Abnormality processing method, apparatus, device, and storage medium

Abstract

The application relates to an abnormality processing method, device and equipment and a storage medium. The method comprises the following steps: in the case that an abnormal event of a front end is listened to, attribute information of the abnormal event is extracted to obtain an attribute information sequence; according to the attribute information sequence, matching reasoning is carried out in a pre-established knowledge graph to obtain at least one initial matching result and confidence related data corresponding to the initial matching result; according to the confidence related data corresponding to each initial matching result, a target matching result is determined from all the initial matching results; and the target matching result is displayed to enable relevant personnel to process the abnormal event according to the display result. The method can avoid the problems of low efficiency, errors and time and labor consumption in the manual investigation mode, and fully adapts to the core demands of rapid iteration and efficient operation of the front-end business.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an exception handling method, apparatus, device, and storage medium. Background Technology

[0002] In front-end scenarios such as general web applications, back-end management systems, and mobile H5 applications, various abnormal events frequently occur during daily operation. These abnormal events not only directly disrupt the user experience and lead to user churn, but may also interrupt core business processes, causing potential economic losses and damage to brand reputation for the business.

[0003] To address the aforementioned front-end anomalies, developers need to spend a significant amount of time and effort checking the causes of each event and developing corresponding solutions. However, this manual troubleshooting method is prone to errors, severely impacting the normal iteration progress of the project and failing to meet the needs of rapid iteration and efficient operation and maintenance of front-end business. Summary of the Invention

[0004] This application provides an anomaly handling method, apparatus, device, and storage medium, which can effectively avoid the problems of low efficiency, error-proneness, and time-consuming and labor-intensive manual troubleshooting mode, and fully adapt to the core requirements of rapid iteration and efficient operation and maintenance of front-end business.

[0005] Firstly, this application provides an exception handling method, the method comprising: Upon detecting an abnormal event on the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence; Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result. Based on the confidence-related data corresponding to each initial matching result, the target matching result is determined among all the initial matching results; The results of the target matching are displayed so that relevant personnel can handle the abnormal events based on the displayed results.

[0006] Optionally, based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result, including: When the attribute information sequence satisfies the first preset condition, the attribute information in the attribute information sequence is matched and reasoned with the nodes in the knowledge graph in turn to obtain the initial matching result. The semantic similarity between each attribute information and the content of the corresponding matching node is the first similarity. Based on the preset first matching strategy, the first similarity, and the preset node matching level, the confidence-related data corresponding to the initial matching result is determined.

[0007] Optionally, based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result, including: When the attribute information sequence satisfies the second preset condition, matching reasoning is performed in the knowledge graph based on the first attribute information in the attribute information sequence to obtain at least one initial matching result. The semantic similarity between the second attribute information in the attribute information sequence and the content of the corresponding matching node is greater than the second similarity, and the similarity between the other attribute information and the content of the corresponding matching node is the first similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset second matching strategy, the semantic similarity between the second attribute information and the corresponding matching node content, and the node matching level.

[0008] Optionally, based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result, including: When the attribute information sequence satisfies the third preset condition, the knowledge graph is subjected to reverse derivation and forward matching reasoning based on the third attribute information in the attribute information sequence to determine at least one initial matching result, wherein the semantic similarity between the third attribute information and the content of the corresponding matching node is greater than the second similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset third matching strategy, the semantic similarity between the third attribute information and the corresponding matching node content, and the node matching level.

[0009] Optionally, based on the confidence-related data corresponding to each initial matching result, the target matching result is determined from all initial matching results, including: For each initial matching result, the matching strategy, semantic similarity, and node matching level are obtained from the corresponding confidence-related data. The first weight corresponding to the matching strategy and the second weight corresponding to the node matching level are determined. Based on the first weight, the second weight, and the semantic similarity, the confidence level corresponding to the initial matching result is obtained. The target matching result is determined based on the initial matching results whose confidence level is greater than the confidence threshold.

[0010] Optionally, the results of the target matching can be displayed, including: Based on the confidence level corresponding to each target matching result, the target matching results are initially sorted to obtain a first sequence; For the target matching results with the same confidence level in the first sequence, a second sequence is obtained by sorting them a second time according to the node matching level. For target matching results with the same node matching level in the second sequence, sort them three times according to the semantic similarity to obtain the target sequence; The display is based on the target sequence.

[0011] Optionally, the method further includes: Retrieve target exception data related to console exceptions; Based on the target abnormal data, a reference triple is constructed, which includes the data corresponding to the current node, the data included in the previous level node, and the data included in the next level node. A knowledge graph is constructed based on the reference triples.

[0012] Secondly, this application provides an anomaly handling apparatus, the apparatus comprising: The extraction unit is used to extract the attribute information of the abnormal event when an abnormal event is detected at the front end, and obtain an attribute information sequence. The matching unit is used to perform matching reasoning in a pre-established knowledge graph based on the attribute information sequence to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result. The determination unit is used to determine the target matching result among all the initial matching results based on the confidence-related data corresponding to each initial matching result; The display unit is used to display the target matching results so that relevant personnel can handle the abnormal event based on the display results.

[0013] Optional, matching unit, used for: When the attribute information sequence satisfies the first preset condition, the attribute information in the attribute information sequence is matched and reasoned with the nodes in the knowledge graph in turn to obtain the initial matching result. The semantic similarity between each attribute information and the content of the corresponding matching node is the first similarity. Based on the preset first matching strategy, the first similarity, and the preset node matching level, the confidence-related data corresponding to the initial matching result is determined.

[0014] Optional, matching unit, used for: When the attribute information sequence satisfies the second preset condition, matching reasoning is performed in the knowledge graph based on the first attribute information in the attribute information sequence to obtain at least one initial matching result. The semantic similarity between the second attribute information in the attribute information sequence and the content of the corresponding matching node is greater than the second similarity, and the similarity between the other attribute information and the content of the corresponding matching node is the first similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset second matching strategy, the semantic similarity between the second attribute information and the corresponding matching node content, and the node matching level.

[0015] Optional, matching unit, used for: When the attribute information sequence satisfies the third preset condition, the knowledge graph is subjected to reverse derivation and forward matching reasoning based on the third attribute information in the attribute information sequence to determine at least one initial matching result, wherein the semantic similarity between the third attribute information and the content of the corresponding matching node is greater than the second similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset third matching strategy, the semantic similarity between the third attribute information and the corresponding matching node content, and the node matching level.

[0016] Optional, define the unit for: For each initial matching result, the matching strategy, semantic similarity, and node matching level are obtained from the corresponding confidence-related data. The first weight corresponding to the matching strategy and the second weight corresponding to the node matching level are determined. Based on the first weight, the second weight, and the semantic similarity, the confidence level corresponding to the initial matching result is obtained. The target matching result is determined based on the initial matching results whose confidence level is greater than the confidence threshold.

[0017] Optional, display unit, used for: Based on the confidence level corresponding to each target matching result, the target matching results are initially sorted to obtain a first sequence; For the target matching results with the same confidence level in the first sequence, a second sequence is obtained by sorting them a second time according to the node matching level. For target matching results with the same node matching level in the second sequence, sort them three times according to the semantic similarity to obtain the target sequence; The display is based on the target sequence.

[0018] Optionally, the apparatus further includes a building unit, the building unit being used for: Retrieve target exception data related to console exceptions; Based on the target abnormal data, a reference triple is constructed, which includes the data corresponding to the current node, the data included in the previous level node, and the data included in the next level node. A knowledge graph is constructed based on the reference triples.

[0019] Thirdly, this application provides an exception handling device, comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; and at least one memory connected to the at least one bus, wherein the processor is configured to: Upon detecting an abnormal event on the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence; Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result. Based on the confidence-related data corresponding to each initial matching result, the target matching result is determined among all the initial matching results; The results of the target matching are displayed so that relevant personnel can handle the abnormal events based on the displayed results.

[0020] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described exception handling method.

[0021] Compared with the prior art, the technical solution provided in this application has the following advantages: When an abnormal event is detected at the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence; based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result; based on the confidence-related data corresponding to each initial matching result, a target matching result is determined among all the initial matching results; the target matching result is displayed so that relevant personnel can handle the abnormal event according to the displayed result. This application can automatically extract the attribute information sequence and perform matching reasoning with the pre-built knowledge graph based on the attribute information sequence, automatically deriving key information such as the root cause of the abnormal event and the appropriate solution; relying on the above information to carry out targeted processing of the abnormal event effectively avoids the problems of low efficiency, error-proneness, and time-consuming and labor-intensive manual investigation mode, fully adapting to the core requirements of rapid iteration and efficient operation and maintenance of front-end business. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0025] Figure 1 A flowchart illustrating an exception handling method provided in an embodiment of this application; Figure 2 A flowchart illustrating a data determination method provided in an embodiment of this application; Figure 3 A flowchart illustrating a data determination method provided in an embodiment of this application; Figure 4 A flowchart illustrating a data determination method provided in an embodiment of this application; Figure 5 A flowchart illustrating another knowledge graph construction method provided in this application embodiment; Figure 6This is a flowchart illustrating an exception handling device provided in an embodiment of this application; Figure 7 This is a schematic diagram of an exception handling device provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] The following disclosure provides numerous different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of the invention. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0028] In front-end scenarios such as general web applications, backend management systems, and mobile H5, various abnormal events frequently occur during daily operation. These abnormal events not only directly disrupt the user experience and lead to user churn, but may also interrupt core business processes, causing potential economic losses and damage to brand reputation for the business. To address these front-end abnormal events, developers need to spend a significant amount of time and effort to investigate their causes one by one and develop corresponding solutions. However, this manual troubleshooting method is prone to errors, severely impacting the normal iteration progress of the project and failing to meet the needs of rapid iteration and efficient operation and maintenance of front-end business.

[0029] In summary, this application provides an anomaly handling method. This method can automatically extract attribute information sequences and perform matching and reasoning based on these sequences with a pre-built knowledge graph to automatically deduce key information such as the root cause of anomalies and suitable solutions. By relying on this information to carry out targeted processing of anomalies, this method effectively avoids the problems of low efficiency, error-proneness, and time-consuming nature of manual investigation, fully meeting the core demands of rapid iteration and efficient operation and maintenance of front-end businesses. Figure 1 As shown, the specific steps include: Step 101: When an abnormal event is detected on the front end, extract the attribute information of the abnormal event to obtain the attribute information sequence.

[0030] The attributes include exception type, exception characteristics, exception scenario, and front-end framework. Exception type refers to the exception category of the event, exception characteristics are the key details of the event, exception scenario is the context of the event, and front-end framework is the runtime environment of the event. The specific information corresponding to these attributes is called attribute information, namely, exception type information, exception characteristic information, exception scenario information, and front-end framework information. Each attribute corresponds to multiple types of attribute information. For example, exception type information includes syntax errors (SyntaxError), type errors (TypeError), reference errors (ReferenceError), resource loading errors (ResourceError), and network errors (NetworkError). Syntax errors are code syntax that does not conform to the ECMAScript specification, such as missing semicolons, mismatched parentheses, and misspelled keywords. These errors are triggered during the compilation phase and have clear syntactic identifiers. Type errors are operations performed on variables that do not match their types, such as calling methods on undefined / null variables or passing incorrect type parameters to functions; these are the most frequent runtime error types. Reference errors occur when referencing undeclared variables or freed memory addresses, such as using undefined variables or accessing properties of deleted objects, often related to improper scope management. Resource loading errors occur when static resources such as HTML / CSS / JS or API dependencies fail to load, such as 404 status codes (resource not found), 503 status codes (service unavailable), or resource blocking due to cross-domain restrictions. Network errors occur when there are abnormalities in the communication link between the front-end and back-end, such as request timeouts, TCP connection failures, or invalid HTTPS certificates. The key difference between network errors and resource loading errors is that the "communication link is interrupted" rather than "there is a problem with the resource itself."

[0031] An attribute information sequence refers to a sequence of attribute information arranged in a preset order, which is directly related to the matching logical order in the knowledge graph. For example, if the knowledge graph adopts a hierarchical matching order of "abnormal type → abnormal feature → abnormal scenario → front-end framework", and all attribute information (all attributes being abnormal type, abnormal feature, abnormal scenario, and front-end framework) is extracted from the abnormal event, then the corresponding attribute information sequence is "abnormal type information, abnormal feature information, abnormal scenario information, and front-end framework information".

[0032] In this step, the exception handling system monitors the front-end's running status in real time to determine if any exception events have occurred. Once an exception event is captured, the system automatically extracts the event's attribute information, fills these attributes into a structured template, and generates standardized structured statements, which are the attribute information sequences.

[0033] For example, the exception type information extracted from the exception event is SyntaxError, the exception feature information is Cannot read property 'x' of undefined (the key error text enclosed in quotes), the exception scenario information is when the Vue component's mounted hook is executed (the lifecycle context of the Vue framework), the front-end framework information is Vue 3, and the structured template is "Exception type information is XXX, exception feature information is XXX, exception scenario information is XXX, front-end framework information is XXX". Filling the extracted information into the structured template yields the attribute information sequence "Exception type information is SyntaxError, exception feature information is Cannot read property 'x' of undefined, exception scenario information is when the Vue component's mounted hook is executed, front-end framework information is Vue 3".

[0034] Step 102: Based on the attribute information sequence, perform matching in the pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to each initial matching result.

[0035] A knowledge graph is a structured semantic knowledge base used to search for solutions. It comprises nodes with various attributes: exception type nodes, exception feature nodes, exception scenario nodes, front-end framework nodes, and solution nodes. Each node corresponds to specific content; for example, the exception type node corresponds to the content "TypeError". These nodes also store relationships. For instance, the parent node of an exception feature node is an exception type node, and its next-level node is an exception feature node; similarly, the parent node of a front-end framework node is an exception scenario node, and its next-level node is a solution node.

[0036] The initial matching results include attributes such as exception type, exception characteristics, exception scenario, front-end framework, and technical solution, along with the corresponding information. The technical solution information is derived through matching and reasoning within the knowledge graph based on the extracted attribute information. For example, the initial matching results might be: exception type is TypeError, exception characteristic is "Can not read property of undefined", exception scenario is Vue componentmethod, front-end framework is Vue 3, and solution is "Check if variable is defined before accessing its property".

[0037] The confidence-related data refers to the data used to calculate the confidence of the matching results. Specifically, it includes: the semantic similarity between attribute information and the corresponding matching node attributes; the node matching level determined by the number of successfully matched nodes (e.g., 4-node / 3-node / 2-node); and the matching strategy type (e.g., exact matching, fuzzy matching, reverse matching). These data quantify the matching quality between attribute information and knowledge graph matching nodes from different dimensions and are key parameters for calculating confidence.

[0038] In this step, the matching strategy to be used is determined based on the attribute information sequence. For example, when the attribute information sequence includes exception type information, exception feature information, exception scenario information, and front-end framework information, a preset precise matching strategy and a fuzzy matching strategy can be used. When the attribute information sequence only includes exception type information, exception feature information, and exception scenario information, or when the attribute information sequence only includes exception type information and exception feature information, a fuzzy matching strategy can be used. When the attribute information sequence does not include exception type information but includes exception feature information, a reverse matching strategy can be used.

[0039] The precise matching strategy requires that all attribute information in the abnormal attribute sequence completely match the content of the corresponding matching node in the knowledge graph, meaning that the semantic similarity between all attribute information and the corresponding node content must reach 1 (or 100%). The fuzzy matching strategy can match based on all or part of the attribute information in the abnormal attribute sequence; where part of the attribute information includes, but is not limited to, the following combinations: abnormal type information + abnormal feature information + abnormal scenario information, and abnormal type information + abnormal feature information. This strategy requires that the semantic similarity between the abnormal type information and the corresponding matching node content must be greater than a preset similarity threshold (e.g., the preset similarity threshold is set to 0.65), and the semantic similarity between the remaining attribute information and the corresponding matching node content must reach 100%. The reverse matching strategy, assuming the abnormal attribute sequence contains abnormal feature information, first deduces the corresponding abnormal type based on the abnormal feature information and the knowledge graph, and then uses the above fuzzy matching strategy to find the target node.

[0040] After determining the matching strategy to be used, the attribute information in the attribute information sequence can be matched and reasoned with the nodes in the knowledge graph according to the matching strategy to obtain at least one initial matching result and confidence-related data corresponding to each initial matching result.

[0041] Step 103: Determine the target matching result based on each initial matching result and the corresponding confidence-related data.

[0042] In this step, you can either retain the matching results with the confidence level data set as preset data from the initial matching results and determine these matching results as the matching results to be used, or you can determine the corresponding confidence level for each initial matching result based on its corresponding confidence level data. If its confidence level is less than the confidence level threshold, it will be discarded; if its confidence level is greater than or equal to the confidence level threshold, it will be determined as the matching result to be used.

[0043] After obtaining the matching results to be used, for each matching result to be used, the content, confidence score, and confidence score-related data of the matching result are fused to obtain the target matching result corresponding to the matching result.

[0044] Step 104: Display the target matching results so that relevant personnel can handle the abnormal events based on the displayed results.

[0045] In this step, the target matching results can be displayed directly, or they can be sorted according to a preset method to obtain a sequence of matching results, which can then be displayed. For example, the target matching results can be sorted according to the confidence level corresponding to each target matching result to obtain a sequence of matching results. Other sorting methods can also be used, and this is not limited here.

[0046] In this embodiment, upon detecting an abnormal event at the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence. Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result. Based on the confidence-related data corresponding to each initial matching result, a target matching result is determined among all the initial matching results. The target matching result is then displayed so that relevant personnel can handle the abnormal event based on the displayed result. This application can automatically extract the attribute information sequence and perform matching reasoning with the pre-built knowledge graph based on the attribute information sequence, automatically deriving key information such as the root cause of the abnormal event and the appropriate solution. Targeted handling of abnormal events based on the above information effectively avoids the problems of low efficiency, error-proneness, and time-consuming labor in the manual investigation mode, fully adapting to the core requirements of rapid iteration and efficient operation and maintenance of front-end business.

[0047] In this embodiment, it is first determined whether the attribute information sequence meets the applicable conditions of the precise matching strategy. If it does, the strategy is enabled, and a matching operation is performed based on the attribute information sequence and a preset front-end anomaly knowledge graph to generate initial matching results and associated confidence-related data. Therefore, this embodiment provides a data determination method, such as... Figure 2 As shown, the specific steps include: Step 201: If the attribute information sequence meets the first preset condition, the attribute information in the attribute information sequence is matched and reasoned with the nodes in the knowledge graph in turn to obtain the initial matching result.

[0048] The first preset condition is that the attribute information sequence includes anomaly type information, anomaly feature information, anomaly scenario information, and front-end framework information. The similarity between each attribute and its corresponding matching node attribute is the first similarity. A first similarity of 1 means that the attribute and its corresponding matching node are exactly matched. The first matching strategy is a precise matching strategy, which requires all attribute information to be precisely matched with their corresponding matching nodes. Precise matching means that the semantic similarity between all attribute information and the content of their corresponding matching nodes reaches 1.

[0049] In this step, when the attribute information sequence is detected to meet the first preset condition, it indicates that the exact matching strategy can be used to perform node matching. At this time, the anomaly type information in the attribute information sequence is extracted first, and its semantic similarity with the node content of each feature type node is calculated; if the semantic similarity between the anomaly type information and the node content of a certain feature type node reaches 1 (i.e., exact matching), then the node is determined as the matching node for the anomaly type information.

[0050] Starting with the matching node of this anomaly type, search for the next-level node associated with it in the knowledge graph, and calculate the semantic similarity between the node content of these nodes and the anomaly feature information in the attribute information sequence; when the semantic similarity between the node content of a node and the anomaly feature information is 1 (i.e., exact match), the node is determined as the matching node of the anomaly feature information.

[0051] Starting with the matching node of the abnormal feature information, continue to search for the next level node associated with it in the knowledge graph, and calculate the semantic similarity between the node content of these nodes and the abnormal scene information; if the semantic similarity between the node content of a certain node and the abnormal scene information reaches 1 (i.e., exact match), then the node is determined as the matching node of the abnormal scene information.

[0052] Starting with the matching node of the abnormal scene information, search for the next-level node associated with it in the knowledge graph, and calculate the semantic similarity between the node content of these nodes and the front-end framework information; when the semantic similarity between the node content of a certain node and the front-end framework information is 1 (i.e., exact match), the node is determined as the matching node of the front-end framework information.

[0053] Subsequently, based on the matching node of this front-end framework information, the next-level node associated with it is searched in the knowledge graph. The content of this node is the solution information. Finally, by combining the exception type information, exception feature information, exception scenario information, front-end framework information, and corresponding solution information, an initial matching result is generated.

[0054] Step 202: Determine the confidence-related data corresponding to the initial matching result based on the preset first matching strategy, the first similarity, and the preset node matching level.

[0055] After obtaining the initial matching result, the confidence-related data corresponding to this initial matching result is determined. Specific steps include: the node matching level is directly related to the number of successfully matched nodes, which essentially corresponds to the number of attribute information items participating in the matching in the attribute information sequence. Exact matching, as the highest priority matching strategy, requires all four types of attribute information in the attribute information sequence (anomaly type information, anomaly feature information, anomaly scenario information, and front-end framework information) to achieve complete matching. Therefore, the preset node matching level corresponding to exact matching is 4-node. Then, based on the preset first matching strategy, the first similarity, and the preset node matching level, the confidence-related data corresponding to the initial matching result is determined.

[0056] In this embodiment, it is first determined whether the attribute information sequence meets the applicable conditions of the fuzzy matching strategy. If it does, the strategy is activated, and a matching inference operation is performed based on the attribute information sequence and a preset front-end anomaly knowledge graph to generate an initial matching result and related confidence-related data. Therefore, this embodiment provides a data determination method, such as... Figure 3 As shown, the specific steps include: Step 301: If the attribute information sequence meets the second preset condition, perform matching reasoning in the knowledge graph based on the first attribute information in the attribute information sequence to obtain at least one initial matching result.

[0057] The first attribute information includes anomaly type information and anomaly feature information. The semantic similarity between the second attribute information in the attribute information sequence and the content of the corresponding matching node is greater than the second similarity; the similarity between other attribute information and the content of the corresponding matching node is the first similarity. The second attribute information is anomaly type information, and the second similarity is a similarity threshold set by technical personnel based on experience, which can be set to 0.65.

[0058] In this step, since anomaly type information and anomaly feature information are the core anchor points for front-end anomaly root cause localization (the former is used for coarse screening of error categories, and the latter is the key basis for semantic matching), the attribute information sequence must contain these two types of attribute information when executing the fuzzy matching strategy. Subsequently, based on the anomaly type and anomaly features, matching reasoning will be carried out in the preset knowledge graph to generate multiple initial matching results. The specific steps are as follows: The semantic similarity between the anomaly type information in the attribute information sequence and the node content of each anomaly feature node is calculated; if the semantic similarity between the anomaly type information and the node content of an anomaly feature node is greater than or equal to 0.65, then that node is determined as the matching node for the anomaly type information. Starting from this matching node for the anomaly type information, the next-level nodes associated with it are determined, and the semantic similarity between the node content of these nodes and the anomaly feature information in the attribute information sequence is calculated; if the semantic similarity between the node content of a node and the anomaly feature information is 1, then that node is determined as the matching node for the anomaly feature. Based on the matched anomaly type nodes, anomaly feature nodes, and corresponding anomaly scenario nodes, front-end framework nodes, and solution nodes in the knowledge graph, an initial matching result is finally formed, which includes root cause, anomaly feature information, scenario information, framework information, and solution information.

[0059] Step 302: For each initial matching result, determine the number of attribute information matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset second matching strategy, the semantic similarity between the second attribute information and the corresponding matching node content, and the node matching level.

[0060] In this step, since the attribute information sequence may also include abnormal scenario information and / or front-end framework information, for each initial matching result, the number of attributes matched between the initial matching result and the attribute information sequence can also be determined. For example, when an initial matching result is obtained based on matching abnormal type information and abnormal feature information, the number of attribute information matched between the initial matching result and the attribute information sequence is 2; when an initial matching result is obtained based on matching abnormal type information, abnormal feature information, and abnormal scenario information, the number of attribute information matched between the initial matching result and the attribute information sequence is 3; when an initial matching result is obtained based on matching abnormal type information, abnormal feature information, abnormal scenario information, and front-end framework information, the number of attribute information matched between the initial matching result and the attribute information sequence is 4.

[0061] Next, the node matching level is determined based on the number. For example, when the number is 2, the corresponding node matching level is 2-node; when the number is 3, the corresponding node matching level is 3-node; and when the number is 4, the corresponding node matching level is 4-node. Finally, based on the preset second matching strategy, the semantic similarity between the second attribute information and the content of the corresponding matching node, and the node matching level, the confidence-related data corresponding to the initial matching result can be determined. For example, if the second matching strategy corresponding to a certain initial matching result is a fuzzy matching strategy, the semantic similarity between the second attribute information and the content of the corresponding matching node is 0.8, and the node matching level is 3-node, then the corresponding confidence-related data are the fuzzy matching strategy, 0.8, and 3-node.

[0062] In this embodiment, it is first determined whether the attribute information sequence meets the applicable conditions of the reverse matching strategy. If it does, the strategy is activated, and a matching inference operation is performed based on the attribute information sequence and a preset knowledge graph to generate an initial matching result and related confidence-related data. Therefore, this embodiment provides a data determination method, such as... Figure 4 As shown, the specific steps include: Step 401: If the attribute information sequence meets the third preset condition, perform reverse deduction in the knowledge graph based on the third attribute information in the attribute information sequence to determine at least one initial matching result.

[0063] The third attribute information is anomaly feature information, and the semantic similarity between the third attribute information and the content of the corresponding matching node is greater than the second similarity. The third preset condition is that the attribute information sequence does not include anomaly type information but includes anomaly feature information.

[0064] In this step, if the attribute information sequence meets the third preset condition, a reverse matching strategy is required. Therefore, the semantic similarity between the abnormal feature information and the node content of each abnormal feature node in the knowledge graph is first calculated. When the semantic similarity between the abnormal feature information and the node content of a certain common feature node is greater than the second similarity, that node is identified as the matching node of the abnormal feature. The parent node of this matching node in the knowledge graph is determined, and these nodes are identified as the nodes corresponding to the abnormal type. Based on the matching nodes of the abnormal type, the matching nodes of the abnormal features, and further association with the corresponding abnormal scenario nodes, front-end framework nodes, and solution nodes in the knowledge graph, an initial matching result containing root cause, abnormal feature, scenario, framework, and solution is finally formed.

[0065] Step 402: For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset third matching strategy, the semantic similarity between the third attribute information and the corresponding matching node content, and the node matching level.

[0066] This step is similar to step 302, and will not be described in detail here.

[0067] In this embodiment, for each initial matching result, the confidence level corresponding to the initial matching result is determined based on the confidence level-related data. Then, initial matching results with confidence levels greater than a confidence threshold are selected based on these confidence levels to determine the target matching result. Therefore, this embodiment provides a method for determining a target matching result, with specific steps including: for each initial matching result, obtaining the matching strategy, semantic similarity, and node matching level from the corresponding confidence level-related data; determining the first weight corresponding to the matching strategy and the second weight corresponding to the node matching level; obtaining the confidence level corresponding to the initial matching result based on the first weight, the second weight, and the semantic similarity; and determining the target matching result based on the initial matching results with confidence levels greater than a confidence threshold.

[0068] In this system, the weight corresponding to the first matching strategy is greater than the weight corresponding to the second matching strategy, the weight corresponding to the second matching strategy is greater than the weight corresponding to the third matching strategy, that is, the weight corresponding to the exact matching strategy is greater than the weight corresponding to the fuzzy matching strategy, and the weight corresponding to the fuzzy matching strategy is greater than the weight corresponding to the reverse matching strategy. For example, the weight corresponding to the exact matching strategy is 1, the weight corresponding to the fuzzy matching strategy is 0.8, and the weight corresponding to the reverse matching strategy is 0.6. The higher the node matching level, the more matching nodes it uses, and the greater its corresponding weight. For example, the weight corresponding to the 4-node node matching level is greater than the weight corresponding to the 3-node node matching level, and the weight corresponding to the 3-node node matching level is greater than the weight corresponding to the 2-node node matching level. For example, the weight corresponding to the 4-node node matching level is 1, the weight corresponding to the 3-node node matching level is 0.7, and the weight corresponding to the 3-node node matching level is 0.4. The confidence threshold is set by technical personnel based on experience; for example, the confidence threshold is 0.4.

[0069] In this step, for each initial matching result, the matching strategy, semantic similarity, and node matching level are obtained from the corresponding confidence-related data. The first weight corresponding to the matching strategy and the second weight corresponding to the node matching level are determined. The first weight, the second weight, and the semantic similarity are multiplied together to obtain the confidence level corresponding to the initial matching result.

[0070] For example, confidence score = first weight × second weight × semantic similarity.

[0071] Then, based on the initial matching results with a confidence level greater than the confidence threshold, the target matching result is determined. The specific steps are as follows: Initial matching results with a confidence level greater than the confidence threshold are identified as the matching results to be used. For each matching result to be used, the matching result, the corresponding confidence-related data, and the confidence level are fused to obtain the target matching result corresponding to that matching result. That is, the target matching result includes data such as anomaly type information, anomaly feature information, anomaly scenario information, front-end framework information, matching strategy type, node matching level, semantic similarity, and confidence level.

[0072] In this embodiment, the matching results are first sorted according to confidence level, then sorted according to node matching level for matching results with the same confidence level, and finally sorted according to semantic similarity for matching results with the same node matching level to obtain a target sequence. Finally, the target sequence is used for display. The specific steps are as follows: Based on the confidence level corresponding to each target matching result, the target matching results are initially sorted to obtain a first sequence; for target matching results with the same confidence level in the first sequence, a second sort is performed according to node matching level to obtain a second sequence; for target matching results with the same node matching level in the second sequence, a third sort is performed according to semantic similarity to obtain a target sequence; and the target sequence is used for display.

[0073] In this step, the target matching results are initially sorted in descending order based on the confidence level corresponding to each target matching result to obtain the first sequence; then, for target matching results with the same confidence level, they are sorted in descending order again according to the priority of 4-node > 3-node > 2-node to obtain the second sequence; for target matching results with the same node matching level, they are sorted in descending order according to semantic similarity (the closer the similarity is to 1.0, the higher the ranking) to obtain the target sequence.

[0074] In this embodiment, target anomaly data can also be extracted from console errors in front-end debugging scenarios, and triples can be constructed using this target anomaly data, thereby building a knowledge graph based on the triples. Therefore, this embodiment provides a method for constructing a knowledge graph, which specifically includes: Step 501: Obtain target exception data related to front-end exceptions.

[0075] In this step, the knowledge graph is built based on the static collection and processing of various unstructured data sources. The foundational data for the knowledge graph primarily comes from four types of data sources: front-end technical documentation, Stack Overflow front-end exception Q&A, exception descriptions in GitHub Issues, and internal enterprise development fault logs. Front-end technical documentation includes official error manuals and frequently asked questions (FAQs) for mainstream front-end frameworks such as Vue and React, providing framework-specific exception descriptions and solutions. Stack Overflow front-end exception Q&A extracts exception questions, frequently asked answers, and solutions related to front-end development from the Stack Overflow community, covering various front-end frameworks and common error scenarios. GitHub Issues extract actual exception questions, discussions, and solutions encountered by developers from Issues in GitHub open-source projects, including bug reports and solutions related to front-end frameworks. Internal enterprise development fault logs (hereinafter referred to as enterprise documentation) include fault records accumulated during internal development, stored in TXT, Markdown, or HTML formats, containing exception descriptions, root cause analyses, and solutions.

[0076] The above data acquisition method involves one-time static data collection, requiring no continuous operation or real-time data updates. After obtaining the above data sources through legal and compliant channels, the raw data is stored in a structured format (CSV, JSON, or database) to provide the foundation for subsequent data processing.

[0077] After obtaining the extracted data, it is necessary to preprocess the data to ensure its quality. The specific steps are as follows: clean and standardize the obtained raw data, including removing HTML tags, special characters and redundant information, unifying the representation of exception types and features, ensuring data consistency and usability, and providing a high-quality data foundation for building a front-end exception knowledge graph.

[0078] After obtaining the cleaned data, inductive analysis is performed on massive amounts of unstructured anomaly data to determine its corresponding attributes. This application abstracts five core attributes affecting root cause localization: 1. Anomaly Type: Major error categories defined at the JavaScript engine and browser underlying mechanism level (e.g., SyntaxError, TypeError), used for the first level of coarse screening. 2. Anomaly Characteristics: Specific, matchable key text in the error message (e.g., "Cannot read property 'x' of undefined"), which are the core anchors directly matching user input. 3. Anomaly Context: The business or technical context in which the error occurred (e.g., "when Vue components are mounted," "after asynchronous React state updates"), used for further precise localization under the same framework and error type. 4. Frontend Framework: The runtime environment of the error; different frameworks have different response mechanisms and best practices, which is a key dimension for distinguishing root causes. 5. Solution: The final target output, closely bound to the above four attributes. Furthermore, statistical analysis of the collected data revealed the top five most frequent error types (accounting for a cumulative 92.3%), which were: SyntaxError (21.5%), TypeError (35.2%), ReferenceError (18.7%), ResourceError (10.9%), and NetworkError (6.0%).

[0079] Next, for each attribute, corresponding information is extracted from the cleaned data. This application uses pre-built rules to achieve information extraction. Specifically: 1. Exception Type: Construct a bilingual core dictionary (["SyntaxError", "Syntax Error", "TypeError", "Type Error", ...]), which is extracted through precise string matching and regular expressions ( / \b(Syntax|Type|Reference|Resource|Network)Error\b / ) to ensure coverage of standardized Chinese and English names.

[0080] 2. ExceptionFeature: Utilize the "quote-wrapped" feature of error messages to extract the text within quotation marks using the regular expression / ["'“”]([^"'“”]+)["'“”] / ; for stack traces, extract relevant phrases by matching keywords ("error", "failed", "undefined").

[0081] 3. ExceptionScenario: Design a "framework + operation" template (e.g., / (Vue|React|component).*?(lifecycle|state update) / ) and extract the context description of the exception occurrence.

[0082] 4. Front-end framework: A dictionary is built based on mainstream frameworks and versions (["Vue 3", "React18", "Angular 14"]). It is extracted by exact matching to avoid misidentification of non-front-end frameworks.

[0083] 5. Solution: Design an "action + goal" semantic template (e.g., / (Check|Repair|Check|Verify).*?(variable|path|semicolon|variable|path) / ) to extract phrases containing the repair action.

[0084] After extracting the above information, it is also necessary to extract the relationship information between the attribute information. This application uses a hybrid relationship extraction strategy of preset rule templates and pre-trained language models to extract the relationship between attribute information from the text. For example, the rule template for extracting the relationship between anomaly type and solution is / (?: solution|method|Fix|Solution):\s*(.*?)(?=\n|$) / .

[0085] To address complex anomaly descriptions that cannot be covered by preset rule templates, a pre-trained language model based on the Transformer architecture, BERT, is introduced for relation extraction. The model combines complete text fragments containing attribute information pairs with the attribute information using delimiters to form an input sequence, which is then fed into the model to obtain the included relations. This allows the model to simultaneously consider local entity information and global textual context.

[0086] Based on the data collected above, this application abstracts five types of nodes, and the descriptions of these nodes are shown in Table 1: Table 1

[0087] Step 502: Construct reference triples based on the target anomaly data.

[0088] The reference triple includes the data corresponding to the current node, the data included in the parent node, and the data included in the next node. See Table 2 for details. Table 2

[0089] Step 503: Construct a knowledge graph based on the reference triples.

[0090] In this step, the reference triples may store triples with related relationships, such as the triples (TypeError, has_solution, Verify variable initialization) and (TypeError, has_solution, Check if variable is defined). These have the same meaning but different expressions. For such triples, only the more important ones need to be retained. Therefore, in the reference triples, triples with the same relationship are found and combined to obtain a set of triples. For each set of triples, generally, the same expression is used for the same data source. Therefore, the confidence level of the data source to which each triple belongs can be determined, and the triple corresponding to the data source with the highest confidence level is taken as the final target triple. Then, a knowledge graph is constructed based on the data in the target triples.

[0091] The specific steps for determining the confidence level for each data source are as follows: Count the frequency of occurrence of the specified triplet in each data source; multiply the frequency of each data source by its preset weight to obtain the product result for each data source; sum the product results of all data sources to obtain the total product value; and determine the confidence level for that data source as the ratio of the product result of each data source to the total product value.

[0092] For example, data sources include Stack Overflow, GitHub Issues, and enterprise documentation. The default weights for Stack Overflow are 0.9, for GitHub Issues 0.8, and for enterprise documentation 0.7. The triple (TypeError, has_solution, Check if variable is defined) appears 5 times on Stack Overflow, 2 times on GitHub Issues, and 1 time in the enterprise documentation.

[0093] Taking the confidence calculation of (TypeError, has_solution, Check if variable is defined) on StackOverflow as an example: Total weights = (0.9 × 5) + (0.8 × 2) + (0.7 × 1) = 4.5 + 1.6 + 0.7 = 6.8 The confidence level of Stack Overflow = (0.9 × 5) / 6.8 = 4.5 / 6.8 ≈ 0.66 The confidence level of the GitHub issue is (0.8 × 2) / 6.8 = 0.24 The confidence level of the enterprise document = (0.7 × 1) / 6.8 = 0.1 After weighted calculation, the confidence score of Stack Overflow is 0.66, the confidence score of GitHub Issue is 0.24, and the confidence score of Enterprise Documentation is 0.1. Only the triple with the highest confidence score is retained, namely the triple from Stack Overflow (TypeError, has_solution, Check if variable is defined).

[0094] Additionally, this application can store the knowledge graph in a Neo4j graph database. The specific steps are as follows: First, convert the merged nodes and relationships into a format supported by Neo4j (such as CSV or JSON). Then, use Neo4j's neo4j-adminimport tool for batch import (suitable for large-scale statically constructed knowledge). Clear the default data directory of the Neo4j database (to ensure a clean environment), execute the import command, and specify the node files, relationship files, and output directory. Finally, start the Neo4j service and verify the import results (confirm the total number of nodes by querying MATCH(n) RETURN count(n) using Cypher).

[0095] like Figure 6 As shown, this application embodiment provides an exception handling device, which corresponds to the method embodiment, and specifically includes: Extraction unit 601 is used to extract attribute information of abnormal events when abnormal events are detected at the front end, and obtain an attribute information sequence. Matching unit 602 is used to perform matching reasoning in a pre-established knowledge graph based on the attribute information sequence to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result; The determining unit 603 is used to determine the target matching result among all the initial matching results based on the confidence-related data corresponding to each initial matching result; The display unit 604 is used to display the target matching results so that relevant personnel can handle the abnormal event based on the display results.

[0096] Optionally, the matching unit 602 is used for: When the attribute information sequence satisfies the first preset condition, the attribute information in the attribute information sequence is matched and reasoned with the nodes in the knowledge graph in turn to obtain the initial matching result. The semantic similarity between each attribute information and the content of the corresponding matching node is the first similarity. Based on the preset first matching strategy, the first similarity, and the preset node matching level, the confidence-related data corresponding to the initial matching result is determined.

[0097] Optionally, the matching unit 602 is used for: When the attribute information sequence satisfies the second preset condition, matching reasoning is performed in the knowledge graph based on the first attribute information in the attribute information sequence to obtain at least one initial matching result. The semantic similarity between the second attribute information in the attribute information sequence and the content of the corresponding matching node is greater than the second similarity, and the similarity between the other attribute information and the content of the corresponding matching node is the first similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset second matching strategy, the semantic similarity between the second attribute information and the corresponding matching node content, and the node matching level.

[0098] Optionally, the matching unit 602 is used for: When the attribute information sequence satisfies the third preset condition, the knowledge graph is subjected to reverse derivation and forward matching reasoning based on the third attribute information in the attribute information sequence to determine at least one initial matching result, wherein the semantic similarity between the third attribute information and the content of the corresponding matching node is greater than the second similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset third matching strategy, the semantic similarity between the third attribute information and the corresponding matching node content, and the node matching level.

[0099] Optionally, the determining unit 603 is used for: For each initial matching result, the matching strategy, semantic similarity, and node matching level are obtained from the corresponding confidence-related data. The first weight corresponding to the matching strategy and the second weight corresponding to the node matching level are determined. Based on the first weight, the second weight, and the semantic similarity, the confidence level corresponding to the initial matching result is obtained. The target matching result is determined based on the initial matching results whose confidence level is greater than the confidence threshold.

[0100] Optionally, display unit 604 is used for: Based on the confidence level corresponding to each target matching result, the target matching results are initially sorted to obtain a first sequence; For the target matching results with the same confidence level in the first sequence, a second sequence is obtained by sorting them a second time according to the node matching level. For target matching results with the same node matching level in the second sequence, sort them three times according to the semantic similarity to obtain the target sequence; The display is based on the target sequence.

[0101] Optionally, the apparatus further includes a construction unit 605, the construction unit 605 being configured to: Retrieve target exception data related to console exceptions; Based on the target abnormal data, a reference triple is constructed, which includes the data corresponding to the current node, the data included in the previous level node, and the data included in the next level node. A knowledge graph is constructed based on the reference triples.

[0102] like Figure 7 As shown in the figure, this application embodiment provides an exception handling device, including a processor 701, a communication interface 702, a memory 703, and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704. Memory 703 is used to store computer programs; In one embodiment of this application, when the processor 701 executes a program stored in the memory 703, it implements the exception handling method provided in any of the foregoing method embodiments, including: Upon detecting an abnormal event on the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence; Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result. Based on the confidence-related data corresponding to each initial matching result, the target matching result is determined among all the initial matching results; The results of the target matching are displayed so that relevant personnel can handle the abnormal events based on the displayed results.

[0103] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps performed by the exception handling method provided in any of the foregoing method embodiments.

[0104] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0106] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0107] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An abnormality processing method characterized by comprising: The method includes: Upon detecting an abnormal event from the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence. Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result. Based on the confidence-related data corresponding to each initial matching result, the target matching result is determined among all the initial matching results; The results of the target matching are displayed so that relevant personnel can handle the abnormal events based on the displayed results.

2. The method of claim 1, wherein, Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result, including: When the attribute information sequence satisfies the first preset condition, the attribute information in the attribute information sequence is matched and reasoned with the nodes in the knowledge graph in turn to obtain the initial matching result. The semantic similarity between each attribute information and the content of the corresponding matching node is the first similarity. Based on the preset first matching strategy, the first similarity, and the preset node matching level, the confidence-related data corresponding to the initial matching result is determined.

3. The method according to claim 1, characterized in that, Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result, including: When the attribute information sequence satisfies the second preset condition, matching reasoning is performed in the knowledge graph based on the first attribute information in the attribute information sequence to obtain at least one initial matching result. The semantic similarity between the second attribute information in the attribute information sequence and the content of the corresponding matching node is greater than the second similarity, and the similarity between the other attribute information and the content of the corresponding matching node is the first similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset second matching strategy, the semantic similarity between the second attribute information and the corresponding matching node content, and the node matching level.

4. The method according to claim 1, characterized in that, Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result, including: When the attribute information sequence satisfies the third preset condition, the knowledge graph is subjected to reverse derivation and forward matching reasoning based on the third attribute information in the attribute information sequence to determine at least one initial matching result, wherein the semantic similarity between the third attribute information and the content of the corresponding matching node is greater than the second similarity. For each initial matching result, determine the number of attributes matched in the initial matching result and the attribute information sequence, determine the node matching level based on the number, and determine the confidence-related data corresponding to the initial matching result based on the preset third matching strategy, the semantic similarity between the third attribute information and the corresponding matching node content, and the node matching level.

5. The method according to claim 1, characterized in that, Based on the confidence-related data corresponding to each initial matching result, the target matching result is determined from all initial matching results, including: For each initial matching result, the matching strategy, semantic similarity, and node matching level are obtained from the corresponding confidence-related data. The first weight corresponding to the matching strategy and the second weight corresponding to the node matching level are determined. Based on the first weight, the second weight, and the semantic similarity, the confidence level corresponding to the initial matching result is obtained. The target matching result is determined based on the initial matching results whose confidence level is greater than the confidence threshold.

6. The method according to claim 5, characterized in that, The results are displayed based on the target matching, including: Based on the confidence level corresponding to each target matching result, the target matching results are initially sorted to obtain a first sequence; For the target matching results with the same confidence level in the first sequence, a second sequence is obtained by sorting them a second time according to the node matching level. For target matching results with the same node matching level in the second sequence, sort them three times according to the semantic similarity to obtain the target sequence; The display is based on the target sequence.

7. The method according to claim 1, characterized in that, The method further includes: Retrieve target exception data related to console exceptions; Based on the target abnormal data, a reference triple is constructed, which includes the data corresponding to the current node, the data included in the previous level node, and the data included in the next level node. A knowledge graph is constructed based on the reference triples.

8. An anomaly handling device, characterized in that, The device includes: The extraction unit is used to extract the attribute information of the abnormal event when an abnormal event is detected at the front end, and obtain an attribute information sequence. The matching unit is used to perform matching reasoning in a pre-established knowledge graph based on the attribute information sequence to obtain at least one initial matching result and confidence-related data corresponding to the initial matching result. The determination unit is used to determine the target matching result among all the initial matching results based on the confidence-related data corresponding to each initial matching result; The display unit is used to display the target matching results so that relevant personnel can handle the abnormal event based on the display results.

9. An anomaly handling device, characterized in that, include: At least one communication interface; At least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; At least one memory connected to the at least one bus, wherein the processor is configured to: Upon detecting an abnormal event from the front end, the attribute information of the abnormal event is extracted to obtain an attribute information sequence. Based on the attribute information sequence, matching reasoning is performed in a pre-established knowledge graph to obtain at least one initial matching result and the confidence-related data corresponding to the initial matching result. Based on the confidence-related data corresponding to each initial matching result, the target matching result is determined among all the initial matching results; The results of the target matching are displayed so that relevant personnel can handle the abnormal events based on the displayed results.

10. 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 exception handling method according to any one of claims 1 to 7.

Images