same URI, and maintain the consistency of URI related links.

[0053] In its implementation, the RDF Editor module is a Linked Data (RDF) editor for cataloging, supporting multi-institutional collaborative cataloging. Users can use the RDF Editor within their resource templates to edit books, revealing and optimizing the triplet RDF data attributes and their relationships through descriptive tags. The created RDF data can be converted to MARC or XML formats and can also be integrated into third-party systems and loaded into ILS and LSPs.

[0054] In summary, this invention, based on a large-scale model-driven intelligent library management system, leverages the semantic understanding and analysis capabilities of large language models (such as Deepseek or other domestically developed open-source large-scale models) to effectively capture fine-grained semantic associations within book topics, abstracts, and application examples, achieving higher accuracy than traditional natural language processing (NLP) paradigms or TFIDF methods. Furthermore, by mapping high-dimensional semantic vectors to a fixed low-dimensional space and combining it with a quadtree partitioning algorithm, it achieves multi-level segmentation and clustering optimization of book resources. This clustering approach overcomes the rigid limitations of traditional classification methods and can adaptively discover implicit associations between books, forming an interconnected knowledge graph network.

[0055] The system also incorporates the triple RDF data model of the BIBFRAME system framework, mapping the descriptive labels of the vector clusters after quadtree segmentation to the triple RDF data attributes. Based on the similarity matching of vector clusters of different books, URI association links are generated based on the BIBFRAME system. This improves the knowledge relevance of the entire collection and builds a high-quality linked data network of book resources.

[0056] In the description of the specification, the descriptions with reference to the terms "further", "in the specific implementation process", "specific example", etc. mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. The above content is merely an example and explanation of the structure of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the scope of protection of the present invention.

Claims

1. A large-scale model-driven smart library management system, characterized by: It includes data analysis layer, BIBFRAME system management layer and intelligent service platform layer; the data analysis layer includes mining and selection module, semantic analysis module and clustering module; The acquisition and selection module is used to collect various types of book resource data and perform data format conversion and cleaning; The semantic analysis module is used to input book resource data into the large language model for analysis and extraction of key semantic knowledge, and convert the extracted key semantic knowledge into semantic vectors; The clustering module is used to cluster the semantic vectors converted from the large language model using the quadtree segmentation algorithm; The smart service platform layer is located above the BIBFRAME system management layer and is used to provide comprehensive reading services to end readers. The comprehensive reading services include book search, personalized recommendations, loan management, and multi-terminal synchronization: The smart service platform layer also includes a reader interaction module, an intelligent recommendation module, and a cross-system interconnection module: Reader interaction module: provides a unified UI for web, mobile, and self-service terminals; Intelligent recommendation module: calls the vector clusters and URI links generated by the present invention and outputs recommendation results in real time, including similar topics, related citations and hot lists; Cross-system interconnection module: Complying with ISOILL and NCIP protocols, it enables interlibrary loan and single sign-on with other libraries and electronic resource platforms; The BIBFRAME system management layer includes the RDF data generation module and the RDF editor module; The RDF data generation module is used to generate triple RDF data, and then create URI association links between books through similarity matching method; The RDF editor module is used to provide a user interactive interface that enables book catalogers to create, edit, verify, and correct triple RDF data, and then create and modify URI association links between books.

2. The large model-driven intelligent library management system according to claim 1, characterized in that: The key semantic knowledge includes book titles, content summaries and application examples.

3. The large model-driven intelligent library management system according to claim 1, characterized in that: The specific process of the semantic analysis module includes: 1) Input book resource data into a large language model for analysis and extraction of key semantic knowledge; 2) Use the Tokenizer tool provided with the large language model to encode the extracted key semantic knowledge; 3) The encoded text is passed through the forward calculation of the Deepseek model to obtain a semantic vector of fixed dimension.

4. The large model-driven intelligent library management system according to claim 3 is characterized by: The specific implementation of process 3) includes: 31) Convert the encoded text into input format and feed it into the large language model for forward propagation calculation. The input format includes word ID, attention mask and sentence type identifier; 32) The large language model outputs a tensor of shape [batch_size, sequence_length, hidden_size]. The last four layers of output are extracted and the weighted average is calculated to generate a semantic vector. 33) PCA dimensionality reduction method is used to initially reduce the output 768-dimensional semantic vector to 64 or 32 dimensions.

5. The large model-driven intelligent library management system according to claim 3 or 4, characterized in that: The large language model is a Deepseek model.

6. The large model-driven intelligent library management system according to claim 1, characterized in that: The specific process of the clustering module includes: a) Use the t-SNE dimensionality reduction method to reduce the dimension of all semantic vectors and map them into a unified two-dimensional space; b) Use the coordinate center point of the two-dimensional space as the root node of the quadtree to construct a quadtree, and then recursively split the two-dimensional space into four quadrants. Each quadrant can be further split; c) setting a threshold condition for triggering further splitting and a maximum depth to limit the splitting; the threshold condition includes the distance between semantic vectors in the quadrant space exceeding a certain threshold or the number of vectors in the quadrant space exceeding a certain threshold; d) After the quadtree split is completed, each quadrant space represents a leaf node, and each leaf node represents a vector cluster of a classification category; e) The word embedding-based method automatically extracts the keywords that best represent the vector cluster as the descriptive label for subsequent mapping.

7. The large model-driven intelligent library management system according to claim 6, characterized in that: The maximum depth is 6-8 layers.

8. The large model-driven intelligent library management system according to claim 1, characterized in that: The specific process of the RDF data generation module includes: Map the descriptive label of each book's vector cluster to the corresponding triple RDF data attribute of the book in the BIBFRAME system; Perform similarity matching based on the triple RDF data attributes mapped to each book; Based on the matching results, URI interconnection links between books are created in the BIBFRAME system.

9. The large model-driven intelligent library management system according to claim 8, characterized in that: The triple RDF data attributes include book title, content summary and application examples; the triple RDF data attributes correspond to key semantic knowledge extracted by the large language model.

10. The large model-driven intelligent library management system according to claim 1, characterized in that: The specific process of the RDF editor module includes: Load the book resource data to be edited from the RDF data generation module and provide a user interaction interface; The cataloger selects the triple RDF data attribute of the corresponding book on the user interface for editing, including modifying the descriptive label or correcting the URI associated link object; Perform SHACL or OWL rule verification on this edit, identify, merge, and split related link objects with the same URI, and maintain the consistency of URI related links.

11. The system according to claim 1, wherein: The intelligent recommendation module triggers a recommendation when the minimum similarity is greater than or equal to 0.

75.

12. The system according to claim 11, characterized in that: The cross-system interconnection module is compatible with ISO10160 / 10161 (ILL) and NCIP2.0 protocols.

13. The system according to claim 12, characterized in that: The reader interaction module supports unified identity authentication for the Web, Android / iOS clients, and self-service borrowing and returning terminals.

Images