Three-dimensional enterprise knowledge base construction method and system

Abstract

The invention relates to the field of multi-modal modeling, and provides a three-dimensional enterprise knowledge base construction method and system. The method comprises the following steps: classifying and marking knowledge assets to be constructed according to a preset dimension system, and extracting semantic feature information corresponding to multiple dimension knowledge nodes to obtain a multi-dimensional cross knowledge framework; knowledge nodes in the multi-dimensional cross knowledge framework are stored in a graph database, a knowledge association network is established, and a dynamic knowledge graph is obtained; collecting user feature data, and generating a user knowledge demand portrait according to the user feature data; performing matching calculation on the user knowledge demand portrait and knowledge nodes in the dynamic knowledge graph, and screening matching results to obtain a target knowledge push list; and determining an access permission level according to the role permission attribute of the user, performing permission verification on the knowledge nodes in the target knowledge pushing list based on the access permission level, and generating personalized knowledge content. According to the invention, the knowledge management efficiency, the delivery accuracy and the security are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of multi-modal modeling, and particularly relates to a three-dimensional enterprise knowledge base construction method and system. BACKGROUND

[0002] At present, enterprise knowledge base systems have been widely used in the field of knowledge management. These systems mainly realize the storage and sharing of enterprise knowledge through basic functions such as document management, classified retrieval, and permission control. Existing knowledge base systems usually organize knowledge content in a tree directory structure or a label classification manner, and users obtain the required information through keyword search or directory browsing. Some systems also introduce knowledge graph technology to establish the association between knowledge nodes through entity recognition and relationship extraction, trying to break down information silos and improve knowledge retrieval efficiency. In terms of permission management, existing systems mostly use role-based access control mechanisms, assign different access permissions according to the user's department and position, and record the user's operation behavior through logs. At the same time, some enterprise knowledge bases also integrate community functions, allowing employees to exchange knowledge through question and answer, comments, and other ways, promoting the explicitness of tacit knowledge.

[0003] However, the existing technology still has significant deficiencies in knowledge presentation, knowledge delivery, and knowledge sharing. First, in the dimension of knowledge presentation, existing systems mostly use a single directory tree or label system to display knowledge, lacking the ability to cross-present knowledge from multiple dimensions such as business processes, personnel roles, application scenarios, and resource types, making it difficult for users to grasp the associated relationships and application scenarios of knowledge from a global perspective, and unable to quickly locate cross-domain and cross-departmental associated knowledge, reducing the efficiency of knowledge acquisition. Second, in terms of knowledge delivery accuracy, existing systems mainly rely on user-initiated search or simple collaborative filtering recommendation, lacking deep understanding of user knowledge needs and scenario adaptation capabilities, and unable to accurately push knowledge according to user job characteristics, work tasks, and real-time scenarios, resulting in blindness and inefficiency in knowledge delivery, and users often need to repeatedly filter through massive amounts of knowledge to find the real content they need. Third, in terms of the balance between knowledge sharing and secrecy, the permission control mechanism of existing systems is relatively static, difficult to adapt to the needs of enterprise organizational structure adjustment and project team dynamic changes, and lacks fine-grained traceability capabilities for knowledge access behavior, promoting knowledge sharing while unable to effectively protect sensitive knowledge at different levels and manage the entire chain of traceability. SUMMARY

[0004] In view of the deficiencies of the above prior art, the purpose of the application is to provide a three-dimensional enterprise knowledge base construction method and system.

[0005] The application provides a three-dimensional enterprise knowledge base construction method, comprising:

[0006] S1: classifying and labeling the knowledge assets to be constructed according to a preset dimension system, extracting semantic feature information corresponding to multiple dimension knowledge nodes through natural language processing technology, and obtaining a multi-dimensional cross-knowledge framework;

[0007] S2: storing the knowledge nodes in the multi-dimensional cross-knowledge framework into a graph database, establishing a knowledge association network according to the association relationship types between the knowledge nodes, and obtaining a dynamic knowledge graph;

[0008] S3: collecting user feature data, and generating a user knowledge demand portrait according to the user feature data;

[0009] S4: matching and calculating the user knowledge demand portrait with the knowledge nodes in the dynamic knowledge graph, filtering the matching results according to the current scene context of the user, and obtaining a target knowledge push list;

[0010] S5: determining an access permission level according to the role permission attribute of the user, performing permission verification on the knowledge nodes in the target knowledge push list based on the access permission level, and generating personalized knowledge content with an access permission identifier.

[0011] According to the three-dimensional enterprise knowledge base construction method provided by the application, the preset dimension system in step S1 includes business dimension, personnel dimension, scene dimension and resource dimension.

[0012] According to the three-dimensional enterprise knowledge base construction method provided by the application, the process of constructing a multi-dimensional knowledge framework in step S1 includes:

[0013] S11: classifying enterprise knowledge assets according to multiple dimensions in the preset dimension system, and labeling attribute information of knowledge nodes in multiple dimensions, to obtain a knowledge set of multiple dimensions;

[0014] S12: performing natural language processing on the knowledge nodes in the knowledge set, extracting keywords, topic labels and entity information of the knowledge nodes, and obtaining semantic feature information of the knowledge nodes;

[0015] S13: cross-fusing the knowledge sets of multiple dimensions based on the semantic feature information, establishing an association index between the dimensions, and forming the multi-dimensional cross-knowledge framework.

[0016] According to the three-dimensional enterprise knowledge base construction method provided by the application, step S2 further includes:

[0017] S21: defining association relationship types between the knowledge nodes according to the semantic feature information of the knowledge nodes in the multi-dimensional cross-knowledge framework, and obtaining a knowledge relationship network structure;

[0018] S22: Store the knowledge nodes and the knowledge relationship network structure in the multidimensional cross-knowledge framework in a graph database in the form of a graph structure to construct and obtain a basic knowledge graph;

[0019] S23: Establish a linking and jumping mechanism between knowledge nodes in the basic knowledge graph to obtain a dynamic knowledge graph.

[0020] According to the method for constructing a three-dimensional enterprise knowledge base provided by the present invention, the relationship types in step S22 include inclusion relationship, dependency relationship, reference relationship and similarity relationship.

[0021] According to the method for constructing a three-dimensional enterprise knowledge base provided by the present invention, step S3 further includes:

[0022] S31: Collect user identity information and behavioral data to obtain basic user characteristics;

[0023] S32: Obtain user interaction content in the community, and use natural language processing technology to segment and extract the topic of the interaction content to obtain user knowledge demand characteristics;

[0024] S33: The user's basic characteristics and the user's knowledge needs characteristics are fused together and dynamically updated in combination with changes in the user's state to generate the user's knowledge needs profile.

[0025] According to the method for constructing a three-dimensional enterprise knowledge base provided by the present invention, step S4 further includes:

[0026] S41: Classify and label the knowledge nodes in the dynamic knowledge graph, and extract the feature labels of the knowledge nodes through machine learning algorithms to obtain a labeled knowledge base;

[0027] S42: Calculate the similarity between the user knowledge demand profile and the knowledge nodes in the tagged knowledge base, filter out candidate knowledge nodes whose similarity exceeds a preset threshold, and obtain preliminary matching results;

[0028] S43: Obtain the user's current scene context information, score the relevance of the preliminary matching result based on the scene context information, and output the target knowledge push list based on the score result.

[0029] According to the method for constructing a three-dimensional enterprise knowledge base provided by the present invention, step S5 further includes:

[0030] S51: Query the permission mapping rules based on the role permission attributes to obtain the user's access permission scope;

[0031] S52: Extract the confidentiality attributes of multiple knowledge nodes in the target knowledge push list, match and verify the confidentiality attributes with the access permission range, and obtain the permission verification result;

[0032] S53: Based on the permission verification result, filter the knowledge nodes in the target knowledge push list, retain the knowledge nodes that the user has the right to access, and generate the personalized knowledge content with the access permission identifier.

[0033] The method for constructing a three-dimensional enterprise knowledge base according to the present invention further includes:

[0034] S6: Collect user actions on accessing the personalized knowledge content across the entire process and write the action event information into the log database to obtain knowledge flow records.

[0035] This invention also provides a three-dimensional enterprise knowledge base construction system, comprising:

[0036] Extraction module: Used to classify and label the knowledge assets to be constructed according to a preset dimension system, and extract semantic feature information corresponding to knowledge nodes of multiple dimensions through natural language processing technology to obtain a multi-dimensional cross-knowledge framework;

[0037] Construction module: Used to store knowledge nodes in the multidimensional cross-knowledge framework into a graph database, establish a knowledge association network according to the types of relationships between knowledge, and obtain a dynamic knowledge graph;

[0038] Description module: Used to collect user feature data and generate a user knowledge need profile based on the user feature data;

[0039] Matching module: used to perform matching calculations between the user's knowledge demand profile and the knowledge nodes in the dynamic knowledge graph, filter the matching results according to the user's current scenario context, and obtain a target knowledge push list;

[0040] Hierarchical module: used to determine the access permission level based on the user's role permission attributes, perform permission verification on the knowledge nodes in the target knowledge push list based on the access permission level, and generate personalized knowledge content with access permission identifiers.

[0041] This invention provides a method and system for constructing a three-dimensional enterprise knowledge base. By building a multi-dimensional cross-knowledge framework and generating a dynamic knowledge graph, it can present enterprise knowledge assets in a three-dimensional way from multiple dimensions such as business processes, personnel roles, application scenarios, and resource types. It breaks through the limitations of the traditional single directory tree structure of knowledge bases, so that users are no longer limited to a single perspective when searching for knowledge. They can quickly discover cross-departmental and cross-domain related knowledge, significantly shorten the knowledge retrieval path, and reduce the cost of information acquisition.

[0042] This invention establishes a panoramic view of the knowledge network by creating relationships such as inclusion, dependency, reference, and similarity between knowledge nodes. Users can clearly understand the context and application scenarios of knowledge, avoiding redundant knowledge construction and resource waste caused by information silos, and improving the reuse rate and value conversion efficiency of knowledge assets. Simultaneously, by collecting user identity information, behavioral data, and community interaction content, and utilizing natural language processing technology to deeply mine user needs, this invention can accurately characterize users' knowledge needs and intelligently match them with the user's current work scenario and task context. This significantly reduces the time cost for users to sift through massive amounts of knowledge, improves the accuracy and timeliness of knowledge acquisition, and promotes the efficient flow of knowledge and collaborative innovation within enterprises. Furthermore, this invention utilizes dynamic permissions... The control mechanism and end-to-end access logging ensure hierarchical confidentiality management while guaranteeing knowledge sharing. This satisfies the need for knowledge openness in cross-departmental collaboration and effectively prevents unauthorized access and leakage of sensitive knowledge. Access permissions are automatically adjusted when user roles change or projects end, avoiding security risks caused by lagging access management. Furthermore, by recording the complete trajectory of user access to knowledge, this invention enables enterprises to audit and analyze knowledge usage, promptly detect abnormal access behavior, and provide reliable technical support for knowledge security management. This access data can also be used to continuously optimize knowledge recommendation strategies, forming a virtuous cycle of knowledge management. Ultimately, this builds a secure, efficient, and sustainably evolving enterprise knowledge management system, comprehensively improving the enterprise's knowledge management level. Attached Figure Description

[0043] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. It is obvious that the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings.

[0044] Figure 1 A schematic diagram of a method for constructing a three-dimensional enterprise knowledge base is provided in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of a three-dimensional enterprise knowledge base construction system provided in an embodiment of the present invention.

[0046] Figure reference numerals: 100, extraction module; 200, construction module; 300, description module; 400, matching module; 500, hierarchical module. Detailed Implementation

[0047] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0048] Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts disclosed in this invention.

[0049] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The terms "installed," "connected," and "linked" should be interpreted broadly; for example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0050] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of methods and systems consistent with some aspects of the invention as detailed in the appended claims.

[0051] The embodiments of the present invention are described below with reference to the figures.

[0052] like Figure 1 As shown, this invention provides a method for constructing a three-dimensional enterprise knowledge base, including:

[0053] S1: The knowledge assets to be constructed are classified and labeled according to a preset dimensional system. Semantic feature information corresponding to knowledge nodes in multiple dimensions is extracted through natural language processing technology to obtain a multi-dimensional cross-knowledge framework.

[0054] Furthermore, the purpose of step S1 is to classify and label the knowledge assets to be constructed according to a preset dimensional system, and extract the semantic feature information corresponding to the knowledge nodes of multiple dimensions through natural language processing technology to obtain a multi-dimensional cross-knowledge framework.

[0055] The preset dimension system in step S1 includes business dimension, personnel dimension, scenario dimension and resource dimension.

[0056] Furthermore, the pre-defined dimension system includes business dimension, personnel dimension, scenario dimension, and resource dimension. Through these four dimensions, a four-dimensional cross-framework of "business-personnel-scenario-resource" is constructed to effectively sort out the relationship between knowledge and target users.

[0057] The process of constructing the multidimensional knowledge framework in step S1 includes:

[0058] S11: Classify the enterprise's knowledge assets according to multiple dimensions in the preset dimension system, and label the attribute information of the knowledge nodes under multiple dimensions to obtain a knowledge set of multiple dimensions.

[0059] In step S11, the present invention classifies enterprise knowledge assets according to multiple dimensions in a preset dimension system, and labels the attribute information of knowledge nodes under multiple dimensions to obtain a knowledge set of multiple dimensions.

[0060] In terms of business dimension, this invention decomposes the enterprise value chain into R&D, production, sales, and after-sales service, with each business link as a business node. Each piece of knowledge is labeled with its corresponding business link identifier, knowledge type tag, and business relevance value. The business relevance value represents the closeness between the knowledge content and the business link, ranging from 0 to 1. This labeling method yields a business-dimensional knowledge set. In terms of personnel dimension, this invention extracts the expert skills, project experience, and department affiliation information of knowledge contributors and users, recording the creator's employee ID, name, department code, job level, list of professional skill tags, and list of projects participated in. It also records the target user group of the knowledge content, associating personnel information with knowledge nodes to obtain a personnel-dimensional knowledge set. In terms of scenario dimension, this invention… This invention identifies knowledge application scenarios such as customer complaint handling and new employee onboarding training. For each application scenario, it defines a scenario name, scenario description, triggering conditions, and required knowledge type. Knowledge content is then bound to the scenario to form a knowledge package for that scenario, resulting in a scenario-dimensional knowledge set. At the resource dimension, the invention categorizes structured data, unstructured documents, multimedia files, and external links. For structured data, it labels the database name, table name, field name, and data update time. For unstructured documents, it extracts the file name, file size, creation time, modification time, and file format. For multimedia files, it records the media type, duration or size, resolution, and encoding format. For external links, it records the link address, link description, and link validity status. Simultaneously, it labels the source information and version information of each knowledge resource, resulting in a resource-dimensional knowledge set.

[0061] S12: Perform natural language processing on the knowledge nodes in the knowledge set to extract keywords, topic tags, and entity information of the knowledge nodes, and obtain the semantic feature information of the knowledge nodes.

[0062] Step S12 performs natural language processing on the knowledge nodes in the knowledge set, extracting keywords, topic tags, and entity information to obtain semantic feature information of the knowledge nodes. This invention uses natural language processing technology to parse the text portion of the knowledge content. First, the document title and body content are extracted. The text content is then input into a word segmenter for Chinese word segmentation to obtain independent word units. Part-of-speech tagging is then performed to identify the part-of-speech type of each word, and nouns and noun phrases are extracted as candidate keywords. This invention uses the TF-IDF algorithm to calculate the importance weight of each candidate keyword. Term frequency (TF) represents the number of times a word appears in the current document divided by the total number of words in the current document, and inverse document frequency (IDF) represents the logarithm of the total number of documents divided by the number of documents containing that word. The TF value and IDF value are multiplied to obtain the TF-IDF weight. The top 10 words with the highest TF-IDF weights are selected as the keywords of the knowledge node. This invention employs the Latent Dirichlet Allocation (LDA) topic model for topic tag extraction. All documents in the knowledge base are input into the LDA model as a corpus, with 50 topics. During model training, the topic distribution and word distribution for each document and topic are iteratively calculated. After training, 50 topics are obtained. The probability distribution of each knowledge node across topics is calculated, and the top three topics with the highest probabilities are selected as the topic tags for that knowledge node. This invention also performs named entity recognition to extract entity information from knowledge content. It uses a deep learning-based named entity recognition model, employing a Bidirectional Long Short-Term Memory (BiLSTM) network combined with a Conditional Random Field (CRF) architecture. The BiLSTM network receives the word sequence after word segmentation, converts each word into a word vector representation, and captures the context information of each word through forward and backward LSTMs, respectively. The CRF layer receives the context representation vector, learns the transition probabilities of entity annotation, and outputs the entity tag sequence for each word. The identified entities are grouped by type, recording lists such as person entity lists, project entity lists, and technology entity lists. The above processing generates semantic feature information for each knowledge node, including a list of keywords and their TF-IDF weights, a list of topic tags and their probability values, and an entity information dictionary.

[0063] S13: Based on the semantic feature information, cross-integrate knowledge sets of multiple dimensions, establish inter-dimensional correlation indexes, and form the multi-dimensional cross-knowledge framework.

[0064] In step S13, this invention cross-integrates knowledge sets from multiple dimensions based on semantic feature information, establishes inter-dimensional association indexes, and forms a multi-dimensional cross-knowledge framework. Specifically, this invention establishes a cross-mapping between business dimensions and personnel dimensions. For each business step in the business dimension, it searches for a list of personnel who create or use knowledge in that business step, creates a personnel association table for each business step, and records personnel number, personnel name, role type, and association strength. The association strength is calculated based on the amount of knowledge created and the frequency of knowledge use by the personnel in that business step. Subsequently, this invention establishes a cross-mapping between business dimensions and scenario dimensions. For each business step, it identifies the application scenarios that will be triggered and creates a scenario association table for each business step, recording scenario name, scenario triggering frequency, and scenario priority. At the same time, this invention also establishes a cross-mapping between scenario dimensions and resource dimensions. For each application scenario, it searches for the knowledge resources required for that scenario and creates a resource association table for each application scenario, recording resource type, resource file path, resource priority, and resource status. This invention also establishes a cross-mapping between personnel dimensions and scenario dimensions. For each personnel role, it identifies the application scenarios commonly used by that role and creates a scenario usage table for each personnel role, recording scenario name, usage frequency, and most recent usage time. By establishing cross-mapping in the four directions mentioned above, we can create inter-dimensional association indexes. These association indexes are stored in the form of relational database tables, including business-person association tables, business-scenario association tables, scenario-resource association tables, and personnel-scenario association tables. These association indexes are then integrated with the knowledge sets of the four dimensions to form a multi-dimensional cross-knowledge framework.

[0065] S2: Store the knowledge nodes in the multidimensional cross-knowledge framework into a graph database, establish a knowledge association network according to the types of relationships between knowledge, and obtain a dynamic knowledge graph.

[0066] Furthermore, step S2 stores the knowledge nodes in the multidimensional cross-knowledge framework into a graph database, establishes a knowledge association network based on the types of relationships between knowledge points, and obtains a dynamic knowledge graph. The graph database uses a graph structure of nodes and edges to store data, where nodes represent entities and edges represent relationships between entities.

[0067] Step S2 further includes:

[0068] S21: Based on the semantic feature information of knowledge nodes in the multidimensional cross-knowledge framework, define the types of association relationships between knowledge nodes to obtain the knowledge relationship network structure.

[0069] Further, in step S21, the present invention defines the types of associations between knowledge nodes based on the semantic feature information of knowledge nodes in the multidimensional cross-knowledge framework, thereby obtaining the knowledge relationship network structure. Specifically, the present invention first identifies the inclusion relationship between knowledge nodes. By analyzing the hierarchical structure and directory organization of knowledge nodes, it traverses all knowledge nodes and checks the parent node attribute of each node. If the parent node attribute value of node A is the identifier of node B, then an inclusion relationship is established between node B and node A. Subsequently, the present invention identifies dependency relationships. By analyzing the preconditions and reference chains of knowledge nodes, it extracts the explicitly declared preconditions of knowledge requirements in each knowledge node, uses regular expressions to match these preconditions, extracts the names of other knowledge nodes mentioned therein, searches for the corresponding knowledge node identifiers in the knowledge base, and establishes a dependency relationship between the current node and the preconditions. Then, it identifies citation relationships by parsing the reference list and in-text citation marks in the knowledge nodes to extract the document. The reference list at the end searches for knowledge nodes in the knowledge base that match the reference titles. If a matching node is found, a citation relationship is established between the current node and the cited node. In addition, this invention also identifies similarity relationships, converting the keyword list, topic tag list, and entity information dictionary of each knowledge node into feature vectors. Keywords and their TF-IDF weights are represented as sparse vectors, topic tags and their probability values ​​are represented as dense vectors, and entity information is converted into one-hot encoded vectors. These are concatenated into a comprehensive feature vector, and the cosine similarity between the comprehensive feature vectors of any two knowledge nodes is calculated. A similarity threshold of 0.6 is set, meaning that when the cosine similarity between two knowledge nodes is greater than 0.6, a similarity relationship is established between them, resulting in a knowledge relationship network structure.

[0070] S22: Store the knowledge nodes and the knowledge relationship network structure in the multidimensional cross-knowledge framework in a graph database in the form of a graph structure to construct a basic knowledge graph.

[0071] Step S22 stores the knowledge nodes and knowledge relationship network structure in the multidimensional cross-knowledge framework in the form of a graph structure to a graph database, thus constructing a basic knowledge graph. This invention uses the Neo4j graph database as the storage engine, converting each knowledge node in the multidimensional cross-knowledge framework into a node object in Neo4j. A label is set for the node object according to the dimension to which the knowledge node belongs: "Business" for business-dimensional knowledge nodes, "Person" for personnel-dimensional knowledge nodes, "Scenario" for scenario-dimensional knowledge nodes, and "Resource" for resource-dimensional knowledge nodes. A second label is also set according to the resource type of the knowledge node. This invention stores the attribute information of knowledge nodes as attribute fields of node objects. These attribute fields include node identifier, node name, node description, creation time, modification time, creator, department, business relevance, security level, and version number. Semantic feature information of knowledge nodes is also stored as attribute fields of node objects. Node objects are created in batches using Neo4j's Cypher query language. Subsequently, this invention converts the relationships in the knowledge relationship network structure into relation objects in Neo4j. Type labels are set for relation objects based on their relationship types: "CONTAINS" for containment relationships, "DEPENDS_ON" for dependency relationships, "REFERENCES" for referencing relationships, and "SIMILAR_TO" for similarity relationships. Relationship weights, creation times, and other information are stored as attribute fields of relation objects. Relation objects are created in batches using Cypher query language. After importing, a unique index is created for the identifier field of the node objects. Finally, full-text indexes are created for the name and keyword fields of the node objects, and an index is created for the type field of the relation objects, thus constructing a basic knowledge graph.

[0072] The relationship types mentioned in step S22 include inclusion relationship, dependency relationship, reference relationship and similarity relationship.

[0073] S23: Establish a linking and jumping mechanism between knowledge nodes in the basic knowledge graph to obtain a dynamic knowledge graph.

[0074] In step S23, in conjunction with hardware, this invention embeds a related knowledge recommendation module into the details page of a knowledge node, comprising three sub-modules: related recommendations, upstream and downstream processes, and citation sources. Specifically, the related recommendations sub-module queries the basic knowledge graph for all outgoing and incoming edges of type SIMILAR_TO for the current node, obtains a list of similar nodes, sorts them in descending order of similarity weight, and displays the top 10 similar nodes. Each similar node displays its name, description, and similarity score; clicking the node name jumps to the node's details page. The upstream and downstream processes sub-module queries all outgoing edges of type DEPENDS_ON for the current node to obtain a list of upstream nodes and queries all incoming edges of type DEPENDS_ON for the current node to obtain a list of downstream nodes. Upstream and downstream nodes are displayed in the upstream and downstream processes area, respectively; clicking the node name jumps to the corresponding node's details page. The citation sources sub-module queries all incoming edges of type REFERENCES for the current node to obtain a list of citation sources and queries all outgoing edges of type REFERENCES for the current node to obtain a list of citation targets. These nodes and their citation context text are displayed in the citation sources area; clicking the node name jumps to the corresponding node's details page. Ultimately, by embedding these related knowledge recommendation modules, a linking and jumping mechanism between knowledge nodes is established to obtain a dynamic knowledge graph.

[0075] S3: Collect user feature data and generate a user knowledge need profile based on the user feature data.

[0076] In step S3 of this invention, user feature data is collected, and a user knowledge demand profile is generated based on the user feature data. The user feature data includes the user's identity information, behavioral data, and interaction content data. By collecting and analyzing this data, the user's knowledge demand characteristics are accurately depicted.

[0077] Step S3 further includes:

[0078] S31: Collect user identity information and behavioral data to obtain basic user characteristics.

[0079] Further, step S31 collects user identity information and behavioral data to obtain basic user characteristics. Identity information is extracted from user registration information. This invention reads user records from the user database and extracts the user's employee number, name, job title, department code, job level, and skill tag fields, assembling these field values ​​into an identity information data structure. The job title represents the user's job title, such as "Algorithm Engineer" or "Sales Manager"; the department code represents the user's department identifier, such as "D001" for R&D and "D002" for Sales; the job level represents the user's rank, such as "Junior," "Intermediate," "Senior," or "Expert"; the skill tags are a list of skill keywords filled in by the user, such as "Python," "Machine Learning," or "Data Analysis"; behavioral data is extracted from the user's operation logs in the knowledge base system. This invention queries the operation log table, filters out all operation records of the current user in the past 90 days, and counts the user's search keyword list, search count, document collection list, document collection count, document like list, and document like count; the search keyword list records all keywords entered by the user in the search box, the search count records the total number of times the user performed a search operation, the document collection list records the list of document identifiers collected by the user, the document collection count records the list length, and the document like list records the list of document identifiers liked by the user, the document like count records the list length. This invention merges identity information data structure and behavioral data statistical results to form a user basic feature data object. This object includes fields such as employee number, name, job title, department code, job level, skill tag list, search keyword list, search count, document collection list, document collection count, document likes list, and document likes count.

[0080] S32: Obtain user interaction content in the community, and use natural language processing technology to segment and extract the topic of the interaction content to obtain user knowledge demand characteristics.

[0081] In step S32, the present invention first queries the community database to extract the questions, comments, and discussion records posted by the current user in the community. Each interaction content record includes a content identifier, a posting time, and a text content field. The present invention concatenates the text content fields of all interaction contents into a long text string, and inputs the long text string into a Chinese word segmentation tool for word segmentation. The word segmentation tool uses a method combining a dictionary and statistics to split the continuous text into independent word units and outputs a word list. Subsequently, the present invention filters out stop words from the word list. Stop words refer to high-frequency words such as "de", "le", "zai", etc. that have no actual meaning. The present invention loads a predefined stop word list, traverses the word list, and deletes the words existing in the stop word list to obtain a filtered word list. After filtering, the present invention performs word frequency statistics on the filtered word list to create a word frequency dictionary. It traverses the word list. For each word, if the word does not exist in the word frequency dictionary, it is added to the dictionary and the count value is set to 1. If the word already exists in the word frequency dictionary, its count value is incremented by 1. After the statistics are completed, a word frequency dictionary is obtained. The keys of the word frequency dictionary are words, and the values are the number of times the word appears in the user interaction content. The present invention sorts the word frequency dictionary in descending order of word frequency values, selects the top 20 words with the highest word frequency values as the keywords that the user is interested in, and forms a list of keywords that the user is interested in with these 20 keywords. Then, the present invention takes all the interaction content texts of the user as a document and inputs it together with other documents in the knowledge base into the LDA topic model. The LDA model has been trained in step S12 and contains 50 topics. Each topic is represented by high-probability words. The present invention uses the trained LDA model to infer the probability distribution of the user interaction content document on each topic and outputs a probability vector with a length of 50. Each element in the vector represents the probability that the document belongs to the corresponding topic. The probability value ranges from 0 to 1, and the sum of all elements is 1. After the output, the present invention selects the top 5 topics with the highest probability values, records the topic numbers and probability values of these 5 topics, and forms a list of topics that the user is interested in. Finally, the present invention merges the list of keywords that the user is interested in and the list of topics that the user is interested in to form a user knowledge demand feature data object, which includes fields such as the list of keywords that the user is interested in and the list of topics that the user is interested in.

[0082] S33: Integrate the user basic features with the user knowledge demand features, and perform dynamic update in combination with the change of the user state to generate the user knowledge demand portrait.

[0083] Furthermore, in step S33, the present invention merges the user basic feature data object obtained in step S31 and the user knowledge demand feature data object obtained in step S32 to create a user knowledge demand profile data object. This object contains all fields of the user basic features and all fields of the user knowledge demand features. Next, the present invention adds a profile creation time field and a profile update time field to the user knowledge demand profile data object. The profile creation time records the timestamp of the first generation of the profile, and the profile update time records the timestamp of the last update of the profile. Then, the present invention serializes the user knowledge demand profile data object into a JSON format string and stores it in the user profile. In the database, the user's employee ID is used as the primary key. Subsequently, this invention establishes a user status monitoring mechanism, periodically querying the user database and organizational structure database to detect changes in user job positions and project participation status. Job position change detection compares the user's current job name, department code, and position level with the corresponding field values ​​recorded in the user's knowledge requirement profile. If a discrepancy is found, a job position change is determined. Project participation status change detection queries the project management system to obtain the user's current project list and compares it with the project participation list recorded in the user's knowledge requirement profile. If a new project is added or a project ends, a change in project participation status is determined. When a change in user job position or project participation status is detected, this invention triggers a profile update process, re-executing steps S31 and S32 to obtain the latest user basic characteristics and user knowledge requirement characteristics. The new feature data is updated in the user knowledge requirement profile data object, the profile update time field is updated to the current timestamp, and the updated user knowledge requirement profile data object is re-stored in the user profile database, overwriting the original record.

[0084] S4: Match the user's knowledge needs profile with the knowledge nodes in the dynamic knowledge graph, filter the matching results according to the user's current scenario context, and obtain a target knowledge push list.

[0085] Step S4 further includes:

[0086] S41: Classify and label the knowledge nodes in the dynamic knowledge graph, and extract the feature labels of the knowledge nodes through machine learning algorithms to obtain a labeled knowledge base.

[0087] In step S41, this invention first establishes a three-level classification system of industry, domain, and scenario. The industry level includes categories such as "manufacturing," "finance," and "internet"; the domain level includes categories such as "production management," "marketing," and "technology research and development"; and the scenario level includes categories such as "equipment maintenance," "customer complaint handling," and "algorithm optimization." Then, this invention trains three text classifiers for industry, domain, and scenario classification, respectively. The classifiers use the Support Vector Machine (SVM) algorithm, using labeled knowledge nodes as training samples. Each sample contains the text content of the knowledge node and its corresponding classification label. The text content is converted into a TF-IDF feature vector as input, and the classification label is used as output. The mapping relationship between feature vectors and classification labels is learned to obtain a classification model. For unlabeled knowledge nodes, the present invention extracts their text content and converts it into TF-IDF feature vectors, which are then input into three classifiers. The classifiers output the industry category, domain category, and scenario category of the knowledge node, respectively. Subsequently, the present invention selects the top 5 keywords with the highest TF-IDF weights from the keyword list in the semantic feature information of the knowledge node as keyword labels. Finally, the present invention merges the classification labels and keyword labels to create a label set for each knowledge node. The label set contains a list of industry labels, domain labels, scenario labels, and keyword labels. The label set is stored in the attribute fields of the knowledge node to obtain a labeled knowledge base.

[0088] S42: Calculate the similarity between the user's knowledge demand profile and the knowledge nodes in the tagged knowledge base, filter out candidate knowledge nodes whose similarity exceeds a preset threshold, and obtain preliminary matching results.

[0089] Furthermore, in step S42, the present invention converts the user's knowledge demand profile into a feature vector. The 20 keywords in the user's keyword list and their word frequency values ​​are represented as a sparse vector. The vector dimension is the total number of keywords in the knowledge base. The value of the dimension corresponding to the keyword that the user is interested in is the word frequency value of that keyword, and the values ​​of other dimensions are 0. Specifically, the 5 topics in the user's topic list and their probability values ​​are represented as a dense vector. The vector dimension is 50. The value of the dimension corresponding to the topic that the user is interested in is the probability value of that topic, and the values ​​of other dimensions are 0. This invention concatenates keyword vectors and topic vectors to form a user profile feature vector. Specifically, this invention converts each knowledge node in the tagged knowledge base into a feature vector. The keywords and their TF-IDF weights in the keyword tag list of the knowledge node are represented as sparse vectors, and the topics and their probability values ​​in the topic tag list of the knowledge node are represented as dense vectors. These two are concatenated to form the knowledge node feature vector. Subsequently, this invention calculates the cosine similarity between the user profile feature vector and the feature vector of each knowledge node. The cosine similarity is equal to the dot product of the two vectors divided by the product of the magnitudes of the two vectors. Finally, this invention sets a preset threshold of 0.3 to filter out knowledge nodes with a similarity greater than 0.3. The identifiers and similarity values ​​of these knowledge nodes are combined into key-value pairs and sorted in descending order of similarity value to obtain a preliminary matching result list.

[0090] S43: Obtain the user's current scene context information, score the relevance of the preliminary matching result based on the scene context information, and output the target knowledge push list based on the score result.

[0091] Furthermore, during the scoring process, this invention obtains the user's current scenario context information. This context information includes the user's current task identifier and the user's current operation interface identifier. The user's current task identifier is obtained from the project management system or task management system, indicating the type of task the user is currently performing. The user's current operation interface identifier is obtained from the user's browser session information, indicating the system page the user is currently accessing. This invention establishes a mapping rule base between tasks and scenarios. The mapping rule base is stored in key-value pairs, where the key is the task type and the value is a list of priority scenario tags corresponding to that task type. For example, the priority scenario tag list corresponding to the task type "handling customer complaints" is "customer complaint handling" and "customer service". "Product Consultation"; Subsequently, this invention queries the mapping rule base based on the user's current task identifier to obtain the corresponding priority scenario tag list. Specifically, this invention traverses each knowledge node in the preliminary matching result list, reads the scenario tag of the knowledge node, and determines whether the scenario tag exists in the priority scenario tag list. If it exists, the knowledge node is assigned a scenario weighting coefficient of 1.5; if it does not exist, it is assigned a scenario weighting coefficient of 1.0. Finally, this invention multiplies the similarity value of each knowledge node in the preliminary matching result list by its scenario weighting coefficient to obtain a weighted similarity value. The weighted similarity values ​​are sorted in descending order, and the top 20 knowledge nodes with the highest weighted similarity values ​​are selected to form the target knowledge push list.

[0092] S5: Determine the access permission level based on the user's role and permission attributes, perform permission verification on the knowledge nodes in the target knowledge push list based on the access permission level, and generate personalized knowledge content with access permission identifiers.

[0093] Step S5 further includes:

[0094] S51: Query the permission mapping rules based on the role permission attributes to obtain the user's access permission scope.

[0095] Further, step S51 queries the permission mapping rules based on the role permission attributes to obtain the user's access permission scope. This invention extracts the user's role permission attributes from the user's knowledge requirement profile. The role permission attributes include the user's job level, department code, and scope of business responsibilities fields. The job level field includes three levels: "Ordinary Employee," "Department Manager," and "Company Leader." The department code field represents the user's department identifier, and the scope of business responsibilities field represents the user's job responsibilities, such as "Sales," "Technology Research and Development," or "Financial Management."

[0096] This invention queries a permission mapping rule table, which is stored in a database table. The table structure includes a role type field, an access permission level field, and an accessible security level range field. The role type field includes values ​​such as "Ordinary Employee," "Department Manager," and "Company Leader." The access permission level field is an integer, ranging from 1 to 5, with higher values ​​indicating higher permission levels. The accessible security level range field is a string array representing the list of knowledge security levels that the role type can access. This invention uses the user's job level field as a query condition to find the corresponding record row in the permission mapping rule table and reads the access permission level field value and the accessible security level range field value of that record row.

[0097] For example, a user with the job level of "Ordinary Employee" will have an access level of 1 and can access the confidentiality levels of "Public" and "Internal Public". A user with the job level of "Department Manager" will have an access level of 3 and can access the confidentiality levels of "Public", "Internal Public", and "Department Confidential". A user with the job level of "Company Leader" will have an access level of 5 and can access the confidentiality levels of "Public", "Internal Public", "Department Confidential", "Company Confidential", and "Top Secret".

[0098] Ultimately, this invention combines the queried access permission levels and accessible security classification ranges into an access permission range data object, which contains two fields: permission level and security classification list.

[0099] S52: Extract the confidentiality attributes of multiple knowledge nodes in the target knowledge push list, match and verify the confidentiality attributes with the access permission scope, and obtain the permission verification result.

[0100] In step S52, this invention traverses each knowledge node in the target knowledge push list, queries the complete attribute information of the node from the graph database based on the knowledge node identifier, and reads the security level identifier field value in the node attributes. The security level identifier field value represents the confidentiality level of the knowledge node, and the values ​​include five levels: "Public", "Internal Public", "Department Confidential", "Company Confidential", and "Top Secret". This invention compares the security level identifier field value of the knowledge node with the security level list in the access permission scope data object to determine whether the security level identifier field value exists in the security level list. If the security level identifier field value exists in the security level list, it is determined that the user has the right to access the knowledge node, and the knowledge node identifier and the access permission determination result "Allowed to Access" are combined into a key-value pair and stored in the access permission verification result list. If the security level identifier field value does not exist in the security level list, it is determined that the user does not have the right to access the knowledge node, and the knowledge node identifier and the access permission determination result "Not Allowed to Access" are combined into a key-value pair and stored in the access permission verification result list. After this invention traverses all knowledge nodes in the target knowledge push list, a complete access permission verification result list is obtained, and each element in the list contains two fields: knowledge node identifier and access permission determination result.

[0101] S53: Based on the permission verification result, filter the knowledge nodes in the target knowledge push list, retain the knowledge nodes that the user has the right to access, and generate the personalized knowledge content with the access permission identifier.

[0102] Furthermore, in step S53, the present invention iterates through each element in the permission verification result list, reads the access permission judgment result field value, and if the field value is "authorized to access", the knowledge node identifier is retained; if the field value is "unauthorized to access", the knowledge node is deleted from the target knowledge push list. Then, a personalized knowledge content list is created based on the retained content. The retained knowledge node identifiers are iterated through, and detailed information of the knowledge nodes is queried from the graph database based on the node identifiers. The detailed information includes fields such as node name, node description, creation time, creator, keyword tags, and scene tags. Subsequently, an access permission identifier field is added to each knowledge node, and the value of the access permission identifier field is set to "accessible", indicating that the user has permission to view the knowledge content. Finally, the present invention returns the personalized knowledge content list to the front-end display module. The front-end display module renders the knowledge content list on the user interface, and each knowledge content displays the node name, node description, creation time, creator, and access permission identifier. Users can click on the knowledge content to jump to the knowledge details page to view the complete content.

[0103] This also includes:

[0104] S6: Collect user actions on accessing the personalized knowledge content across the entire process and write the action event information into the log database to obtain knowledge flow records.

[0105] In step S6, this invention embeds operation behavior listeners in the front-end display module and the back-end service module. The listeners capture all user operation events related to personalized knowledge content. Operation event types include viewing events, download events, favorite events, like events, comment events, and share events. When a user clicks on knowledge content to jump to the knowledge details page, the listener captures a viewing event and records the event type as "VIEW". When a user clicks the download button to download knowledge attachments, the listener captures a download event and records the event type as "DOWNLOAD". When a user clicks the favorite button to add knowledge content to their favorites, the listener captures a favorite event and records the event type as "FAVORITE". When a user clicks the like button to like knowledge content, the listener captures a like event and records the event type as "LIKE". When a user submits a comment on the knowledge details page, the listener captures a comment event and records the event type as "COMMENT". When a user shares knowledge content with other users through the sharing function, the listener captures a share event and records the event type as "SHARE". Then, this invention extracts operation event information for each captured operation event. The information includes fields such as event type, event timestamp, user identifier, knowledge node identifier, and operation details. The event type field contains the event type code, the event timestamp field contains the precise time of the event (accurate to milliseconds), the user identifier field contains the user's employee ID, the knowledge node identifier field contains the unique identifier of the knowledge node being operated on, and the operation details field records additional information based on the event type, such as the filename and size of the downloaded file in a download event, the comment text content in a comment event, and a list of employee IDs of the target users in a share event. Subsequently, this invention encapsulates the operation event information into JSON format log records. The JSON structure of the log records includes key-value pairs such as event_type, timestamp, user_id, node_id, and details. Then, this invention writes the log records to a log database, which uses a time-series database or a NoSQL database, supporting high-concurrency writes and queries by time range. Finally, this invention creates a composite index for the log records, with index fields including the event timestamp and user identifier, accelerating query operations filtered by time range and user. Ultimately, this invention combines all log records written to the log database into a knowledge flow record. This knowledge flow record fully records the user's operation trajectory when accessing knowledge content, supporting administrators to perform audit analysis and trace the access history and propagation path of knowledge.

[0106] like Figure 2 As shown, the present invention also provides a three-dimensional enterprise knowledge base construction system, including:

[0107] Extraction Module 100: Used to classify and label the knowledge assets to be constructed according to a preset dimension system, and extract semantic feature information corresponding to knowledge nodes of multiple dimensions through natural language processing technology to obtain a multi-dimensional cross-knowledge framework;

[0108] Construction module 200: used to store the knowledge nodes in the multidimensional cross-knowledge framework into a graph database, establish a knowledge association network according to the type of association between knowledge, and obtain a dynamic knowledge graph;

[0109] Description module 300: Used to collect user feature data and generate a user knowledge need profile based on the user feature data;

[0110] Matching module 400: used to perform matching calculations between the user knowledge demand profile and the knowledge nodes in the dynamic knowledge graph, filter the matching results according to the user's current scenario context, and obtain a target knowledge push list;

[0111] Hierarchical module 500: Used to determine the access permission level based on the user's role permission attributes, perform permission verification on the knowledge nodes in the target knowledge push list based on the access permission level, and generate personalized knowledge content with access permission identifiers.

[0112] This invention provides a method and system for constructing a three-dimensional enterprise knowledge base. Through structured classification and summarization, the knowledge base centrally manages scattered quality control points, avoiding operational oversights caused by fragmented information and significantly improving the overall system's processing quality. Secondly, it pushes customized knowledge based on user roles and scenarios, using keywords, semantic search, and association analysis to help users accurately locate the information they need. This increases the proportion of data-driven decision-making and reduces reliance on human experience, thereby reducing the risk of misjudgment and improving the accuracy of output results. Furthermore, this invention enables cross-departmental and cross-regional teams to access shared knowledge (such as project documents) in real time through knowledge sharing. This invention utilizes case studies and experience to avoid repetitive work, improve efficiency, and record knowledge update history through traceability. This allows for rapid identification of defects and modification records during system failures, reducing debugging time and improving maintenance efficiency. Furthermore, through multi-dimensional aggregation and knowledge sharing, it preserves historical corporate experience, avoids redundant investment, achieves resource reuse and efficiency enhancement, and makes hidden costs explicit, thereby significantly reducing labor, training, and time costs. Additionally, by storing knowledge electronically to replace paper documents, this invention reduces paper consumption and waste generation, enhancing the company's green image. It also extends the lifespan of knowledge assets by recording knowledge maintenance history and knowledge iteration logs. By training new employees through hierarchical access control, it avoids knowledge gaps caused by the departure of key personnel, thus indirectly extending the system's lifespan.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a three-dimensional enterprise knowledge base, characterized in that, include: S1: Classify and label the knowledge assets to be constructed according to a preset dimension system, and extract semantic feature information corresponding to knowledge nodes of multiple dimensions through natural language processing technology to obtain a multi-dimensional cross-knowledge framework; S2: Store the knowledge nodes in the multidimensional cross-knowledge framework into a graph database, establish a knowledge association network according to the type of association between knowledge, and obtain a dynamic knowledge graph; S3: Collect user feature data and generate a user knowledge demand profile based on the user feature data; S4: Match the user knowledge demand profile with the knowledge nodes in the dynamic knowledge graph, filter the matching results according to the user's current scenario context, and obtain a target knowledge push list; S5: Determine the access permission level based on the user's role and permission attributes, perform permission verification on the knowledge nodes in the target knowledge push list based on the access permission level, and generate personalized knowledge content with access permission identifiers.

2. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, The preset dimension system in step S1 includes business dimension, personnel dimension, scenario dimension and resource dimension.

3. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, The process of constructing the multidimensional knowledge framework in step S1 includes: S11: Classify the enterprise's knowledge assets according to multiple dimensions in the preset dimension system, and label the knowledge nodes under multiple dimensions with attribute information to obtain a knowledge set of multiple dimensions; S12: Perform natural language processing on the knowledge nodes in the knowledge set to extract keywords, topic tags and entity information of the knowledge nodes, and obtain semantic feature information of the knowledge nodes; S13: Based on the semantic feature information, cross-integrate knowledge sets of multiple dimensions, establish inter-dimensional correlation indexes, and form the multi-dimensional cross-knowledge framework.

4. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, Step S2 further includes: S21: Based on the semantic feature information of knowledge nodes in the multidimensional cross-knowledge framework, define the types of association relationships between knowledge nodes to obtain the knowledge relationship network structure; S22: Store the knowledge nodes and the knowledge relationship network structure in the multidimensional cross-knowledge framework in a graph database in the form of a graph structure to construct and obtain a basic knowledge graph; S23: Establish a linking and jumping mechanism between knowledge nodes in the basic knowledge graph to obtain a dynamic knowledge graph.

5. The method for constructing a three-dimensional enterprise knowledge base according to claim 4, characterized in that, The types of relationships mentioned in step S22 include containment relationships, dependency relationships, reference relationships, and similarity relationships.

6. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, Step S3 further includes: S31: Collect user identity information and behavioral data to obtain basic user characteristics; S32: Obtain user interaction content in the community, and use natural language processing technology to segment and extract the topic of the interaction content to obtain user knowledge demand characteristics; S33: The user's basic characteristics and the user's knowledge needs characteristics are fused together and dynamically updated in combination with changes in the user's state to generate the user's knowledge needs profile.

7. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, Step S4 further includes: S41: Classify and label the knowledge nodes in the dynamic knowledge graph, and extract the feature labels of the knowledge nodes through machine learning algorithms to obtain a labeled knowledge base; S42: Calculate the similarity between the user knowledge demand profile and the knowledge nodes in the tagged knowledge base, filter out candidate knowledge nodes whose similarity exceeds a preset threshold, and obtain preliminary matching results; S43: Obtain the user's current scene context information, score the relevance of the preliminary matching result based on the scene context information, and output the target knowledge push list based on the score result.

8. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, Step S5 further includes: S51: Query the permission mapping rules based on the role permission attributes to obtain the user's access permission scope; S52: Extract the confidentiality attributes of multiple knowledge nodes in the target knowledge push list, match and verify the confidentiality attributes with the access permission range, and obtain the permission verification result; S53: Based on the permission verification result, filter the knowledge nodes in the target knowledge push list, retain the knowledge nodes that the user has the right to access, and generate the personalized knowledge content with the access permission identifier.

9. The method for constructing a three-dimensional enterprise knowledge base according to claim 1, characterized in that, Also includes: S6: Collect user actions on accessing the personalized knowledge content across the entire process and write the action event information into the log database to obtain knowledge flow records.

10. A three-dimensional enterprise knowledge base construction system, characterized in that, include: Extraction module: Used to classify and label the knowledge assets to be constructed according to a preset dimension system, and extract semantic feature information corresponding to knowledge nodes of multiple dimensions through natural language processing technology to obtain a multi-dimensional cross-knowledge framework; Construction module: Used to store knowledge nodes in the multidimensional cross-knowledge framework into a graph database, establish a knowledge association network according to the types of relationships between knowledge, and obtain a dynamic knowledge graph; Description module: Used to collect user feature data and generate a user knowledge need profile based on the user feature data; Matching module: used to perform matching calculations between the user's knowledge demand profile and the knowledge nodes in the dynamic knowledge graph, filter the matching results according to the user's current scenario context, and obtain a target knowledge push list; Hierarchical module: used to determine the access permission level based on the user's role permission attributes, perform permission verification on the knowledge nodes in the target knowledge push list based on the access permission level, and generate personalized knowledge content with access permission identifiers.

Images