Knowledge classification method and device

Abstract

The invention discloses a knowledge classification method and device. The method comprises the following steps: acquiring a geographic information ontology conceptual model; processing the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge comprises N pieces of ontology knowledge information, and each piece of ontology knowledge information comprises an ontology knowledge title and ontology knowledge content; and processing the ontology knowledge set to obtain a knowledge classification result. According to the method, the ontology knowledge set is extracted and constructed from the geographic information ontology concept model, and the obtained ontology knowledge title and content naturally bear the core concept, attribute association and hierarchical logic of the geographic information field, so that the subsequent classification result is ensured to completely accord with the internal system of the geographic information knowledge; the problems of disordered classification logic and poor field adaptability caused by knowledge source mixing of a traditional classification method are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of knowledge management and classification, and particularly relates to a knowledge classification method and device. BACKGROUND

[0002] In the field of knowledge management, a reasonable knowledge classification method is the basis for realizing efficient retrieval, association and utilization of knowledge. At present, the existing knowledge classification methods mostly adopt single face classification or line classification, and have problems such as incomplete classification logic, hierarchical redundancy, poor adaptability to actual application scenarios, etc. Especially in the field related to knowledge management, knowledge covers multiple dimensions such as general space-time, regional elements and events, involves multiple different classification coding specifications, and the existing classification methods are difficult to balance the systematization and convenience of classification: if only face classification is adopted, it is easy to cause category overlap; if only line classification is adopted, it will cause too many knowledge levels, affecting the retrieval efficiency. At the same time, the existing classification system does not fully combine the latest research results of disciplines, cannot completely express the entity elements of the real world, and is inconvenient for knowledge source association and tracing, and is difficult to meet the demand for fine management of knowledge. Therefore, there is an urgent need for a knowledge classification method that can integrate multiple specifications, optimize the hierarchical structure, and balance the systematization and convenience. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a knowledge classification method and device, which solves the problems of hierarchical redundancy, incomplete classification logic, poor adaptability to multiple specifications and inconvenient knowledge tracing of the existing knowledge classification methods, and realizes the unity of systematization, logic and convenience of knowledge classification.

[0004] To solve the above technical problems, the first aspect of the embodiment of the present application discloses a knowledge classification method, which comprises: S1, acquiring a geographic information ontology concept model; S2, processing the geographic information ontology concept model to obtain an ontology knowledge set; the ontology knowledge comprises N pieces of ontology knowledge information, and each piece of ontology knowledge information comprises an ontology knowledge title and ontology knowledge content; S3, processing the ontology knowledge set to obtain a knowledge classification result.

[0005] As an optional implementation manner, in the first aspect of the embodiment of the present application, the processing of the ontology knowledge set to obtain the knowledge classification result comprises: S31, processing the ontology knowledge set to obtain a first-level classification result; the first-level classification result comprises factual knowledge, rule knowledge and decision knowledge; S32, performing second-level classification on the factual knowledge to obtain a factual knowledge classification result; the factual knowledge classification result comprises general space-time knowledge, regional element knowledge and event knowledge; S33, performing secondary classification on the rule type knowledge to obtain a rule type knowledge classification result; the rule type knowledge classification result comprises natural law and change rule knowledge and human society and man-made activity rule knowledge; S34, performing secondary classification on the decision type knowledge to obtain a decision type knowledge classification result; the decision type knowledge classification result comprises acquisition decision knowledge, aggregation decision knowledge, protection decision knowledge and guarantee decision knowledge; S35, integrating the fact type knowledge classification result, the rule type knowledge classification result and the decision type knowledge classification result to obtain a knowledge classification result.

[0006] As an optional implementation, in the first aspect of the embodiment of the present application, the processing of the ontology knowledge set to obtain a primary classification result comprises: S311, performing feature extraction on an ontology knowledge title in the ontology knowledge set to obtain title feature information; the title feature information comprises title theme feature information and title semantic feature information; S312, performing feature extraction on ontology knowledge content in the ontology knowledge set to obtain deep-level semantic feature information; S313, processing the title feature information by using a preset first classification model to obtain a first classification result; S314, processing the deep-level semantic feature information by using a preset second classification model to obtain a second classification result; S315, performing probability fusion on the first classification result and the second classification result to obtain a primary classification result.

[0007] As an optional implementation, in the first aspect of the embodiment of the present application, the feature extraction on the ontology knowledge title in the ontology knowledge set to obtain title feature information comprises: S3111, performing standardization processing on the ontology knowledge title to obtain a standardized ontology knowledge title; S3112, performing word segmentation processing on the standardized ontology knowledge title to obtain a pretreated ontology knowledge title; S3113, performing feature extraction on the pretreated ontology knowledge title to obtain title theme feature information; S3114, performing semantic feature extraction on the pretreated ontology knowledge title to obtain title semantic feature information.

[0008] As an optional implementation, in the first aspect of the embodiment of the present application, the feature extraction on the pretreated ontology knowledge title to obtain title theme feature information comprises: The title theme feature information extraction model is used for feature extraction on the preprocessed ontology knowledge title to obtain title theme feature information. The title theme feature information extraction model expression is: wherein, is the title theme feature information, is attribute information, is an attribute value, is is the mth element in the set, is the mth element in the set.

[0009] As an optional implementation, in the first aspect of the embodiment of the present application, the feature extraction on the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information comprises: S3121, document embedding is performed on the ontology knowledge content to obtain an embedding vector; S3122, dimension reduction processing is performed on the embedding vector to obtain a dimension-reduced embedding vector; S3123, clustering processing is performed on the dimension-reduced embedding vector to obtain a class vector of K topics, K being a positive integer; S3124, topic representation is performed on each topic class vector to obtain a keyword list of each topic cluster; S3125, information extraction is performed on the keyword list of each topic cluster to obtain deep semantic feature information.

[0010] As an optional implementation, in the first aspect of the embodiment of the present application, the topic representation on each topic class vector to obtain a keyword list of each topic cluster comprises: S31241, importance degree calculation is performed on each topic class vector to obtain an importance degree score of each topic; The importance degree score expression is: wherein, is the importance degree score of the topic cluster, denotes a candidate word, denotes a set of all topic clusters, denotes the number of other topic clusters containing the word denotes the number of topic clusters, denotes the size of the topic cluster denotes a word in the topic cluster . ​​​ Frequency of occurrence in; S31242, Based on the importance score of each topic, select the top N words as the topic keywords for each topic cluster, and obtain the keyword list for each topic cluster.

[0011] A second aspect of this invention discloses a knowledge classification device, the device comprising: The data acquisition module is used to acquire the geographic information ontology conceptual model; The ontology knowledge set construction module is used to process the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge includes N ontology knowledge information, and each ontology knowledge information includes an ontology knowledge title and ontology knowledge content; The knowledge classification module is used to process the ontology knowledge set to obtain knowledge classification results.

[0012] As an optional implementation, in the second aspect of the present invention, the processing of the ontology knowledge set to obtain knowledge classification results includes: S31, The ontology knowledge set is processed to obtain a first-level classification result; the first-level classification result includes factual knowledge, rule knowledge, and decision-making knowledge; S32, perform secondary classification on the factual knowledge to obtain factual knowledge classification results; the factual knowledge classification results include general spatiotemporal knowledge, regional element knowledge, and event knowledge; S33, perform secondary classification on the rule-based knowledge to obtain the rule-based knowledge classification result; the rule-based knowledge classification result includes knowledge of natural laws and rules of change and knowledge of human society and human activities rules; S34, perform secondary classification on the decision-making knowledge to obtain the decision-making knowledge classification results; the decision-making knowledge classification results include acquisition decision-making knowledge, aggregation decision-making knowledge, protection decision-making knowledge, and safeguard decision-making knowledge; S35, integrate the factual knowledge classification results, the rule-based knowledge classification results, and the decision-based knowledge classification results to obtain the knowledge classification results.

[0013] As an optional implementation, in the second aspect of the present invention, the processing of the ontology knowledge set to obtain a first-level classification result includes: S311, extract features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information; the title feature information includes title topic feature information and title semantic feature information. S312, extract features from the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information; S313, using a preset first classification model, the title feature information is processed to obtain a first classification result; S314, using a preset second classification model, the deep semantic feature information is processed to obtain the second classification result; S315, perform probability fusion on the first classification result and the second classification result to obtain the first-level classification result.

[0014] As an optional implementation, in the second aspect of the present invention, the step of extracting features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information includes: S3111, Standardize the ontology knowledge title to obtain a standardized ontology knowledge title; S3112, Perform word segmentation on the standardized ontology knowledge title to obtain a preprocessed ontology knowledge title; S3113, Perform feature extraction on the preprocessed ontology knowledge title to obtain title topic feature information; S3114, Semantic features are extracted from the preprocessed ontology knowledge title to obtain title semantic feature information.

[0015] As an optional implementation, in the second aspect of the present invention, the step of extracting features from the preprocessed ontology knowledge title to obtain title topic feature information includes: Using a title-topic feature information extraction model, feature extraction is performed on the preprocessed ontology knowledge title to obtain title-topic feature information; The expression for the title / topic feature information extraction model is: in, For title and theme feature information, For attribute information, For attribute values, for The m-th element in for The m-th element in.

[0016] As an optional implementation, in the second aspect of the present invention, the step of extracting features from the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information includes: S3121, Document embedding is performed on the ontology knowledge content to obtain an embedding vector; S3122, Perform dimensionality reduction processing on the embedding vector to obtain a dimensionality-reduced embedding vector; S3123, perform clustering processing on the reduced-dimensional embedding vector to obtain K class vectors of topics, where K is a positive integer; S3124, Represent the class vector of each topic as a topic to obtain a list of keywords for each topic cluster; S3125, extract information from the keyword list of each topic cluster to obtain deep semantic feature information.

[0017] As an optional implementation, in the second aspect of the present invention, the step of representing the class vector of each topic to obtain a keyword list for each topic cluster includes: S31241, calculate the importance of each topic's class vector to obtain an importance score for each topic; The expression for the importance score is as follows: in, For thematic clusters Importance score Indicates a candidate word. Represents the set of all topic clusters. Indicates the inclusion word The number of other topic clusters, Indicates the number of topic clusters, Represents a topic cluster Size, word In thematic clusters Frequency of occurrence in; S31242, Based on the importance score of each topic, select the top N words as the topic keywords for each topic cluster, and obtain the keyword list for each topic cluster.

[0018] A third aspect of the present invention discloses another knowledge classification device, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the knowledge classification method disclosed in the first aspect of the present invention.

[0019] The fourth aspect of the present invention discloses a computer-storable medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the knowledge classification method disclosed in the first aspect of the present invention.

[0020] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: This invention discloses a knowledge classification method that extracts and constructs an ontology knowledge set from a geographic information ontology conceptual model. The acquired ontology knowledge titles and contents naturally carry the core concepts, attribute associations, and hierarchical logic of the geographic information field, ensuring that the subsequent classification results fully conform to the inherent system of geographic information knowledge. This solves the problems of chaotic classification logic and poor domain adaptability caused by the mixed sources of knowledge in traditional classification methods. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a knowledge classification method disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of the geographic information ontology conceptual model disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a knowledge classification device disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of another knowledge classification device disclosed in an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] In all embodiments of the present invention, the variables involved in all computational expressions or mathematical functions are dimensionless before calculation. The values ​​of the independent variables in all computational expressions or mathematical functions in these embodiments conform to the reasonable requirements of the input range of the computational expression or mathematical function, ensuring that the computational expression or mathematical function can be calculated smoothly without violating physical laws or mathematical rules.

[0027] This invention discloses a knowledge classification method and apparatus. The method includes acquiring a geographic information ontology conceptual model; processing the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge includes N ontology knowledge information entries, each entry including an ontology knowledge title and content; and processing the ontology knowledge set to obtain a knowledge classification result. This invention extracts and constructs an ontology knowledge set from a geographic information ontology conceptual model. The acquired ontology knowledge titles and content naturally carry the core concepts, attribute relationships, and hierarchical logic of the geographic information domain, ensuring that the subsequent classification results fully conform to the inherent system of geographic information knowledge. This solves the problems of chaotic classification logic and poor domain adaptability caused by the mixed sources of knowledge in traditional classification methods. Detailed explanations follow.

[0028] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a knowledge classification method disclosed in an embodiment of the present invention. Figure 1 The described knowledge classification method is applied in the field of knowledge management and classification technology, and the embodiments of this invention are not limited thereto. Figure 1 As shown, this knowledge classification method may include the following operations: S1, Obtain the geographic information ontology conceptual model; Geographic information decision support knowledge is characterized by its multidisciplinary, multimodal, and multi-type nature, as well as complex spatiotemporal features and relationships. Based on the analysis of domain data, a conceptual model of the geographic information ontology should be further determined, clarifying the top-level concepts of the geographic information ontology and their interrelationships, thereby guiding the next step in determining the geographic information knowledge system. At its top level, geographic information decision support knowledge is divided into factual knowledge, rule-based knowledge, and decision-making knowledge.

[0029] Factual knowledge addresses the "what" question, including general spatiotemporal knowledge representing domain knowledge, knowledge of objectively existing entity elements, and knowledge of real-world events. Rule-based knowledge addresses the "why" question, referring to axioms, theorems, laws, and computational rules governing the objective existence, evolution, interaction, and influence of various factual knowledge within the domain. Decision-based knowledge addresses the "how" question, referring to knowledge that provides decision-making support for solving various application tasks within the domain.

[0030] Entity element knowledge within factual knowledge forms the core and foundation of geographic information decision support knowledge. It supports the definition of event knowledge, rule knowledge, and decision knowledge; that is, the definition of event knowledge, rule knowledge, and decision knowledge requires the reference of entity elements as the subject or object. From the perspective of the three-dimensional world, entity element knowledge is divided into natural environment, socio-cultural environment, and information environment knowledge. Decision knowledge can further reference rule knowledge to support the generation of decision-making schemes.

[0031] Figure 2 This is a schematic diagram of the geographic information ontology conceptual model disclosed in an embodiment of the present invention; S2, process the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge includes N ontology knowledge information, each ontology knowledge information including ontology knowledge title and ontology knowledge content; S3, process the ontology knowledge set to obtain knowledge classification results.

[0032] Optionally, the processing of the ontology knowledge set to obtain knowledge classification results includes: S31, The ontology knowledge set is processed to obtain a first-level classification result; the first-level classification result includes factual knowledge, rule knowledge, and decision-making knowledge; S32, perform secondary classification on the factual knowledge to obtain factual knowledge classification results; the factual knowledge classification results include general spatiotemporal knowledge, regional element knowledge, and event knowledge; The specific classification steps are as follows: Feature extraction is performed on the factual knowledge to obtain factual knowledge feature information; The factual knowledge is subjected to empirical mode decomposition to obtain n IMF component signals and one margin signal; The n IMF component signals are processed to obtain k high-frequency components and m low-frequency components; Wavelet threshold denoising is performed on the k high-frequency components to obtain the denoised k high-frequency components; The denoised k high-frequency components, the residual signal, and the m low-frequency components are reconstructed to obtain factual knowledge feature information.

[0033] Using the factual knowledge features, a pre-defined classification model is trained to obtain an optimized classification model; The default classification model is the CNN-Transformer model; Using the optimized classification model, the factual knowledge feature information to be processed is processed to obtain the factual knowledge classification result; S33, perform secondary classification on the rule-based knowledge to obtain the rule-based knowledge classification result; the rule-based knowledge classification result includes knowledge of natural laws and rules of change and knowledge of human society and human activities rules; The specific steps are the same as those for calculating the results of factual knowledge classification; S34, perform secondary classification on the decision-making knowledge to obtain the decision-making knowledge classification results; the decision-making knowledge classification results include acquisition decision-making knowledge, aggregation decision-making knowledge, protection decision-making knowledge, and safeguard decision-making knowledge; The specific steps are the same as those for calculating the results of factual knowledge classification; S35, integrate the factual knowledge classification results, the rule-based knowledge classification results, and the decision-based knowledge classification results to obtain the knowledge classification results.

[0034] Optionally, the processing of the ontology knowledge set to obtain the first-level classification result includes: S311, extract features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information; the title feature information includes title topic feature information and title semantic feature information. S312, extract features from the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information; S313, using a preset first classification model, the title feature information is processed to obtain a first classification result; The first classification model is the OC-CNN model; S314, using a preset second classification model, the deep semantic feature information is processed to obtain the second classification result; The second classification model is the OC-SVM model; S315, perform probability fusion on the first classification result and the second classification result to obtain the first-level classification result.

[0035] The first and second classification results are unified into target class probabilities of [0,1]. The weight coefficients of the two classification results are determined and weighted fusion is performed to obtain the first-level classification result.

[0036] Optionally, the step of extracting features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information includes: S3111, Standardize the ontology knowledge title to obtain a standardized ontology knowledge title; S3112, Perform word segmentation on the standardized ontology knowledge title to obtain a preprocessed ontology knowledge title; S3113, Perform feature extraction on the preprocessed ontology knowledge title to obtain title topic feature information; S3114, Semantic features are extracted from the preprocessed ontology knowledge title to obtain title semantic feature information.

[0037] Optionally, the step of extracting features from the preprocessed ontology knowledge title to obtain title topic feature information includes: Using a title-topic feature information extraction model, feature extraction is performed on the preprocessed ontology knowledge title to obtain title-topic feature information; The expression for the title / topic feature information extraction model is: in, For title and theme feature information, For attribute information, For attribute values, for The m-th element in for The m-th element in the dataset. In practical applications, it is necessary to... The elements in the table are concatenated line by line to obtain the title feature information, which is used for subsequent classification tasks.

[0038] Alternatively, another method for extracting features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information is as follows: The ontology knowledge title is segmented into words to obtain a word sequence; The vocabulary sequence is cleaned to remove stop words and special symbols, while retaining core vocabulary. The core vocabulary is standardized to obtain a standardized vocabulary sequence; 1) Extract title and theme feature information (1) Set the basic configuration of the Sparrow Search algorithm and the Empirical Mode Decomposition of Adaptive Noise Complete Set (CEEMDAN) parameters. Basic parameters of the Sparrow Search Algorithm (SSA): population size N, maximum number of iterations T, proportion of discoverers (0.2~0.3), proportion of vigilants (0.1~0.2), and safety threshold (0.6~0.8). CEEMDAN core optimization parameters and search range: Additional white noise amplitude coefficient (Usually taken as 0.01~0.2), set average frequency (Integers, usually between 10 and 100); Constraints: M is a positive floating-point number, and M is a positive integer. After the parameters are updated, it is necessary to check whether they are within the search range.

[0039] (2) Initialize the sparrow population and determine the initial optimal individual. Based on population size N, in An initial population is randomly generated within the search range of M, with each sparrow individual corresponding to a set of CEEMDAN parameters. ,in Rounding Preserve floating-point numbers; Parameters for each individual sparrow The CEEMDAN decomposition is performed on the standardized vocabulary sequence to obtain several intrinsic mode functions (IMFs). Calculate the envelope entropy of all IMF components, select the minimum envelope entropy as the objective function value for that individual, and the individual with the minimum objective function value is the initial optimal sparrow. Record its parameters. And the optimal objective function value.

[0040] (3) Sparrow search algorithm foraging behavior, updating population parameters A global search for parameters is achieved through a discoverer-follower position update mechanism, and the update rules for CEEMDAN parameters are adapted as follows: Discoverer position update: The discoverer moves to the globally optimal region, expanding the search range, and the update formula is adapted as follows: Where t is the current iteration number, round() is the rounding operation, and M is guaranteed to be an integer; rand(0,1) is a random number in the range (0,1], and rand(0,1) is a random number between 0 and 1. Let be the i-th M parameter in the t-th iteration. For the i-th iteration in t iterations The parameter, T, represents the total number of iterations. The optimal M parameters for iteration t. The best for t iterations parameter; Follower position update: Some followers quickly move towards the discoverer's position, while others randomly search in local areas to avoid getting trapped in local optima; Parameter validity verification: for the updated and Perform range validation, and truncate parameters that exceed the boundary to the corresponding boundary value.

[0041] (4) Sparrow search algorithm anti-predation behavior, secondary optimization of population parameters By leveraging the danger perception mechanism of the vigilant, a localized, fine-grained search for parameters is achieved. The core steps are as follows: Randomly select 10% to 20% of individuals from the population as vigilants, and calculate the parameter distance and objective function value difference between the vigilants and the current best sparrow; If the watcher senses danger (parameters are too close or the objective function value deteriorates), it randomly updates its position in a direction away from the local optimum; if it does not sense danger, it moves closer to the current best sparrow and performs a fine-grained local search. Perform rounding and range checks on the updated parameters of all vigilants to ensure... and The constraints are met; All sparrows are processed according to the updated parameters. The preprocessed speech signal is subjected to CEEMDAN decomposition to obtain a new set of IMF components.

[0042] (5) Iterate to find the optimal CEEMDAN parameters after the termination condition is met.

[0043] For all sparrow individuals after iterative update, recalculate their objective function value (minimum value of IMF component envelope entropy). Iterate through all individuals, update the parameters and objective function value of the globally optimal sparrow, and retain the parameter combination with the minimum envelope entropy; Repeat steps (3) to (4) until the maximum number of iterations T is reached or the objective function value converges (the change in envelope entropy over multiple consecutive generations is less than the set threshold). After the iteration terminates, output the parameters corresponding to the globally optimal sparrow—the optimal additional white noise amplitude coefficient. and the average number of times the optimal set .

[0044] (6) CEEMDAN decomposition under optimal parameters to extract title theme feature information.

[0045] 2) Extraction of semantic features from titles Based on the pre-trained language model (BERT), the entire standardized vocabulary sequence is directly encoded to obtain sentence embedding vectors that include contextual semantics; (1) Perform EEMD analysis on the sentence embedding vector to obtain the IMF components. Filter the high-frequency IMF components and other low-frequency components, and select M IMF components containing the main information. ,; (2) Calculate the total energy of each IMF component. : ( j =1,2,...,M) (3) Construct a new feature vector using energy as an element. : , in, Represents the preset scale factor ; (4) Normalize the eigenvectors to obtain normalized eigenvectors. : , The title feature information is obtained by fusing the title theme feature information and the title semantic feature information; The fusion method is as follows: Let X represent the title's key features. This refers to the semantic features of the title. in, is the hyperbolic tangent function, Norm is the normalization process, and W and b are learnable model parameters obtained by training with preset data; Optionally, the step of extracting features from the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information includes: S3121, Document embedding is performed on the ontology knowledge content to obtain an embedding vector; S3122, Perform dimensionality reduction processing on the embedding vector to obtain a dimensionality-reduced embedding vector; S3123, perform clustering processing on the reduced-dimensional embedding vector to obtain K class vectors of topics, where K is a positive integer; The clustering method is: 1) Initialize the number of clusters and hyperparameters, initialize the cluster centers, membership matrices, transition confidence matrices, and feature weight matrices of the target and source domains, and initialize the dynamic adjustment coefficients. and .

[0046] 2) Iteratively optimize and alternately update the following variables until the objective function converges: Update the membership matrix U: Calculate the membership degree of the sample to the target domain cluster based on the fitting error weighted by adaptive features.

[0047] Update the feature weight matrix W: combine the fitting error, normalization regularization and cross-domain alignment term to optimize the weights of each feature.

[0048] Update the migration confidence matrix R: Adjust the confidence between the target domain and source domain clusters based on the dynamic migration matching error and the adaptive penalty term.

[0049] Update dynamic adjustment coefficient and Based on the dispersion and distribution similarity of the current cluster centers, the importance of features and the migration intensity are dynamically adjusted.

[0050] Update cluster center and Using the updated membership and migration confidence, the cluster centers of the target and source domains are recalculated.

[0051] 3) Convergence judgment: When the change in the objective function is less than the set threshold, or when the maximum number of iterations is reached, the iteration stops and the final clustering result is output.

[0052] Objective function: In the formula, , For matrix The elements in the expression represent the membership degree of sample j to cluster center i in the target domain. For matrix The elements in the table represent the values ​​of the target domain cluster center i in the feature dimension k. This represents the value of the cluster center l in the target domain along the feature dimension k, where K is the number of clusters and N is the number of samples in the dimensionality-reduced embedding vector. This represents the dimension-reduced embedding vector in the i-th row and k-th column. It is a weighted vector matrix of the source domain. It is the cluster center matrix of the source domain, where d is the feature dimension. Let be the element in the i-th row and l-th column of matrix R, and let be the target domain clustering element. i Clustering with source domain l The transfer confidence (similarity) between them, where M is the number of source domain clusters. It is the weighted vector matrix of the target domain. It is the cluster center matrix of the target domain. These are adaptive feature weighting coefficients, automatically calculated from the dispersion of the current cluster centers. This allows the algorithm to dynamically identify and amplify highly discriminative features during iterations, while suppressing interference from noisy features. This represents the Hadamard product operation. The dynamic transfer confidence score is determined by the similarity of the distributions of the target domain and the source domain clusters. When the distributions of the two clusters are similar, The magnitude of the change will increase, thereby strengthening the migration intensity of the path, and vice versa; The weighting coefficients are set experimentally. As a penalty factor, it assists in learning source domain knowledge; , in, The value ranges from 1.5 to 2.5.

[0053] The clustering method in this embodiment introduces adaptive feature weighting coefficients, which automatically adjust the weights based on the intra-class dispersion and inter-class discriminative power of the features. This automatically amplifies effective features with high discriminative power and suppresses the interference of noisy features, thereby improving the stability of the clustering results. The algorithm can adaptively adjust the transfer strength based on the similarity of cluster distributions between the target and source domains. Clusters with high similarity receive stronger transfer weights, while interference from irrelevant clusters is weakened, avoiding the negative impact of blind transfer. A new cross-domain feature weight alignment term has been added. This approach forces the feature weights of the target domain and the matching source domain to remain consistent, improving the efficiency and reliability of knowledge transfer. The original sparse regularization term is improved to a normalized regularization term, avoiding the loss of effective information caused by excessively sparse feature weights. At the same time, it ensures the comparability between different feature weights, making the physical meaning of the feature weights clearer. The normalized penalty is used for the transfer confidence, rather than a fixed penalty, so that the algorithm pays more attention to the target clusters with high matching degree with the source domain, reduces the interference of low matching degree clusters, and improves the robustness of the overall transfer process.

[0054] S3124, Represent the class vector of each topic as a topic to obtain a list of keywords for each topic cluster; S3125, extract information from the keyword list of each topic cluster to obtain deep semantic feature information.

[0055] First, each word in the keyword list is converted into a low-dimensional, high-density semantic vector. Then, for the semantic vector set of a single topic cluster, through three steps of vector aggregation, core feature mining, and semantic dimension extraction, the deep semantic features that can represent the topic cluster are extracted. Finally, structured deep information such as semantic core words, semantic feature vectors, and semantic dimension labels are output.

[0056] Optionally, the step of representing the class vector of each topic to obtain a keyword list for each topic cluster includes: S31241, calculate the importance of each topic's class vector to obtain an importance score for each topic; The expression for the importance score is as follows: in, For thematic clusters Importance score Indicates a candidate word. Represents the set of all topic clusters. Indicates the inclusion word The number of other topic clusters, Indicates the number of topic clusters, Represents a topic cluster Size, word In thematic clusters Frequency of occurrence in; S31242, Based on the importance score of each topic, select the top N words as the topic keywords for each topic cluster, and obtain the keyword list for each topic cluster.

[0057] Example 2 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a knowledge classification device disclosed in an embodiment of the present invention. Figure 3 The described knowledge classification device is applied in the field of knowledge management and classification technology, and the embodiments of this invention are not limited thereto. Figure 3 As shown, the knowledge classification device may include the following operations: S301, Data Acquisition Module, used to acquire geographic information ontology conceptual models; S302, Ontology knowledge set construction module, used to process the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge includes N ontology knowledge information, each ontology knowledge information including an ontology knowledge title and ontology knowledge content; S303, Knowledge Classification Module, is used to process the ontology knowledge set to obtain knowledge classification results.

[0058] Example 3 Please see Figure 4 , Figure 4 This is a schematic diagram of another knowledge classification device disclosed in an embodiment of the present invention. Figure 4 The described knowledge classification device is applied in the field of knowledge management and classification technology, and the embodiments of this invention are not limited thereto. Figure 4 As shown, the knowledge classification device may include the following operations: Memory 401 storing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the knowledge classification method described in Embodiment 1.

[0059] Example 4 This invention discloses a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program enables a computer to perform the steps in the knowledge classification method described in Embodiment 1.

[0060] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0061] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0062] Finally, it should be noted that the knowledge classification method and apparatus disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. 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. Such 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.

Claims

1. A knowledge classification method, characterized in that, The method includes: S1, Obtain the geographic information ontology conceptual model; S2, process the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge includes N ontology knowledge information, each ontology knowledge information including ontology knowledge title and ontology knowledge content; S3, process the ontology knowledge set to obtain knowledge classification results.

2. The knowledge classification method according to claim 1, characterized in that, The process of processing the ontology knowledge set to obtain knowledge classification results includes: S31, The ontology knowledge set is processed to obtain a first-level classification result; the first-level classification result includes factual knowledge, rule knowledge, and decision-making knowledge; S32, perform secondary classification on the factual knowledge to obtain factual knowledge classification results; the factual knowledge classification results include general spatiotemporal knowledge, regional element knowledge, and event knowledge; S33, perform secondary classification on the rule-based knowledge to obtain the rule-based knowledge classification result; the rule-based knowledge classification result includes knowledge of natural laws and rules of change and knowledge of human society and human activities rules; S34, perform secondary classification on the decision-making knowledge to obtain the decision-making knowledge classification results; the decision-making knowledge classification results include acquisition decision-making knowledge, aggregation decision-making knowledge, protection decision-making knowledge, and safeguard decision-making knowledge; S35, integrate the factual knowledge classification results, the rule-based knowledge classification results, and the decision-based knowledge classification results to obtain the knowledge classification results.

3. The knowledge classification method according to claim 2, characterized in that, The processing of the ontology knowledge set to obtain the first-level classification result includes: S311, extract features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information; the title feature information includes title topic feature information and title semantic feature information. S312, extract features from the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information; S313, using a preset first classification model, the title feature information is processed to obtain a first classification result; S314, using a preset second classification model, the deep semantic feature information is processed to obtain the second classification result; S315, perform probability fusion on the first classification result and the second classification result to obtain the first-level classification result.

4. The knowledge classification method according to claim 3, characterized in that, The step of extracting features from the ontology knowledge titles in the ontology knowledge set to obtain title feature information includes: S3111, Standardize the ontology knowledge title to obtain a standardized ontology knowledge title; S3112, Perform word segmentation on the standardized ontology knowledge title to obtain a preprocessed ontology knowledge title; S3113, Perform feature extraction on the preprocessed ontology knowledge title to obtain title topic feature information; S3114, Semantic features are extracted from the preprocessed ontology knowledge title to obtain title semantic feature information.

5. The knowledge classification method according to claim 4, characterized in that, The step of extracting features from the preprocessed ontology knowledge title to obtain title topic feature information includes: Using a title-topic feature information extraction model, feature extraction is performed on the preprocessed ontology knowledge title to obtain title-topic feature information; The expression for the title / topic feature information extraction model is: in, For title and theme feature information, For attribute information, For attribute values, for The m-th element in for The m-th element in.

6. The knowledge classification method according to claim 3, characterized in that, The step of extracting features from the ontology knowledge content in the ontology knowledge set to obtain deep semantic feature information includes: S3121, Document embedding is performed on the ontology knowledge content to obtain an embedding vector; S3122, Perform dimensionality reduction processing on the embedding vector to obtain a dimensionality-reduced embedding vector; S3123, perform clustering processing on the reduced-dimensional embedding vector to obtain K class vectors of topics, where K is a positive integer; S3124, Represent the class vector of each topic as a topic to obtain a list of keywords for each topic cluster; S3125, extract information from the keyword list of each topic cluster to obtain deep semantic feature information.

7. The knowledge classification method according to claim 6, characterized in that, The process of representing the class vector of each topic to obtain a keyword list for each topic cluster includes: S31241, calculate the importance of each topic's class vector to obtain an importance score for each topic; The expression for the importance score is as follows: in, For thematic clusters Importance score Indicates a candidate word. Represents the set of all topic clusters. Indicates the inclusion word The number of other topic clusters, Indicates the number of topic clusters, Represents a topic cluster Size, word In thematic clusters Frequency of occurrence in; S31242, Based on the importance score of each topic, select the top N words as the topic keywords for each topic cluster, and obtain the keyword list for each topic cluster.

8. A knowledge classification device, characterized in that, The device includes: The data acquisition module is used to acquire the geographic information ontology conceptual model; The ontology knowledge set construction module is used to process the geographic information ontology conceptual model to obtain an ontology knowledge set; the ontology knowledge includes N ontology knowledge information, and each ontology knowledge information includes an ontology knowledge title and ontology knowledge content; The knowledge classification module is used to process the ontology knowledge set to obtain knowledge classification results.

9. A knowledge classification method apparatus, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the knowledge classification method as described in any one of claims 1-7.

10. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the knowledge classification method as described in any one of claims 1-7.

Images