Dynamic knowledge graph fusing method and system based on multi-modal data fusion

Abstract

This invention proposes a dynamic knowledge graph circuit breaker optimization method and system based on multimodal data fusion, belonging to the field of multi-source heterogeneous data processing technology. The method includes: acquiring multi-source data at the current moment; evaluating the multi-source data input at the current moment using entanglement relationships to obtain a dynamic quality score; when the blank point acceptance rate is less than a second threshold or the dynamic quality score is less than a first threshold, activating a dynamic knowledge graph circuit breaker mechanism and switching the knowledge graph algorithm using an algorithm selection strategy; collecting multi-source data after the algorithm switch and optimizing the first threshold, second threshold, and algorithm selection strategy using historical multi-source data. The method of this invention can automatically switch the corresponding algorithm, and the switching threshold can be updated in real time, ensuring the reliability of the knowledge graph generation results and eliminating the problem of reduced accuracy of the knowledge graph due to terminology conflicts.

Description

Technical Field

[0001] This invention relates to the field of multi-source heterogeneous data processing technology, specifically providing a dynamic knowledge graph circuit breaker optimization method and system based on multimodal data fusion. Background Technology

[0002] Knowledge graphs are widely used in areas such as technology insights and competitive intelligence analysis, but traditional systems are mostly based on static rules or a single data source (such as patent data only). Some commercial platforms use predefined clustering algorithms (such as K-means) to process patent data, but do not address the dynamic fusion of multimodal data and adaptive algorithm switching. For example, CN112527826A, "A Knowledge Graph Update Method Based on Quality Assessment," triggers graph updates through data quality scoring, but only for a single data source, failing to address the issues of multimodal data conflicts and real-time algorithm switching. Existing knowledge graph technologies face the following core problems when processing multi-source heterogeneous data (such as patents, papers, and market reports): 1. High sensitivity to data quality fluctuations: When the quality of input data declines (e.g., the blank point acceptance rate is lower than the threshold), the reliability of the generated map drops sharply, and the processing strategy cannot be dynamically switched.

[0003] 2. Inflexible algorithms: Core algorithms such as clustering and entity alignment rely on fixed rules and cannot adapt to data characteristics, resulting in a high rate of invalid suggestions.

[0004] 3. Low efficiency of multimodal data fusion: The lack of real-time resolution mechanisms for semantic conflicts and temporal differences among patent, academic, and market data affects the speed and accuracy of map updates.

[0005] 4. Data quality assessment relies solely on static indicators (such as completeness and timeliness), failing to consider the dynamic correlations between multiple data sources. Algorithm switching depends on manually set thresholds and cannot be automatically optimized based on business objectives (such as the efficiency of blank data discovery). Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a dynamic knowledge graph circuit breaker optimization method and system based on multimodal data fusion.

[0007] Firstly, a dynamic knowledge graph circuit breaking optimization method based on multimodal data fusion is proposed, including: Step S1: Obtain multi-source data at the current moment, evaluate the multi-source data input at the current moment using entanglement relationship, and obtain a dynamic quality score; Step S2: When the dynamic quality score is greater than or equal to the first threshold, the multi-source data at the current moment is semantically aligned using a multilingual alignment anchor matrix to obtain the normalized entities and relations at the current moment. Step S3: When the blank point adoption rate is less than the second threshold or the dynamic quality score is less than the first threshold, the dynamic knowledge graph circuit breaker mechanism is activated, and the algorithm selection strategy is used to switch the knowledge graph algorithm. Step S4: Collect multi-source data after switching algorithms, proceed to step S1, and use historical multi-source data to optimize the first threshold, the second threshold, and the algorithm selection strategy.

[0008] The entanglement relationship is used to evaluate the multi-source data input at the current moment to obtain a dynamic quality score, calculated as follows: Q score (t)=Q initial e λt +ΔQ event ; Among them, Q score (t) represents the dynamic quality score at time t, Q initial The dynamic quality score at the initial time, λ is the weighting coefficient, and ΔQ event Let e ​​be the quantum superposition state of the event, i.e., the entanglement relationship of the multi-source data at the current moment, and let e be the natural logarithm.

[0009] The events mentioned include the completeness, authority, and timeliness of multi-source data at the current moment.

[0010] The multilingual alignment anchor matrix, based on a technical terminology database, is constructed by including: Collect core terms in a specific field and their standard expressions in different languages ​​to form a technical terminology database; A multilingual pre-trained model is used to convert each term into a high-dimensional semantic vector; Align and merge the high-dimensional semantic vectors of the same term in different languages ​​to form term-vector pairs; A multilingual alignment anchor matrix is ​​constructed using term-vector pairs, where terms are row vectors of the multilingual alignment anchor matrix.

[0011] The anchor point matrix, which employs multilingual alignment, performs semantic alignment of the multi-source data at the current moment to obtain the normalized entities and relations at the current moment, including: The BERT model is used to calculate the similarity between the context and the anchor matrix of the multi-source data at the current time. When the similarity is greater than the preset similarity threshold, the terms in the multi-source data at the current time are unambiguous. If the similarity is less than or equal to the preset similarity threshold, then according to the pre-screening rules, one of the ambiguous terms is selected as the normalized entity at the current moment.

[0012] The optimization strategy for the first threshold, the second threshold, and the algorithm selection using historical multi-source data includes: Using historical multi-source data as input and a first threshold, a second threshold, and an algorithm selection strategy as output, a reinforcement learning model is established. The increase in blank point adoption rate and the mean of dynamic quality score are used as reward signals. Through policy gradient or deep Q-learning algorithms, the reinforcement learning model is iteratively trained on historical multi-source data to obtain a trained reinforcement learning model. After inputting the current multi-source data into the trained reinforcement learning model, the optimized first threshold, second threshold, and algorithm selection strategy are obtained.

[0013] The knowledge graph algorithm includes a main algorithm, alternative algorithms, and a degradation algorithm. The main algorithm includes at least a clustering algorithm. The alternative algorithms include at least an association rule mining algorithm, a graph neural network, and a time series prediction model. The degradation algorithm includes at least a keyword matching algorithm.

[0014] Secondly, a dynamic knowledge graph circuit breaker optimization system based on multimodal data fusion is proposed to implement the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion described in the first aspect, including: The data acquisition module is used to acquire multi-source data at the current moment, evaluate the multi-source data input at the current moment using entanglement relationships, and obtain a dynamic quality score; The graph output module is used to semantically align the multi-source data at the current moment using a multilingual aligned anchor matrix when the dynamic quality score is greater than or equal to the first threshold, so as to obtain the normalized entities and relations at the current moment. The algorithm switching module is used to activate the dynamic knowledge graph circuit breaker mechanism and switch the knowledge graph algorithm by using an algorithm selection strategy when the blank point adoption rate is less than the second threshold or the dynamic quality score is less than the first threshold. The threshold update module is used to collect multi-source data after switching algorithms, proceed to step S1, and optimize the first threshold, the second threshold, and the algorithm selection strategy using historical multi-source data.

[0015] Thirdly, this application proposes an electronic device, including: one or more processors, and a memory for storing instructions, which, when executed by the one or more processors, cause the one or more processors to execute the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion.

[0016] Fourthly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the described dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion.

[0017] Beneficial effects: The dynamic knowledge graph circuit breaking optimization method and system based on multimodal data fusion disclosed in this invention can automatically switch the corresponding algorithm, and the switching threshold can be updated in real time, ensuring the reliability of the knowledge graph generation results and eliminating the problem of reduced accuracy of knowledge graphs due to terminology conflicts. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention or the prior art will be briefly introduced below. Obviously, the 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.

[0019] Figure 1 This is a flowchart of the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion, as shown in an embodiment of the present invention. Figure 2 This is a schematic diagram of the dynamic knowledge graph circuit breaker optimization system based on multimodal data fusion, as shown in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0021] Existing technologies suffer from several problems, such as blank spot discovery failure: when data quality fluctuates (e.g., blank spot adoption rate <5%), fixed clustering algorithms fail to identify valid technological opportunities, resulting in an invalid suggestion rate exceeding 90%. Multimodal data conflicts: differences in terminology and timeliness among patents, papers, and market data lead to semantic ambiguity, requiring manual intervention for resolution, which is inefficient. Redundant algorithm computation: low-quality data still triggers full-process computation, resulting in significant resource waste.

[0022] This application proposes a dynamic knowledge graph circuit breaker optimization method and system based on multimodal data fusion, which can automatically switch algorithms, eliminate ambiguity in terminology, and reduce resource waste without triggering full-process calculation.

[0023] Example 1: This embodiment proposes a dynamic knowledge graph circuit breaking optimization method based on multimodal data fusion, such as... Figure 1 As shown, it includes: Step S1: Obtain multi-source data at the current moment, evaluate the multi-source data input at the current moment using entanglement relationship, and obtain a dynamic quality score; In this embodiment, multi-source data at time t is input into the quantum heuristic data quality dynamic evaluation module to generate a dynamic quality score.

[0024] Among them, the multi-source data comes from the verification environment built on nine-dimensional data sources and proprietary data logic (patent database, academic papers, market reports, upstream and downstream supply chains, judicial precedents, corporate information, financing database, technical standard reports, industry and government reports).

[0025] The entanglement relationship is used to evaluate the multi-source data input at the current moment to obtain a dynamic quality score, calculated as follows: Q score (t)=Q initial e λt +Q event ; Among them, Q score (t) represents the dynamic quality score at time t, Q initial The dynamic quality score at the initial time, λ is the weighting coefficient, and ΔQ event Let e ​​be the quantum superposition state of the event, i.e., the entanglement relationship of the multi-source data at the current moment, and let e be the natural logarithm.

[0026] In this embodiment, the quantum-heuristic data quality dynamic evaluation module simulates the data quality dimensions (completeness, authority, timeliness) as a quantum superposition state, and quantifies the collaborative quality of multi-source data through entanglement. The entanglement can be understood as the correlation between multiple events. This embodiment uses quantum-like probability to describe the relationship between multiple events. The correlation between multiple events can be implemented using various existing technologies, which will not be elaborated here.

[0027] Step S2: When the dynamic quality score is greater than or equal to the first threshold, the multi-source data at the current moment is semantically aligned using a multilingual alignment anchor matrix to obtain the normalized entities and relations at the current moment. The multilingual alignment anchor matrix, based on a technical terminology database, is constructed by including: Collect core terms in a specific field and their standard expressions in different languages ​​to form a technical terminology database; A multilingual pre-trained model is used to convert each term into a high-dimensional semantic vector; Align and merge the high-dimensional semantic vectors of the same term in different languages ​​to form term-vector pairs; A multilingual alignment anchor matrix is ​​constructed using term-vector pairs, where terms are row vectors of the multilingual alignment anchor matrix.

[0028] In this embodiment, firstly, a terminology database is constructed: core terms in a specific domain (such as "solid-state battery") and their standard expressions in different languages ​​(Chinese, English, Japanese, etc.) are collected to form a structured terminology database. Secondly, vectorization is performed: each term (and its contextual definition) is converted into a high-dimensional semantic vector using a multilingual pre-trained model (such as XLM-RoBERTa, a cross-lingual version of BERT). Thirdly, matrix formation is performed: the semantic vectors of the same term in different languages ​​are aligned and merged (e.g., by calculating the average vector or establishing a mapping relationship) to form "term-vector" pairs. Finally, the semantic vectors of all core concepts are used as rows to construct an "anchor matrix." This matrix serves as a benchmark for semantic alignment.

[0029] The anchor point matrix, which employs multilingual alignment, performs semantic alignment of the multi-source data at the current moment to obtain the normalized entities and relations at the current moment, including: Step S2.1: Use the BERT model to calculate the similarity between the context of the multi-source data and the anchor matrix at the current time. Step S2.2: When the similarity is greater than the preset similarity threshold, the terms in the multi-source data at the current time are unambiguous; Step S2.3: When the similarity is less than or equal to the preset similarity threshold, one of the ambiguous terms is selected as the normalized entity at the current moment according to the pre-screening rules.

[0030] In this embodiment, firstly, a multilingual aligned anchor matrix is ​​constructed based on a specific technical terminology library (e.g., "sulfide electrolyte"). Secondly, context-aware disambiguation is used, employing a BERT model to calculate the similarity between the new data context and the anchor points, resolving terminological ambiguities (e.g., "Toyota" refers to an automaker in a patent but a research institution in a paper). Finally, a conflict resolution strategy library with preset rules (e.g., patent dates take precedence over paper dates) automatically resolves terminological ambiguities in the data.

[0031] Step S3: When the blank point adoption rate is less than the second threshold or the dynamic quality score is less than the first threshold, the dynamic knowledge graph circuit breaker mechanism is activated, and the algorithm selection strategy is used to switch the knowledge graph algorithm. The knowledge graph algorithm includes a main algorithm, alternative algorithms, and a degradation algorithm. The main algorithm includes at least a clustering algorithm. The alternative algorithms include at least an association rule mining algorithm, a graph neural network, and a time series prediction model. The degradation algorithm includes at least a keyword matching algorithm.

[0032] In this embodiment, the circuit breaker condition is as follows: when the blank point acceptance rate is less than 5% or the data quality score is lower than the threshold, the algorithm is switched. The switching logic is as follows: the main algorithm (clustering algorithm) switches to the alternative algorithm (association rule mining), and then to the downgraded algorithm (keyword matching). Real-time feedback: After the circuit breaker is triggered, the system collects data on the effect of the new algorithm and adjusts the switching threshold through reinforcement learning.

[0033] Step S4: Collect multi-source data after switching algorithms, proceed to step S1, and use historical multi-source data to optimize the first threshold, the second threshold, and the algorithm selection strategy.

[0034] The optimization strategy for the first threshold, the second threshold, and the algorithm selection using historical multi-source data includes: Using historical multi-source data as input and a first threshold, a second threshold, and an algorithm selection strategy as output, a reinforcement learning model is established. The increase in blank point adoption rate and the mean of dynamic quality score are used as reward signals. Through policy gradient or deep Q-learning algorithms, the reinforcement learning model is iteratively trained on historical multi-source data to obtain a trained reinforcement learning model. After inputting the current multi-source data into the trained reinforcement learning model, the optimized first threshold, second threshold, and algorithm selection strategy are obtained.

[0035] In this embodiment, the optimization steps are as follows: A reinforcement learning model is established. The first threshold, the second threshold, and the algorithm selection strategy are treated as adjustable actions. After inputting historical multi-source data into the system, the increase in the blank point adoption rate and the average dynamic quality score are used as reward signals. Through policy gradient or Deep Q-Learning algorithms, the reinforcement learning model is iteratively trained on historical data to learn the optimal threshold and policy combination that maximizes long-term rewards (i.e., the graph generation effect).

[0036] This embodiment proposes a dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion, which achieves the following aspects: (1) Improved accuracy: Through dynamic alignment of multimodal data, the false alarm rate of blank point identification is reduced by 90% (compared to the static rules of existing technologies).

[0037] (2) Efficiency optimization: The circuit breaker mechanism avoids redundant calculation of low-quality data, and the processing speed is increased by more than 50%.

[0038] (3) Adaptive capability: Reinforcement learning enables the system to continuously optimize the algorithm selection strategy and adapt to the rapid evolution of technologies such as solid-state batteries.

[0039] Example 2: This embodiment proposes a dynamic knowledge graph circuit breaker optimization system based on multimodal data fusion, used to implement the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion described in Embodiment 1, such as... Figure 2 As shown, it includes: a data acquisition module, a map output module, an algorithm switching module, and a threshold update module. The data acquisition module is connected to both the map output module and the algorithm switching module, and the algorithm switching module is connected to the threshold update module.

[0040] The data acquisition module is used to acquire multi-source data at the current moment, evaluate the multi-source data input at the current moment using entanglement relationships, and obtain a dynamic quality score; The graph output module is used to semantically align the multi-source data at the current moment using a multilingual aligned anchor matrix when the dynamic quality score is greater than or equal to the first threshold, so as to obtain the normalized entities and relations at the current moment. The algorithm switching module is used to activate the dynamic knowledge graph circuit breaker mechanism and switch the knowledge graph algorithm by using an algorithm selection strategy when the blank point adoption rate is less than the second threshold or the dynamic quality score is less than the first threshold. The threshold update module is used to collect multi-source data after switching algorithms, proceed to step S1, and optimize the first threshold, the second threshold, and the algorithm selection strategy using historical multi-source data.

[0041] Example 3: This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion.

[0042] The electronic device may be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0043] The processor is used to execute all or part of the steps in the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in the electronic device, as well as application-related data.

[0044] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the dynamic knowledge graph circuit breaking optimization method based on multimodal data fusion described in the above embodiments.

[0045] Example 4: This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0046] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion described in the various embodiments of this application.

[0047] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) app stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion described above.

[0048] The above-described 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 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, and should all be included within the protection scope of the present invention.

[0049] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention.

Claims

1. A dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion, characterized in that, include: Step S1: Obtain multi-source data at the current moment, evaluate the multi-source data input at the current moment using entanglement relationship, and obtain a dynamic quality score; Step S2: When the dynamic quality score is greater than or equal to the first threshold, the multi-source data at the current moment is semantically aligned using a multilingual alignment anchor matrix to obtain the normalized entities and relations at the current moment. Step S3: When the blank point adoption rate is less than the second threshold or the dynamic quality score is less than the first threshold, the dynamic knowledge graph circuit breaker mechanism is activated, and the algorithm selection strategy is used to switch the knowledge graph algorithm. Step S4: Collect multi-source data after switching algorithms, proceed to step S1, and use historical multi-source data to optimize the first threshold, the second threshold, and the algorithm selection strategy.

2. The dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion according to claim 1, characterized in that, The entanglement relationship is used to evaluate the multi-source data input at the current moment to obtain a dynamic quality score, calculated as follows: Q score (t)=Q initial e λt +ΔQ event ; Among them, Q score (t) represents the dynamic quality score at time t, Q initial The dynamic quality score at the initial time, λ is the weighting coefficient, and ΔQ event Let e ​​be the quantum superposition state of the event, i.e., the entanglement relationship of the multi-source data at the current moment, and let e be the natural logarithm.

3. The dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion according to claim 2, characterized in that, The events mentioned include the completeness, authority, and timeliness of multi-source data at the current moment.

4. The dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion according to claim 1, characterized in that, The multilingual alignment anchor matrix, based on a technical terminology database, is constructed by including: Collect core terms in a specific field and their standard expressions in different languages ​​to form a technical terminology database; A multilingual pre-trained model is used to convert each term into a high-dimensional semantic vector; Align and merge the high-dimensional semantic vectors of the same term in different languages ​​to form term-vector pairs; A multilingual alignment anchor matrix is ​​constructed using term-vector pairs, where terms are row vectors of the multilingual alignment anchor matrix.

5. The dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion according to claim 1, characterized in that, The anchor point matrix, which employs multilingual alignment, performs semantic alignment of the multi-source data at the current moment to obtain the normalized entities and relations at the current moment, including: The BERT model is used to calculate the similarity between the context and the anchor matrix of the multi-source data at the current time. When the similarity is greater than the preset similarity threshold, the terms in the multi-source data at the current time are unambiguous. If the similarity is less than or equal to the preset similarity threshold, then according to the pre-screening rules, one of the ambiguous terms is selected as the normalized entity at the current moment.

6. The dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion according to claim 1, characterized in that, The optimization strategy for the first threshold, the second threshold, and the algorithm selection using historical multi-source data includes: Using historical multi-source data as input and a first threshold, a second threshold, and an algorithm selection strategy as output, a reinforcement learning model is established. The increase in blank point adoption rate and the mean of dynamic quality score are used as reward signals. Through policy gradient or deep Q-learning algorithms, the reinforcement learning model is iteratively trained on historical multi-source data to obtain a trained reinforcement learning model. After inputting the current multi-source data into the trained reinforcement learning model, the optimized first threshold, second threshold, and algorithm selection strategy are obtained.

7. The dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion according to claim 1, characterized in that, The knowledge graph algorithm includes a main algorithm, alternative algorithms, and a degradation algorithm. The main algorithm includes at least a clustering algorithm. The alternative algorithms include at least an association rule mining algorithm, a graph neural network, and a time series prediction model. The degradation algorithm includes at least a keyword matching algorithm.

8. A dynamic knowledge graph circuit breaker optimization system based on multimodal data fusion, comprising the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion as described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire multi-source data at the current moment, evaluate the multi-source data input at the current moment using entanglement relationships, and obtain a dynamic quality score; The graph output module is used to semantically align the multi-source data at the current moment using a multilingual aligned anchor matrix when the dynamic quality score is greater than or equal to the first threshold, so as to obtain the normalized entities and relations at the current moment. The algorithm switching module is used to activate the dynamic knowledge graph circuit breaker mechanism and switch the knowledge graph algorithm by using an algorithm selection strategy when the blank point adoption rate is less than the second threshold or the dynamic quality score is less than the first threshold. The threshold update module is used to collect multi-source data after the algorithm is switched, and to optimize the first threshold, the second threshold, and the algorithm selection strategy using historical multi-source data.

9. An electronic device, characterized in that, include: One or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed, cause the processor to perform the dynamic knowledge graph circuit breaker optimization method based on multimodal data fusion as described in any one of claims 1 to 7.

Images