Artificial intelligence-based knowledge graph proposition correction method and system

Abstract

The application discloses an artificial intelligence-based knowledge graph proposition correction method and system, relates to the technical field of artificial intelligence, and comprises the following steps: performing feature extraction on knowledge graph data, wherein the knowledge graph data comprises text data and structure data, and obtaining multi-modal feature data; constructing a local semantic verification model based on a graph attention network, performing local semantic consistency detection on the multi-modal feature data, and outputting a local semantic verification result; constructing a global semantic verification model based on a knowledge graph embedding technology, performing global semantic consistency detection on the multi-modal feature data, and outputting a global semantic verification result; and inputting the local semantic verification result and the global semantic verification result into a conditional random field model to generate a semantic correction result of the knowledge graph. The application not only significantly improves the accuracy and usability of the knowledge graph, but also provides reliable technical support for the maintenance and optimization of the knowledge graph, and has important practical value.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a knowledge graph-based proposition error correction method and system. Background Technology

[0002] With the rapid development of artificial intelligence technology, knowledge graphs, as a structured knowledge representation method, play an important role in fields such as intelligent question answering, recommendation systems, and information retrieval. However, due to factors such as noise in the knowledge acquisition process, the dynamic nature of knowledge evolution, and the complexity of knowledge representation, semantic errors inevitably exist in knowledge graphs. Traditional knowledge graph error correction methods mainly rely on rule matching and statistical analysis. These methods often focus only on single-dimensional features, such as considering only textual semantics or graph structure features, resulting in limited error correction performance. Furthermore, most existing error correction techniques employ independent local or global verification strategies, lacking comprehensive analysis of multi-level semantic relationships, making it difficult to accurately identify and correct complex semantic errors in knowledge graphs. In addition, traditional methods have low computational efficiency when processing large-scale knowledge graphs and insufficient robustness to noisy data.

[0003] With the advancement of deep learning technology, neural network-based knowledge graph processing methods have demonstrated excellent performance. However, existing deep learning-based knowledge graph error correction methods still suffer from several problems: First, they lack an effective fusion mechanism for multimodal features of the knowledge graph, making it difficult to fully utilize textual and structural information; second, during semantic verification, existing methods often neglect the interaction between local context and global knowledge constraints, leading to insufficient accuracy and reliability of verification results; and third, they lack adaptive error correction strategies, making it difficult to provide precise correction suggestions based on different types of semantic errors. Therefore, designing a knowledge graph error correction method that can comprehensively utilize multimodal features and fuse local and global semantic verification has become an important research topic.

[0004] Existing knowledge graph error correction technologies suffer from problems such as insufficient feature utilization, simplistic verification strategies, and imprecise correction suggestions. This invention provides an artificial intelligence-based knowledge graph proposition error correction method and system, aiming to improve the accuracy and practicality of knowledge graph error correction through multimodal feature extraction, two-layer semantic verification, and adaptive error correction fusion. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] Therefore, this invention provides a knowledge graph proposition error correction method and system based on artificial intelligence, which can solve the problems mentioned in the background art.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] In a first aspect, the present invention provides a knowledge graph proposition error correction method based on artificial intelligence, which includes extracting features from knowledge graph data, wherein the knowledge graph data includes text data and structural data, to obtain multimodal feature data;

[0010] A local semantic verification model is constructed based on a graph attention network to perform local semantic consistency detection on the multimodal feature data and output the local semantic verification result.

[0011] A global semantic verification model is constructed based on knowledge graph embedding technology, and global semantic consistency detection is performed on the multimodal feature data to output global semantic verification results.

[0012] The local semantic verification results and the global semantic verification results are input into the conditional random field model to generate semantic error correction results for the knowledge graph.

[0013] As a preferred embodiment of the knowledge graph-based propositional error correction method based on artificial intelligence described in this invention, the feature extraction includes:

[0014] The text data in the knowledge graph is segmented into head entity text, relation text, and tail entity text according to the triplet structure. The calculation formula is as follows:

[0015] ;

[0016] Where Triple represents a triple in a knowledge graph, h represents a head entity, r represents a relation, and t represents a tail entity;

[0017] The head entity text, relation text, and tail entity text are encoded separately to obtain text feature vectors. The encoding formula is as follows:

[0018] ;

[0019] ;

[0020] ;

[0021] in, Represents the encoding function. This represents the pre-trained language model encoder;

[0022] A multi-head self-attention matrix based on attention weights is constructed, and semantic correlation is calculated on the text feature vector to obtain a semantically enhanced feature vector;

[0023] The structural data in the knowledge graph is converted into an adjacency matrix, and the adjacency matrix is ​​convolved based on a graph neural network to obtain the graph structure feature vector.

[0024] A multilayer perceptron is used to perform a spatial mapping transformation on the graph structure feature vectors to obtain structure-enhanced feature vectors;

[0025] A feature fusion layer is constructed to concatenate the semantically enhanced feature vector and the structurally enhanced feature vector into a fused feature matrix.

[0026] The fused feature matrix is ​​reorganized using a residual connection structure to output multimodal feature data.

[0027] As a preferred embodiment of the knowledge graph proposition error correction method based on artificial intelligence described in this invention, the formula for calculating the semantically enhanced feature vector is as follows:

[0028] ;

[0029] ;

[0030] Where Attention represents the attention function, Q represents the query matrix, K represents the key matrix, V represents the value matrix, d represents the feature dimension, MultiHead represents the multi-head attention function, and head i W represents the i-th attention head. O This indicates the output projection matrix; Concat represents the concatenation operation.

[0031] The formula for calculating the structurally enhanced feature vector is as follows:

[0032] ;

[0033] Where MLP() represents the multilayer perceptron function, and x represents the input feature. and Represents the weight matrix. and Represents the bias vector, and ReLU() represents the activation function;

[0034] The calculation formula for the feature recombination is as follows:

[0035] ;

[0036] ;

[0037] in, Indicates fusion characteristics, Semantic features represents structural features, || represents feature concatenation operation, LayerNorm() represents layer normalization function, SubLayer() represents sublayer network, and Output represents final output features.

[0038] As a preferred embodiment of the knowledge graph proposition error correction method based on artificial intelligence described in this invention, the local semantic consistency detection includes:

[0039] Extract the feature representation of the current triplet to be verified from the multimodal feature data, and obtain the set of neighboring triplets directly connected to the triplet.

[0040] The computational unit constructs the graph attention network and assigns initial attention weight coefficients to each triple in the set of neighbor triples. The calculation formula for assigning initial attention weights to the set of neighbor triples is as follows:

[0041] ;

[0042] in, Indicates from node To the node The initial attention coefficient, Represents the feature transformation weight matrix. and Representing nodes respectively and nodes eigenvectors, represents the attention vector parameters, and || represents the vector concatenation operation. Represents a node The set of neighboring nodes, where LeakyReLU represents the activation function;

[0043] For each triplet in the neighbor triplet set, calculate the semantic similarity matrix with the current triplet to be verified, and generate a triplet comparison score.

[0044] Based on the triplet comparison score update of the attention weight coefficient in the graph attention network, key neighbor triplets with semantic relevance higher than a preset threshold are selected.

[0045] The features of the key neighbor triplet and the features of the current triplet to be verified are used to calculate the local semantic consistency, and the local consistency feature vector is obtained.

[0046] A bidirectional gated recurrent unit network is constructed, and the local consistency feature vector is input into the network for temporal dependency analysis to obtain local semantic verification results.

[0047] As a preferred embodiment of the knowledge graph proposition error correction method based on artificial intelligence described in this invention, the temporal dependency analysis includes:

[0048] Construct a bidirectional GRU network to process local consistency features;

[0049] The validation results are obtained by fusing bidirectional features, and the calculation formula is as follows:

[0050]

[0051] in, This represents the final local semantic verification result vector. This represents the output layer weight matrix. This represents the final hidden state of the forward GRU. This represents the final hidden state of the backward GRU. Indicates the bias term. This represents the vector concatenation operation, and softmax represents the normalized activation function.

[0052] As a preferred embodiment of the knowledge graph proposition error correction method based on artificial intelligence described in this invention, the global semantic consistency detection includes:

[0053] A multidimensional tensor decomposition model is constructed to map the multimodal feature data to a unified semantic space, thereby obtaining an initial semantic embedding vector;

[0054] Establish a global semantic constraint rule matrix for the knowledge graph, which includes entity type constraints, relation transitivity constraints, and mutual exclusion constraints;

[0055] The global semantic constraint rule matrix for constructing a knowledge graph includes,

[0056] Construct entity type constraint matrix Define type compatibility between entities;

[0057] Constructing the relation transitivity constraint matrix Describes the logical reasoning rules between relationships;

[0058] The comprehensive constraint matrix is ​​generated using the following formula:

[0059] ;

[0060] in, This represents the final global constraint matrix. Represents the type constraint matrix. Represents the transitivity constraint matrix. Represents the mutual exclusion constraint matrix. Indicates the weight coefficient of each constraint;

[0061] The initial semantic embedding vector is combined with the global semantic constraint rule matrix to perform tensor operations, generating a constraint enhancement vector.

[0062] A scoring model based on an energy function is constructed to calculate the semantic association score between the constraint enhancement vector and the existing triples in the knowledge graph. The calculation formula for the energy function scoring model is as follows:

[0063] ;

[0064] in, This represents the energy score of the triple (h,r,t). The constraint weights are represented by || ||2, which represents the L2 norm.

[0065] A set of comparison samples is generated using negative sampling techniques. The semantic association score is then evaluated against the scores of the comparison samples to form a global consistency score.

[0066] The global consistency score is grouped based on a hierarchical clustering algorithm to identify potential semantic conflict points, output global semantic verification results, and define a consistency clustering criterion. The calculation formula is as follows:

[0067] ;

[0068] in, Represents a triplet Verification results Represents a triplet Consistency score, This indicates the high confidence threshold. This indicates the low confidence threshold.

[0069] As a preferred embodiment of the knowledge graph proposition error correction method based on artificial intelligence described in this invention, the semantic error correction result of the generated knowledge graph includes:

[0070] The local semantic verification result and the global semantic verification result are assigned weight coefficients respectively, and the fused feature vector is obtained by weighted averaging.

[0071] A directed graph is used to represent the association between triples. The fused feature vector is used as the initial state value of the node in the graph. The state transition probability between nodes is calculated using the normalized adjacency matrix to generate the error state probability distribution.

[0072] The probability distribution of the error states is clustered into triples according to a preset probability threshold to obtain high-confidence error clusters, low-confidence error clusters, and correct clusters;

[0073] Retrieve triplets from the knowledge graph that have the same head or tail entity as the triplets in the high-confidence error cluster, and construct a set of correction candidates;

[0074] For each candidate in the modified candidate set, calculate its conformity score with the type constraints, logical rules and global consistency in the original knowledge graph, and generate a candidate score list;

[0075] Based on the scores of the candidate scoring list, output an error correction report containing the error type, suggested correction triples and their scores.

[0076] Secondly, the present invention provides an artificial intelligence-based knowledge graph proposition error correction system, which includes: a multimodal feature extraction module, a local semantic verification module, a global semantic verification module, and a semantic error correction fusion module;

[0077] The multimodal feature extraction module is used to extract features from knowledge graph data, which includes text data and structural data, to obtain multimodal feature data.

[0078] The local semantic verification module is used to construct a local semantic verification model based on a graph attention network, perform local semantic consistency detection on the multimodal feature data, and output the local semantic verification result.

[0079] The global semantic verification module is used to construct a global semantic verification model based on knowledge graph embedding technology, perform global semantic consistency detection on the multimodal feature data, and output the global semantic verification result;

[0080] The semantic error correction fusion module is used to input the local semantic verification result and the global semantic verification result into the conditional random field model to generate the semantic error correction result of the knowledge graph.

[0081] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a knowledge graph proposition error correction method based on artificial intelligence.

[0082] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements the steps of a knowledge graph proposition error correction method based on artificial intelligence.

[0083] Compared with existing technologies, the advantages of this invention are as follows: First, by decomposing knowledge graph data into textual and structural data through feature extraction, and combining multi-head self-attention mechanisms and graph neural networks for dual enhancement, the comprehensiveness and accuracy of feature representation are achieved. Second, a local semantic verification model constructed through a graph attention network, utilizing dynamic attention weight allocation and a bidirectional gated recurrent unit network, achieves accurate judgment of semantic consistency between neighboring nodes. Third, a global semantic verification model constructed through knowledge graph embedding technology, integrating multidimensional tensor decomposition and multiple semantic constraint rules, achieves systematic verification of the overall semantic consistency of the knowledge graph. Finally, a conditional random field model integrates local and global verification results, combined with differentiated error correction strategies and candidate scoring mechanisms, forming a complete error correction closed loop. This method not only significantly improves the accuracy and usability of knowledge graphs but also provides reliable technical support for the maintenance and optimization of knowledge graphs, possessing significant practical value. Attached Figure Description

[0084] To more clearly illustrate the technical solutions of 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.

[0085] Figure 1 A flowchart illustrating a knowledge graph-based proposition error correction method and system based on artificial intelligence, provided in one embodiment of the present invention;

[0086] Figure 2 This is an internal structural diagram of a computer device for an artificial intelligence-based knowledge graph proposition error correction system provided in one embodiment of the present invention. Detailed Implementation

[0087] To make the above-mentioned objects, features, and advantages of the present invention more readily understood, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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 protection scope of the present invention.

[0088] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0089] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0090] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a knowledge graph proposition error correction method based on artificial intelligence, including:

[0091] Before detailing the embodiments of this application, some related concepts will be explained for clarity.

[0092] Multi-head self-attention mechanism: a feature processing method in deep learning. By projecting input features in parallel to multiple subspaces and independently calculating attention weights, each "head" can focus on different aspects of the features. Finally, the results of multiple heads are combined, thereby enhancing the model's ability to understand the input data from multiple perspectives. It is particularly suitable for processing complex feature data that requires consideration of multiple correlations.

[0093] DBSCAN clustering algorithm: a density-based clustering method that groups density-connected data points into the same cluster by setting two parameters: neighborhood radius and minimum number of samples. It does not require pre-specifying the number of clusters, can discover clusters of arbitrary shapes, and has strong anti-interference ability against noisy data, making it particularly suitable for processing non-uniformly distributed datasets.

[0094] Bidirectional gated recurrent unit network: an improved recurrent neural network structure that includes two processing directions, forward and backward. By controlling the flow of information through update gates and reset gates, it can simultaneously consider the historical and future information of sequence data, overcome the gradient vanishing problem of traditional recurrent neural networks when processing long sequences, and is particularly suitable for sequence data processing tasks that require consideration of contextual dependencies.

[0095] Conditional Random Field (CRF) model: A probabilistic graphical model that takes into account the dependencies and contextual information between elements in the input sequence. It achieves globally optimal sequence prediction by modeling the transition probabilities between sequence labels, supports joint inference of multiple features, and is widely used in sequence labeling and structured prediction tasks.

[0096] Figure 1 A flowchart illustrating a knowledge graph-based proposition error correction method and system based on artificial intelligence is provided, including:

[0097] S1: Extract features from the knowledge graph data, which includes text data and structural data, to obtain multimodal feature data;

[0098] Furthermore, feature extraction includes,

[0099] The text data in the knowledge graph is segmented into head entity text, relation text, and tail entity text according to the triple structure, specifically:

[0100] ;

[0101] Where Triple represents a triple in a knowledge graph, h represents a head entity, r represents a relation, and t represents a tail entity;

[0102] The head entity text, relation text, and tail entity text are encoded separately to obtain text feature vectors. The encoding formula is as follows:

[0103] ;

[0104] ;

[0105] ;

[0106] in, Represents the encoding function. This represents the pre-trained language model encoder;

[0107] A multi-head self-attention matrix based on attention weights is constructed, and semantic correlation is calculated on the text feature vector to obtain a semantically enhanced feature vector;

[0108] ;

[0109] ;

[0110] Where Attention represents the attention function, Q represents the query matrix, K represents the key matrix, V represents the value matrix, d represents the feature dimension, MultiHead represents the multi-head attention function, and head i W represents the i-th attention head. O This indicates the output projection matrix; Concat represents the concatenation operation.

[0111] The structural data in the knowledge graph is converted into an adjacency matrix, and the adjacency matrix is ​​convolved based on a graph neural network to obtain the graph structure feature vector.

[0112] A multilayer perceptron is used to perform a spatial mapping transformation on the graph structure feature vector to obtain a structure-enhanced feature vector. The calculation formula is as follows:

[0113] ;

[0114] Where MLP() represents the multilayer perceptron function, and x represents the input feature. and Represents the weight matrix. and Represents the bias vector, and ReLU() represents the activation function;

[0115] A feature fusion layer is constructed to concatenate the semantically enhanced feature vector and the structurally enhanced feature vector into a fused feature matrix.

[0116] The fused feature matrix is ​​reorganized using a residual connection structure to output multimodal feature data. The calculation formula is as follows:

[0117] ;

[0118] ;

[0119] in, Indicates fusion characteristics, Semantic features represents structural features, || represents feature concatenation operation, LayerNorm() represents layer normalization function, SubLayer() represents sublayer network, and Output represents final output features.

[0120] S2: Construct a local semantic verification model based on a graph attention network, perform local semantic consistency detection on the multimodal feature data, and output the local semantic verification result;

[0121] Furthermore, local semantic consistency detection includes,

[0122] Extract the feature representation of the current triplet to be verified from the multimodal feature data, and obtain the set of neighboring triplets directly connected to the triplet.

[0123] The computational unit constructs the graph attention network and assigns initial attention weight coefficients to each triple in the set of neighbor triples. The calculation formula for assigning initial attention weights to the set of neighbor triples is as follows:

[0124] ;

[0125] in, Indicates from node To the node The initial attention coefficient, Represents the feature transformation weight matrix. and Representing nodes respectively and nodes eigenvectors, represents the attention vector parameters, and || represents the vector concatenation operation. Represents a node The set of neighboring nodes, where LeakyReLU represents the activation function;

[0126] For each triplet in the neighbor triplet set, calculate the semantic similarity matrix with the current triplet to be verified, and generate a triplet comparison score.

[0127] Based on the triplet comparison score update of the attention weight coefficient in the graph attention network, key neighbor triplets with semantic relevance higher than a preset threshold are selected.

[0128] The features of the key neighbor triplet and the features of the current triplet to be verified are used to calculate the local semantic consistency, and the local consistency feature vector is obtained.

[0129] A bidirectional gated recurrent unit network is constructed, and the local consistency feature vector is input into the network for temporal dependency analysis to obtain local semantic verification results.

[0130] Further, time-series dependency analysis includes:

[0131] Construct a bidirectional GRU network to process local consistency features;

[0132] The validation results are obtained by fusing bidirectional features, and the calculation formula is as follows:

[0133] ;

[0134] in, This represents the final local semantic verification result vector. This represents the output layer weight matrix. This represents the final hidden state of the forward GRU. This represents the final hidden state of the backward GRU. Indicates the bias term. This represents the vector concatenation operation, and softmax represents the normalized activation function.

[0135] S3: Construct a global semantic verification model based on knowledge graph embedding technology, perform global semantic consistency detection on the multimodal feature data, and output the global semantic verification result;

[0136] Furthermore, global semantic consistency detection includes,

[0137] A multidimensional tensor decomposition model is constructed to map the multimodal feature data to a unified semantic space, thereby obtaining an initial semantic embedding vector;

[0138] Establish a global semantic constraint rule matrix for the knowledge graph, which includes entity type constraints, relation transitivity constraints, and mutual exclusion constraints;

[0139] The global semantic constraint rule matrix for constructing a knowledge graph includes,

[0140] Construct entity type constraint matrix Define type compatibility between entities;

[0141] Constructing the relation transitivity constraint matrix Describes the logical reasoning rules between relationships;

[0142] The comprehensive constraint matrix is ​​generated using the following formula:

[0143] ;

[0144] in, This represents the final global constraint matrix. Represents the type constraint matrix. Represents the transitivity constraint matrix. Represents the mutual exclusion constraint matrix. Indicates the weight coefficient of each constraint;

[0145] The initial semantic embedding vector is combined with the global semantic constraint rule matrix to perform tensor operations, generating a constraint enhancement vector.

[0146] A scoring model based on an energy function is constructed to calculate the semantic association score between the constraint enhancement vector and the existing triples in the knowledge graph. The calculation formula for the energy function scoring model is as follows:

[0147] ;

[0148] in, This represents the energy score of the triple (h,r,t). The constraint weights are represented by || ||2, which represents the L2 norm.

[0149] A set of comparison samples is generated using negative sampling techniques. The semantic association score is then evaluated against the scores of the comparison samples to form a global consistency score.

[0150] The global consistency score is grouped based on a hierarchical clustering algorithm to identify potential semantic conflict points, output global semantic verification results, and define a consistency clustering criterion. The calculation formula is as follows:

[0151] ;

[0152] in, Represents a triplet Verification results Represents a triplet Consistency score, This indicates the high confidence threshold. This indicates the low confidence threshold.

[0153] S4: Input the local semantic verification result and the global semantic verification result into the conditional random field model to generate the semantic error correction result of the knowledge graph.

[0154] Furthermore, the semantic error correction results of the generated knowledge graph include:

[0155] The local semantic verification result and the global semantic verification result are assigned weight coefficients α and β, respectively. A fused feature vector F is obtained by weighted averaging. The weight coefficients α and β are dynamically adjusted based on the confidence score of the verification result, and satisfy α + β = 1. It should be noted that this invention first sets the initial weight coefficients α = 0.5 and β = 0.5, and normalizes the confidence scores in the local and global semantic verification results. When the confidence score of the local semantic verification result is greater than 0.7, the value of α is increased by 0.1, and the value of β is decreased by 0.1 accordingly; conversely, when the confidence score of the global semantic verification result is greater than 0.7, the value of β is increased by 0.1, and the value of α is decreased by 0.1 accordingly. Finally, the fused feature vector is calculated by weighted summation, retaining the top 50% of the feature values ​​with the highest confidence scores in the fused feature vector.

[0156] A directed graph is used to represent the relationships between triples. The fused feature vector is used as the initial state value of the nodes in the graph. The state transition probability between nodes is calculated using the normalized adjacency matrix W to generate the error state probability distribution P. Specifically, a directed graph G is constructed, where the nodes represent triples in the knowledge graph. The fused feature vector is used as the initial state value of the nodes. If two triples share a head entity or a tail entity, a directed edge is established between the corresponding nodes. The weight of the edge is obtained by calculating the number of shared entities between the nodes divided by the total number of entities, and a normalized adjacency matrix is ​​constructed. A random walk algorithm is used for 5 rounds of iterative calculation to update the error state probability value of the nodes until the node state probability distribution tends to stabilize.

[0157] Based on the DBSCAN clustering algorithm, the error state probability distribution P is clustered into triples according to preset probability thresholds θ1 and θ2 to obtain high-confidence error clusters, low-confidence error clusters, and correct clusters;

[0158] Retrieve triples from the knowledge graph that have the same head or tail entity as the triples in the high-confidence error cluster, and construct a set of corrected candidate options M; specifically, for each triple in the high-confidence error cluster, retrieve all triples containing the same head or tail entity from the original knowledge graph; retain candidate options with the same relation type as the original triples to form a set of corrected candidate options; and label the overlapping entity information of each candidate option with the original triples.

[0159] For each candidate in the modified candidate set M, its compliance score with the type constraints, logical rules, and global consistency in the original knowledge graph is calculated, generating a candidate score list S. Specifically, a scoring standard for the three types of constraints is constructed: entity type constraints check whether the correspondence between entity types and relations in the candidate is reasonable, scoring 1 point if satisfied and 0 points if not; relational logic constraints verify whether the candidate violates the transitivity and symmetry of relations, scoring 1 point if there is no violation and 0 points if there is a violation; global consistency constraints count the number of supporting evidences for the candidate in the knowledge graph and assign points according to the proportion of evidence. The weighted average of the scores of the three types of constraints is used as the final score of the candidate.

[0160] Based on the scores of the candidate scoring list S, output an error correction result report R containing the error type, suggested correction triplet and its score.

[0161] Example 2, refer to Figure 2 This is the second embodiment of the present invention. This embodiment also provides an artificial intelligence-based knowledge graph proposition correction system, including: a multimodal feature extraction module, a local semantic verification module, a global semantic verification module, and a semantic correction fusion module;

[0162] The multimodal feature extraction module is used to extract features from knowledge graph data, which includes text data and structural data, to obtain multimodal feature data.

[0163] The local semantic verification module is used to construct a local semantic verification model based on a graph attention network, perform local semantic consistency detection on the multimodal feature data, and output the local semantic verification result.

[0164] The global semantic verification module is used to construct a global semantic verification model based on knowledge graph embedding technology, perform global semantic consistency detection on the multimodal feature data, and output the global semantic verification result.

[0165] The semantic error correction fusion module is used to input the local semantic verification result and the global semantic verification result into the conditional random field model to generate the semantic error correction result of the knowledge graph.

[0166] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as follows. Figure 2As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an artificial intelligence-based knowledge graph proposition error correction method. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0167] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it performs the following steps: extracting features from knowledge graph data, wherein the knowledge graph data includes text data and structural data, to obtain multimodal feature data.

[0168] A local semantic verification model is constructed based on a graph attention network to perform local semantic consistency detection on the multimodal feature data and output the local semantic verification result.

[0169] A global semantic verification model is constructed based on knowledge graph embedding technology, and global semantic consistency detection is performed on the multimodal feature data to output global semantic verification results.

[0170] The local semantic verification results and the global semantic verification results are input into the conditional random field model to generate semantic error correction results for the knowledge graph.

[0171] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0172] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0173] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0174] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0175] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0176] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0177] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A knowledge graph error correction method based on artificial intelligence, characterized in that: This includes extracting features from knowledge graph data, which includes text data and structural data, to obtain multimodal feature data; A local semantic verification model is constructed based on a graph attention network to perform local semantic consistency detection on the multimodal feature data and output the local semantic verification result. A global semantic verification model is constructed based on knowledge graph embedding technology, and global semantic consistency detection is performed on the multimodal feature data to output global semantic verification results. The local semantic verification results and the global semantic verification results are input into the conditional random field model to generate semantic error correction results for the knowledge graph. The global semantic consistency detection includes, A multidimensional tensor decomposition model is constructed to map the multimodal feature data to a unified semantic space, thereby obtaining an initial semantic embedding vector; Establish a global semantic constraint rule matrix for the knowledge graph, which includes entity type constraints, relation transitivity constraints, and mutual exclusion constraints; The global semantic constraint rule matrix for constructing a knowledge graph includes, Construct entity type constraint matrix Define type compatibility between entities; Constructing the relation transitivity constraint matrix Describes the logical reasoning rules between relationships; The comprehensive constraint matrix is ​​generated using the following formula: ； in, This represents the final global constraint matrix. Represents the type constraint matrix. Represents the transitivity constraint matrix. Represents the mutual exclusion constraint matrix. Indicates the weight coefficient of each constraint; The initial semantic embedding vector is combined with the global semantic constraint rule matrix to perform tensor operations, generating a constraint enhancement vector. A scoring model based on an energy function is constructed to calculate the semantic association score between the constraint enhancement vector and the existing triples in the knowledge graph. The calculation formula for the energy function scoring model is as follows: ； in, This represents the energy score of the triple (h,r,t). The constraint weights are represented by || ||2, which represents the L2 norm. A set of comparison samples is generated using negative sampling techniques. The semantic association score is then evaluated against the scores of the comparison samples to form a global consistency score. The global consistency score is grouped based on a hierarchical clustering algorithm to identify potential semantic conflict points, output global semantic verification results, and define a consistency clustering criterion. The calculation formula is as follows: ； in, Represents a triplet Verification results Represents a triplet Consistency score, This indicates the high confidence threshold. Indicates the low confidence threshold; The semantic error correction results of the generated knowledge graph include, The local semantic verification result and the global semantic verification result are assigned weight coefficients respectively, and the fused feature vector is obtained by weighted averaging. A directed graph is used to represent the association between triples. The fused feature vector is used as the initial state value of the node in the graph. The state transition probability between nodes is calculated using the normalized adjacency matrix to generate the error state probability distribution. The probability distribution of the error states is clustered into triples according to a preset probability threshold to obtain high-confidence error clusters, low-confidence error clusters, and correct clusters; Retrieve triplets from the knowledge graph that have the same head or tail entity as the triplets in the high-confidence error cluster, and construct a set of correction candidates; For each candidate in the modified candidate set, calculate its conformity score with the type constraints, logical rules and global consistency in the original knowledge graph, and generate a candidate score list; Based on the scores of the candidate scoring list, output an error correction report containing the error type, suggested correction triples and their scores.

2. The knowledge graph error correction method based on artificial intelligence as described in claim 1, characterized in that: The feature extraction includes, The text data in the knowledge graph is segmented into head entity text, relation text, and tail entity text according to the triplet structure. The calculation formula is as follows: ； Where Triple represents a triple in a knowledge graph, h represents a head entity, r represents a relation, and t represents a tail entity; The head entity text, relation text, and tail entity text are encoded separately to obtain text feature vectors. The encoding formula is as follows: ； ； ； in, Represents the encoding function. This represents the pre-trained language model encoder; A multi-head self-attention matrix based on attention weights is constructed, and semantic correlation is calculated on the text feature vector to obtain a semantically enhanced feature vector; The structural data in the knowledge graph is converted into an adjacency matrix, and the adjacency matrix is ​​convolved based on a graph neural network to obtain the graph structure feature vector. A multilayer perceptron is used to perform a spatial mapping transformation on the graph structure feature vectors to obtain structure-enhanced feature vectors; A feature fusion layer is constructed to concatenate the semantically enhanced feature vector and the structurally enhanced feature vector into a fused feature matrix. The fused feature matrix is ​​reorganized using a residual connection structure to output multimodal feature data.

3. The knowledge graph error correction method based on artificial intelligence as described in claim 2, characterized in that: The formula for calculating the semantically enhanced feature vector is as follows: ； ； Where Attention represents the attention function, Q represents the query matrix, K represents the key matrix, V represents the value matrix, d represents the feature dimension, MultiHead represents the multi-head attention function, and head i W represents the i-th attention head. O This indicates the output projection matrix; Concat represents the concatenation operation. The formula for calculating the structurally enhanced feature vector is as follows: ； Where MLP() represents the multilayer perceptron function, and x represents the input feature. and Represents the weight matrix. and Represents the bias vector, and ReLU() represents the activation function; The calculation formula for the feature recombination is as follows: ； ； in, Indicates fusion characteristics, Semantic features represents structural features, || represents feature concatenation operation, LayerNorm() represents layer normalization function, SubLayer() represents sublayer network, and Output represents final output features.

4. The knowledge graph error correction method based on artificial intelligence as described in claim 3, characterized in that: The local semantic consistency detection includes, Extract the feature representation of the current triplet to be verified from the multimodal feature data, and obtain the set of neighboring triplets directly connected to the triplet. The computational unit constructs the graph attention network and assigns initial attention weight coefficients to each triple in the set of neighbor triples. The calculation formula for assigning initial attention weights to the set of neighbor triples is as follows: ； in, Indicates from node To the node The initial attention coefficient, Represents the feature transformation weight matrix. and Representing nodes respectively and nodes eigenvectors, represents the attention vector parameters, and || represents the vector concatenation operation. Represents a node The set of neighboring nodes, where LeakyReLU represents the activation function; For each triplet in the neighbor triplet set, calculate the semantic similarity matrix with the current triplet to be verified, and generate a triplet comparison score. Based on the triplet comparison score update of the attention weight coefficient in the graph attention network, key neighbor triplets with semantic relevance higher than a preset threshold are selected. The features of the key neighbor triplet and the features of the current triplet to be verified are used to calculate the local semantic consistency, and the local consistency feature vector is obtained. A bidirectional gated recurrent unit network is constructed, and the local consistency feature vector is input into the network for temporal dependency analysis to obtain local semantic verification results.

5. The knowledge graph error correction method based on artificial intelligence as described in claim 4, characterized in that: The time-series dependency analysis includes: Construct a bidirectional GRU network to process local consistency features; The validation results are obtained by fusing bidirectional features, and the calculation formula is as follows: ； in, This represents the final local semantic verification result vector. This represents the output layer weight matrix. This represents the final hidden state of the forward GRU. This represents the final hidden state of the backward GRU. Indicates the bias term. This represents the vector concatenation operation, and softmax represents the normalized activation function.

6. An artificial intelligence-based knowledge graph error correction system, based on the artificial intelligence-based knowledge graph error correction method according to any one of claims 1 to 5, characterized in that: It includes a multimodal feature extraction module, a local semantic verification module, a global semantic verification module, and a semantic error correction fusion module; The multimodal feature extraction module is used to extract features from knowledge graph data, which includes text data and structural data, to obtain multimodal feature data. The local semantic verification module is used to construct a local semantic verification model based on a graph attention network, perform local semantic consistency detection on the multimodal feature data, and output the local semantic verification result. The global semantic verification module is used to construct a global semantic verification model based on knowledge graph embedding technology, perform global semantic consistency detection on the multimodal feature data, and output the global semantic verification result; The semantic error correction fusion module is used to input the local semantic verification result and the global semantic verification result into the conditional random field model to generate the semantic error correction result of the knowledge graph.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the knowledge graph error correction method based on artificial intelligence as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the knowledge graph error correction method based on artificial intelligence as described in any one of claims 1 to 5.

Images