Method for constructing fault knowledge graph, and computing apparatus

Abstract

A method for constructing a fault knowledge graph. The method comprises: after a sentence to be processed is acquired according to fault discrimination text, determining protocol signaling and an atomic flow of said sentence from an atomic flow chart according to a signaling character string of said sentence; after an atomic flow clause is generated according to the atomic flow and said sentence, determining a logic expression corresponding to the atomic flow clause; generating a logic expression of said sentence according to the logic expression of the atomic flow clause; then generating a logic expression of the fault discrimination text according to the logic expression of said sentence; and adding the logic expression of the fault discrimination text to a fault knowledge graph. By means of the method, corresponding protocol signaling and a corresponding atomic flow can be searched for in an atomic flow chart according to a signaling character string of fault discrimination text, such that the accuracy of identifying the signaling can be improved, and the success rate of converting the fault discrimination text into a logic expression is improved. Further provided in the present application is a computing apparatus capable of implementing the method.

Description

A method and computing device for constructing a fault knowledge graph

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office on June 16, 2023, with application number 202310718965.3 and invention name “A method and computing device for constructing a fault knowledge graph”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of wireless communications, and in particular to a method and computing device for constructing a fault knowledge graph. Background Art

[0003] The fault documentation of the telecommunications network system records various fault phenomena, problem analysis and location, and solution processes.

[0004] Currently, there is a method for constructing a fault knowledge graph that is roughly as follows: extract entities from the fault discrimination text based on the entity dictionary, perform part-of-speech tagging and dependency syntactic analysis on the sentences in the fault discrimination text; generate logical expressions based on the parameters obtained from the part-of-speech tagging and the logical relationships obtained from the dependency syntactic analysis, and form a fault knowledge graph based on the logical expressions of a large number of fault discrimination texts.

[0005] Since a large amount of signaling in fault judgment texts is written in natural language, it is difficult to identify non-standard signaling based on entity dictionaries. Therefore, a large amount of fault judgment texts cannot be converted into logical expressions.

[0006] Summary of the Invention

[0007] In light of this, this application provides a method for constructing a fault knowledge graph. This method can search for corresponding protocol signaling and atomic processes in an atomic process graph based on the signaling string of the fault diagnosis text. This method can improve the success rate of identifying the signaling string, thereby improving the success rate of converting the fault diagnosis text into a logical expression. This application also provides a computing device capable of implementing the above method.

[0008] The first aspect provides a method for constructing a fault knowledge graph, which includes: after obtaining a sentence to be processed according to a fault judgment text, obtaining a signaling string of the sentence to be processed, and then determining the protocol signaling of the sentence to be processed from the atomic flow map according to the signaling string of the sentence to be processed, and determining the atomic flow to which the protocol signaling of the sentence to be processed belongs from the atomic flow map, and then generating at least one atomic flow clause according to the atomic flow and the sentence to be processed, determining the question template corresponding to the atomic flow clause, and then generating a logical expression of the atomic flow clause according to the logical expression template corresponding to the question template, the protocol signaling corresponding to the atomic flow clause, and the non-signaling entity of the atomic flow clause, generating a logical expression of the sentence to be processed according to the logical expression of at least one atomic flow clause; generating a logical expression of the fault judgment text according to the logical expression of the sentence to be processed; and adding the logical expression of the fault judgment text to the fault knowledge graph.

[0009] According to this implementation, the corresponding protocol signaling and atomic process can be found in the atomic process map based on the signaling string of the fault judgment text. Since the protocol signaling and atomic process are standardized objects, the success rate of identifying the signaling string can be improved, thereby improving the success rate of converting the fault judgment text into a logical expression.

[0010] In a possible implementation of the first aspect, obtaining sentences to be processed based on the fault discrimination text includes: when the sentences in the fault discrimination text do not include logical operation keywords, treating the sentences in the fault discrimination text as sentences to be processed; when the sentences in the fault discrimination text include logical operation keywords, determining multiple subjects in the sentences in the fault discrimination text based on the logical operation keywords, treating the text obtained by removing the logical operation keywords and multiple subjects from the sentences in the fault discrimination text as a common expression text, and generating a sentence to be processed based on the common expression text and each subject. In this way, sentences including logical operation keywords can be split and reorganized into multiple sentences, reducing the inconsistency between the signaling of the sentences to be processed and the signaling of the atomic process, and improving the recognition success rate of the atomic process.

[0011] In another possible implementation of the first aspect, obtaining the signaling string of the sentence to be processed includes: when the sentence to be processed includes a keyword from a preset vocabulary, replacing the keyword from the preset vocabulary with an English string corresponding to the keyword; extracting at least one candidate string from the replaced sentence, and selecting a candidate string including a preset signaling type as the signaling string of the sentence to be processed. In this way, the signaling written in Chinese can be converted into an English string, thereby increasing the number of recognizable signaling strings. Since the length of the signaling string is greater than 2 and the signaling string includes a preset signaling type (such as ACK, REQ, REL, or RGNB), some strings that do not belong to signaling can be removed based on the string length, reducing the computational overhead of extracting the signaling string.

[0012] Based on the previous possible implementation, in another possible implementation, the method for constructing a fault knowledge graph of the present application further includes: before selecting a candidate string including a preset signaling type as the signaling string of the sentence to be processed, removing a target candidate string from at least one candidate string. The target candidate string includes one or more of a string containing a "-", a string beginning with a number, a standard name, a network element name, or a port name. In this way, strings that do not belong to signaling can be removed, reducing the computational overhead of subsequent signaling matching and improving the efficiency of identifying signaling strings.

[0013] In another possible implementation, determining the protocol signaling of the sentence to be processed from the atomic flow map based on the signaling string of the sentence to be processed includes: selecting a set of candidate protocol signaling from the atomic flow map based on the signaling type of the signaling string of the sentence to be processed, and then calculating the fuzzy matching score corresponding to the candidate protocol signaling based on the length of the signaling string of the sentence to be processed, the length of the candidate protocol signaling, and the edit distance between the signaling string of the sentence to be processed and the candidate protocol signaling; and determining the protocol signaling of the sentence to be processed based on the candidate protocol signaling corresponding to the highest fuzzy matching score. The candidate protocol signaling set includes one or more candidate protocol signalings, and the protocol signaling is a standard signaling in the 3GPP protocol. This provides a method for searching for protocol signaling based on a signaling string, which has a high accuracy rate.

[0014] In another possible implementation, determining the protocol signaling of the sentence to be processed based on the candidate protocol signaling corresponding to the highest fuzzy matching score includes: when there are multiple candidate protocol signalings corresponding to the highest fuzzy matching score, selecting a candidate protocol signaling from the candidate protocol signalings with the highest fuzzy matching score as the protocol signaling of the sentence to be processed according to a preset matching rule. The preset matching rule includes at least one of the following rules: the atomic process to which the signaling string belongs is the same as the atomic process to which the candidate protocol signaling belongs, the signaling type of the signaling string and the signaling type of the candidate protocol signaling respond to each other, and the port type of the signaling string is the same as the port type of the candidate protocol signaling. When there are multiple candidate protocol signalings corresponding to the highest fuzzy matching score, the accuracy of finding the protocol signaling can be improved based on the preset matching rule.

[0015] In another possible implementation, determining the protocol signaling of the signaling to be processed from the atomic flow map based on the signaling string of the sentence to be processed includes: when the signaling string preceding the signaling string of the sentence to be processed matches the first target protocol signaling, determining the protocol signaling of the signaling to be processed as the next protocol signaling of the first target protocol signaling in the atomic flow to which the first target protocol signaling belongs. For signaling strings that include partial signaling (such as ACK or REQ), the protocol signaling can be searched based on the preceding signaling string, thereby improving the success rate of finding the protocol signaling.

[0016] In another possible implementation, determining the protocol signaling of the signaling to be processed from the atomic flow map based on the signaling string of the sentence to be processed includes: when a subsequent signaling string of the signaling string of the sentence to be processed matches a second target protocol signaling, determining that the protocol signaling of the sentence to be processed is the protocol signaling preceding the second target protocol signaling in the atomic flow to which the second target protocol signaling belongs. For signaling strings that include partial signaling (e.g., ACK or REQ), the protocol signaling can be searched based on the subsequent signaling string, thereby improving the success rate of searching for the protocol signaling.

[0017] In another possible implementation, the method for constructing a fault knowledge graph of the present application further includes: when a previous signaling string fails to match the protocol signaling of the atomic flow graph, and / or when a subsequent signaling string fails to match the protocol signaling of the atomic flow graph, searching for a target signaling string in the text following the signaling string of the sentence to be processed according to the signaling type of the signaling string of the sentence to be processed; when the target signaling string matches the third target protocol signaling, determining that the protocol signaling of the sentence to be processed is the third target protocol signaling. This can improve the success rate of finding the protocol signaling.

[0018] In another possible implementation, determining the question template corresponding to the atomic process clause includes: replacing the signaling string and non-signaling entity in the atomic process clause with a mask string to obtain a query statement; after obtaining the set of candidate question templates corresponding to the atomic process, selecting a subset of candidate question templates from the candidate question template set according to the signaling quantity of the query statement; for each candidate question template in the candidate question template subset, calculating the fuzzy matching score between the candidate question template and the query statement; determining that the question template corresponding to the atomic process clause is the candidate question template corresponding to the highest matching score. The signaling quantity of each candidate question template in the candidate question template subset is the same as the signaling quantity of the query statement. The mask string of the query statement is the same as the signaling string corresponding to the question template, which can improve the success rate of finding the question template. Finding the question template according to the fuzzy matching score has a high accuracy rate.

[0019] In another possible implementation, generating the logical expression of the atomic flow clause based on the logical expression template corresponding to the question template, the protocol signaling corresponding to the atomic flow clause, and the non-signaling entity of the atomic flow clause includes: filling the slots of the logical expression template corresponding to the question template with the protocol signaling corresponding to the atomic flow clause and the non-signaling entity of the atomic flow clause to obtain the logical expression of the atomic flow clause. This provides a specific method for generating the logical expression of the atomic flow clause.

[0020] In another possible implementation, generating a logical expression for a sentence to be processed based on a logical expression of at least one atomic flow clause includes: when there are multiple atomic flow clauses, obtaining logical relationships between the multiple atomic flow clauses; and generating a logical expression for the sentence to be processed based on the logical relationships between the multiple atomic flow clauses and the logical expressions of the multiple atomic flow clauses. The logical relationships between the multiple atomic flow clauses include AND and OR, so that the logical expressions of the multiple atomic flow clauses can be combined into a logical expression for the sentence to be processed based on the logical relationships between the atomic flow clauses.

[0021] In another possible implementation, the method of constructing a fault knowledge spectrum graph of the present application also includes: when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text do not include logical operation keywords, determining the inter-sentence logical relationship of all sentences to be processed according to the hierarchical identifier; when the fault judgment text does not include a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determining the inter-sentence logical relationship of all sentences to be processed according to the logical operation keywords; when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determining the inter-sentence logical relationship of all sentences to be processed according to the hierarchical identifier and the logical operation keywords; and then generating a logical expression of the fault judgment text based on the inter-sentence logical relationship of all sentences to be processed and the logical expression of the sentences to be processed.

[0022] When the fault judgment text includes hierarchical sentences and / or parallel sentences, the inter-sentence logical relations of the sentences to be processed can be obtained, and the logical expressions of the sentences to be processed can be combined into the logical expressions of the fault judgment text according to the inter-sentence logical relations of the sentences to be processed.

[0023] In another possible implementation, the method of constructing a fault knowledge graph of the present application also includes: extracting an atomic flow chart from a protocol document, creating an atomic flow chart based on the atomic flow chart of the protocol document, obtaining a set of fault description sentences from the fault document; obtaining a problem template corresponding to each fault description sentence in the set of fault description sentences; when all the signaling of the target problem template belongs to the signaling set of the target atomic process, associating the target problem template with the atomic process identifier of the target atomic process. The atomic flow chart includes the atomic process identifier and the signaling set of the atomic process. The target problem template is any problem template. In this way, a correspondence between the problem template and the atomic process can be established, and based on the correspondence, the problem template corresponding to the atomic process can be automatically found.

[0024] In another possible implementation, the method for constructing a fault knowledge graph in the present application further includes: receiving logs sent by network devices; searching for a target logical expression in the fault knowledge graph based on the signaling strings and non-signaling entities included in the logs; and determining whether the logs contain a fault event based on the output of the target logical expression. After constructing the fault knowledge graph, fault detection can be automatically performed based on the fault knowledge graph, thereby improving fault detection speed.

[0025] The second aspect provides a computing device, which includes an acquisition module, a first matching module, a second matching module and a generation module. The acquisition module is used to acquire a sentence to be processed based on a fault judgment text and acquire a signaling string of the sentence to be processed; the first matching module is used to determine the protocol signaling of the sentence to be processed from the atomic flow map based on the signaling string of the sentence to be processed; the second matching module is used to determine the atomic flow to which the protocol signaling of the sentence to be processed belongs from the atomic flow map; generate at least one atomic flow clause based on the atomic flow and the sentence to be processed; determine the question template corresponding to the atomic flow clause; the generation module is used to generate a logical expression of the atomic flow clause based on the logical expression template corresponding to the question template, the protocol signaling corresponding to the atomic flow clause and the non-signaling entity of the atomic flow clause; generate a logical expression of the sentence to be processed based on the logical expression of at least one atomic flow clause; generate a logical expression of the fault judgment text based on the logical expression of the sentence to be processed; and add the logical expression of the fault judgment text to the fault knowledge graph.

[0026] In one possible implementation, the acquisition module is specifically used to treat the sentences of the fault judgment text as sentences to be processed when the sentences of the fault judgment text do not include logical operation keywords; when the sentences of the fault judgment text include logical operation keywords, determine multiple subjects in the sentences of the fault judgment text according to the logical operation keywords, and generate a sentence to be processed based on the common expression text and each subject.

[0027] In another possible implementation, the acquisition module is specifically used to replace the keywords in the preset vocabulary with English strings corresponding to the keywords when the sentence to be processed includes keywords in the preset vocabulary; extract at least one candidate string from the replaced sentence, and select the candidate string including the preset signaling type as the signaling string of the sentence to be processed.

[0028] In another possible implementation, the acquisition module is further configured to remove the target candidate character string from the at least one candidate character string.

[0029] In another possible implementation, the first matching module is specifically used to select a set of candidate protocol signalings from the atomic flow chart according to the signaling type of the signaling string of the sentence to be processed; calculate the fuzzy matching score corresponding to the candidate protocol signaling according to the length of the signaling string of the sentence to be processed, the length of the candidate protocol signaling, and the edit distance between the signaling string of the sentence to be processed and the candidate protocol signaling; and determine the protocol signaling of the sentence to be processed according to the candidate protocol signaling corresponding to the highest fuzzy matching score.

[0030] In another possible implementation, the first matching module is specifically configured to select a candidate protocol signaling from the candidate protocol signaling with the highest fuzzy matching score as the protocol signaling of the sentence to be processed according to a preset matching rule when there are multiple candidate protocol signalings corresponding to the highest fuzzy matching score.

[0031] In another possible implementation, the first matching module is specifically used to determine that the protocol signaling of the signaling to be processed is the subsequent protocol signaling of the first target protocol signaling in the atomic process to which the first target protocol signaling belongs when the previous signaling string of the signaling string of the sentence to be processed matches the first target protocol signaling.

[0032] In another possible implementation, the first matching module is specifically used to determine that the protocol signaling of the sentence to be processed is the previous protocol signaling of the second target protocol signaling in the atomic process to which the second target protocol signaling belongs when the latter signaling string of the signaling string of the sentence to be processed matches the second target protocol signaling.

[0033] In another possible implementation, the first matching module is specifically used to search for a target signaling string in the target text according to the signaling type of the signaling string of the sentence to be processed when the current signaling string fails to match the protocol signaling of the atomic flowchart, and / or when the subsequent signaling string fails to match the protocol signaling of the atomic flowchart, and when the target signaling string matches the third target protocol signaling, determine that the protocol signaling of the sentence to be processed is the third target protocol signaling.

[0034] In another possible implementation, the second matching module is specifically used to replace the signaling strings and non-signaling entities in the atomic process clause with mask strings to obtain a query statement; obtain a set of candidate question templates corresponding to the atomic process; select a subset of candidate question templates from the candidate question template set according to the signaling quantity of the query statement, and for each candidate question template in the candidate question template subset, calculate the fuzzy matching score between the candidate question template and the query statement, and determine that the question template corresponding to the atomic process clause is the candidate question template corresponding to the highest matching score.

[0035] In another possible implementation, the generation module is specifically used to fill the protocol signaling corresponding to the atomic process clause and the non-signaling entity of the atomic process clause in the slot of the logical expression template corresponding to the problem template to obtain the logical expression of the atomic process clause.

[0036] In another possible implementation, the generation module is specifically used to obtain the inter-sentence logical relationship of multiple atomic flow clauses when there are multiple atomic flow clauses; and generate the logical expression of the sentence to be processed based on the inter-sentence logical relationship of multiple atomic flow clauses and the logical expressions of multiple atomic flow clauses.

[0037] In another possible implementation, the acquisition module is also used to determine the inter-sentence logical relationship of all sentences to be processed according to the hierarchical identifier when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text do not include logical operation keywords; when the fault judgment text does not include a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determine the inter-sentence logical relationship of all sentences to be processed according to the logical operation keywords; when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determine the inter-sentence logical relationship of all sentences to be processed according to the hierarchical identifier and the logical operation keywords; the generation module is specifically used to generate a logical expression of the fault judgment text based on the inter-sentence logical relationship of all sentences to be processed and the logical expression of the sentences to be processed.

[0038] In another possible implementation, the computing device of the present application also includes a first creation module, a second creation module and an association module, the first creation module is used to extract an atomic flow chart from a protocol document; create an atomic flow chart based on the atomic flow chart of the protocol document; the second creation module is used to obtain a set of fault description sentences from a fault document; obtain the problem template corresponding to each fault description sentence in the set of fault description sentences; the association module is used to associate the target problem template with the atomic flow identifier of the target atomic flow when all signaling of the target problem template belongs to the signaling set of the target atomic flow.

[0039] In another possible implementation, the computing device of the present application also includes a receiving module and a fault detection module. The receiving module is used to receive logs sent by network equipment; the fault detection module is used to search for a target logical expression in a fault knowledge graph based on the signaling strings and non-signaling entities included in the log; and determine whether the log contains a fault event based on the output result of the target logical expression.

[0040] For the explanation of terms in the second aspect, the steps performed by each module and the beneficial effects, please refer to the corresponding description of the first aspect.

[0041] The third aspect provides a computing device, which includes a processor and an interface circuit, the interface circuit is used to receive signals from other computing devices and transmit them to the processor or send signals from the processor to other computing devices, and the processor is used to implement the method of the first aspect through logic circuits or executing code instructions.

[0042] A fourth aspect provides a computer-readable storage medium, in which a computer program or instruction is stored. When the computer program or instruction is executed by a computing device, the method of the first aspect is implemented.

[0043] A fifth aspect provides a computer program product comprising instructions, which, when run on a computer, causes the computer to perform the method of the first aspect.

[0044] A sixth aspect provides a chip comprising a processor and a memory, wherein the memory stores instructions, and the processor is configured to execute the instructions stored in the memory. When the instructions are executed, the processor is configured to perform the method of the first aspect. The chip can be a field-programmable gate array (FPGA) or an application-specific integrated circuit (ASIC). BRIEF DESCRIPTION OF THE DRAWINGS

[0045] FIG1 is a schematic diagram of a scenario for constructing a fault knowledge graph in an embodiment of the present application;

[0046] FIG2 is a schematic diagram of a fault detection scenario in an embodiment of the present application;

[0047] FIG3 is a schematic diagram of constructing telecommunication knowledge middleware and generating logical expressions in an embodiment of the present application;

[0048] FIG4 is a schematic diagram of an atomic process in an embodiment of the present application;

[0049] FIG5 is a schematic diagram of a logical expression of a fault discrimination text in an embodiment of the present application;

[0050] FIG6 is a schematic diagram of a question template in an embodiment of the present application;

[0051] FIG7 is a flowchart of a method for constructing a fault knowledge graph in an embodiment of the present application;

[0052] FIG8 is a schematic diagram of sentence reorganization in an embodiment of the present application;

[0053] FIG9 is another schematic diagram of sentence reorganization according to an embodiment of the present application;

[0054] FIG10 is a schematic diagram of converting an atomic process clause into a query statement in an embodiment of the present application;

[0055] FIG11 is a schematic diagram of a logical expression for generating an atomic flow clause in an embodiment of the present application;

[0056] FIG12 is a structural diagram of a computing device according to an embodiment of the present application;

[0057] FIG13 is another structural diagram of the computing device in an embodiment of the present application. DETAILED DESCRIPTION

[0058] The method for constructing a fault knowledge graph in this application can be applied to fault detection scenarios based on fault knowledge graphs in wireless communication systems. The following describes the scenario for constructing a fault knowledge graph. Referring to Figure 1, in one example, telecommunications knowledge middleware includes a problem template and a signaling interaction process. After a computing device obtains a fault judgment text, a problem template, and a signaling interaction process, it can generate a logical expression for the fault judgment text based on the fault judgment text, the problem template, and the signaling interaction process. The logical expression of the fault judgment text is added to the fault knowledge graph, and fault detection is performed based on the fault knowledge graph.

[0059] The fault identification text is a natural language text describing the fault. For example, "After 4G initiates an SGNB add req and does not receive an ACK within 5 seconds, it sends an SNB rel req with the cause value tDCoverall-expiry."

[0060] Telecom knowledge middleware includes atomic process data and problem templates for the 3rd Generation Partnership Project (3GPP) protocol. Atomic process data includes atomic processes and their signaling. An atomic process is the smallest signaling interaction process in a communication protocol with complete, independent functionality, typically consisting of one to three signaling interaction steps between two network elements.

[0061] Problem templates are generated according to preset rules and can describe the fault phenomenon, for example, [send / send / initiate] <sig:req>Pass <time>[Not received / Not received / Not received] <sig:ack>. " / " means or. <sig:req>Indicates REQ signaling, <sig:ack>Indicates ACK signaling, <time>Indicates time.

[0062] Logical expressions are composed of operators and named entities. The operators in this application include but are not limited to operators for logical relationships between parameters and constants, operators for logical relationships between parameters, operators for logical operations, and operators for numerical operations.

[0063] In an example, the logical relationship between parameters and constants is shown in Table 1:

[0064] Table 1

[0065] In an example, the logical relationship between parameters is shown in Table 2:

[0066] Table 2

[0067] In an example, the numerical operation relationship characters are shown in Table 3:

[0068] Table 3

[0069] In an example, the logical operation relationship is shown in Table 4:

[0070] Table 4

[0071] Named entities include parameters, values, time, numbers, etc. For example, the logical expression is: ((time(SGNB_ADD_REQ_ACK)-time(SGNB_ADD_REQ)>5000)and(SGNB_REL_REQ:cause==tDCoverall-expiry). Among them, SGNB_ADD_REQ_ACK, SGNB_ADD_REQ, and SGNB_REL_REQ are signaling entities, time() represents the time when the signaling is obtained, 5000 represents 5000 microseconds (i.e., 5 seconds), and tDCoverall-expiry represents a parameter. Named entities are things identified by proper nouns, including signaling names, names of people, places, and so on.

[0072] Referring to FIG2 , the network device sends the log to the computing device, which searches for the corresponding logical expression in the fault knowledge graph based on the signaling and signaling parameters included in the log, and outputs the fault detection result based on the found logical expression.

[0073] The process of building electronic middleware includes the process of building an atomic flow map and the process of building a fault template library. The following is a brief introduction to the process of building an atomic flow map and the process of building a fault template library. In one embodiment, the process of building a principle flow map includes steps 301 and 302.

[0074] Step 301: Extract the atomic flow chart.

[0075] This application can extract atomic flowcharts from 3GPP protocol documents. Tools for batch image extraction can include, but are not limited to, Office.Files.Images. The extracted images may include miscellaneous images and unnecessary atomic flowcharts, which can be removed by expert screening. For example, an atomic flowchart of the 3GPP protocol is shown in Figure 4.

[0076] Step 302: Generate an atomic flow chart.

[0077] Specifically, signaling attributes, process attributes, and signaling transmission direction are obtained from the atomic flow graph. The nodes of the atomic flow graph are formed according to the signaling attributes and process attributes, and the edges of the atomic flow graph are formed according to the signaling transmission direction. The atomic flow graph refers to a knowledge graph that includes atomic process information. The knowledge graph is a graph-based data structure that can be considered a semantic network composed of nodes and edges. Each node represents an entity, and each edge represents the relationship between entities. Methods for obtaining information from the atomic flow graph include, but are not limited to, optical character recognition (OCR).

[0078] Signaling attributes include at least one of the following: signaling sender, signaling receiver, signaling name, signaling type, port type, and signaling sequence number. The signaling sender is the network element that sends the signaling. The signaling receiver is the network element that receives the signaling. Signaling types include request types (such as REQ), response types (such as ACK), information types (such as INFO), reconfiguration types (such as RECFG), etc. Port types include S1, NG, X2, UU, etc. It should be understood that 3GPP signaling types and port types are not limited to the above examples. The signaling sequence number is the sequence number of the step in which the signaling is located, and the signaling sequence number can be numbered starting from 1.

[0079] Flow attributes include atomic flow identifiers (such as atomic flow names) and atomic flow types. Atomic flow types include signaling types that are concatenated according to the order of signaling interactions. For example, REQ-ACK represents an atomic flow consisting of a request and a response in sequence; RECFG-CMP represents an atomic flow consisting of a reconfiguration and a completion in sequence.

[0080] The process of building a fault template library includes step 303 and step 304 .

[0081] Step 303: Extract the atomic process clause.

[0082] An atomic process clause is a sentence used to describe a signaling fault and corresponds to an atomic process.

[0083] Step 304: Build a question template library.

[0084] Specifically, a question template and a question template identifier are generated according to the atomic process clause. The question template includes named entities and relationships between named entities. Named entities include but are not limited to keywords, parameter types, signaling entities or non-signaling entities.

[0085] Step 305: Associate the atomic process and the problem template.

[0086] For the signaling type involved in the problem template and the signaling type involved in the atomic process, the problem template is associated with the atomic process type of the atomic process map.

[0087] The following describes a method for generating a logical expression. Referring to FIG3 , in one embodiment, the method for generating a logical expression includes:

[0088] Step 306: Text preprocessing.

[0089] Text preprocessing includes data cleaning and sentence segmentation. Data cleaning includes at least one of removing illegal characters and removing excess whitespace. Sentence segmentation divides multiple sentences into a sequence of sentences based on line breaks. Text preprocessing is optional.

[0090] Step 307: logical clauses.

[0091] Logical clause segmentation includes embedded logical splitting and inter-sentence nested logical analysis.

[0092] Embedded logic splitting includes embedded logical relationship identification, sentence component analysis, and sentence splitting and reorganization. Embedded logical relationship identification refers to the logical relationship corresponding to keywords (for example, or, either, and, and). Sentence component analysis uses regular expressions to identify the subjects of embedded logical connections and the public expression text. Sentence splitting and reorganization refers to recombining the subjects and the public expression text into a complete sentence. The logical relationship between the reorganized sentences is the logical relationship corresponding to the keywords.

[0093] Inter-sentence nesting logic parsing includes: using regular expressions to identify the level to which each sentence belongs, and for sentences at the same level, reorganizing the sentences based on embedded logic splitting. The logical relationship of sentences includes the hierarchical relationship and the logical relationship corresponding to the keywords. In one example, the identification of the hierarchical relationship includes the first-level serial number and the second-level serial number as shown in Table 5:

[0094] Table 5

[0095] Step 308: Parameter identification.

[0096] Parameter identification includes string filtering and rule filtering.

[0097] The string screening includes: extracting English strings with a length greater than k after replacing keywords using a preset word list. The Chinese characters included in the preset word list correspond to signaling or signaling parameters.

[0098] Rule filtering includes: filtering strings using business rules. Business rules include one or more of the following:

[0099] Rule 1: A character string containing a preset signaling type is a signaling character string. The preset signaling type includes but is not limited to ACK, REQ, REL, and SGNB.

[0100] Rule 2: A string containing "-" is not a signaling string.

[0101] Rule 3: A string starting with a number is not a signaling string.

[0102] Rule 4: The standard name is not a signaling string.

[0103] Rule 5: The network element name is not a signaling string.

[0104] Rule 6: Port names are not signaling strings.

[0105] Step 309: Parameter linking.

[0106] Parameter linking is a method of linking the signaling string in the fault judgment text to the atomic process signaling. Parameter linking includes basic signaling matching, multi-candidate matching processing, and fuzzy signaling matching.

[0107] Basic signaling matching involves the following: If the signaling type of the signaling string differs from the signaling type of the candidate protocol signaling, the match score for the candidate protocol signaling is -1. If the signaling type of the signaling string and the candidate protocol signaling are the same, a fuzzy match score is calculated between the signaling string and the candidate protocol signaling. Fuzzy matching uses custom matching rules, which differ from strict matching rules or regular expressions.

[0108] The multi-candidate matching process includes: selecting a target protocol signaling from a plurality of candidate protocol signalings having the same fuzzy matching score according to a preset matching rule.

[0109] Preset matching rules include one or more of the following:

[0110] Rule 7: If the atomic process to which the signaling string belongs is consistent with the atomic process to which the candidate protocol signaling belongs, 1 point is awarded.

[0111] Rule 8: If the signaling type of the signaling string and the candidate protocol signaling type are mutually responsive, 1 point is awarded. Mutually responsive signaling types include REQ and ACK, REQUIRED-CONFIRM, etc.

[0112] Rule 9: If the port type of the signaling string is the same as the port type of the candidate protocol signaling, 1 point is awarded.

[0113] Fuzzy signaling matching includes: after obtaining the previous signaling of the signaling string, determining the protocol signaling corresponding to the previous signaling of the signaling string, and then using the next signaling of the protocol signaling in the atomic process as the protocol signaling corresponding to the signaling string. Alternatively, after obtaining the next signaling of the signaling string, determining the protocol signaling corresponding to the next signaling of the signaling string, and then using the previous signaling of the protocol signaling in the atomic process as the protocol signaling corresponding to the signaling string. Alternatively, when the previous signaling string fails to match the protocol signaling of the atomic process map, searching for a signaling string of the same signaling type in the text following the signaling string based on the signaling type of the signaling string; when a signaling string of the same signaling type matches the protocol signaling, determining that the protocol type is the protocol signaling of the sentence to be processed.

[0114] Step 310: Question template matching.

[0115] Question template matching includes text segmentation, non-signaling entity recognition, query statement construction and fuzzy matching of signaling constraints.

[0116] Text segmentation involves determining the atomic process to which each signaling request belongs based on parameter linking results, then segmenting the sentence into clauses according to the atomic process, ensuring that the signaling in each clause belongs to the same atomic process. For example, the sentence "4G initiates SGNB add req, but does not receive an ACK within 5 seconds, and then sends SNB rel req with the cause value tDCoverall-expiry" can be divided into "4G initiates SGNB add req, but does not receive an ACK within 5 seconds" and "Sends SNB rel req with the cause value tDCoverall-expiry," corresponding to the "Access Add Preparation" and "Release" processes, respectively.

[0117] Non-signaling entity recognition refers to the use of regular expressions to identify non-signaling entities in clauses. Non-signaling entities include time, standard, port, number, and other parameter values.

[0118] Query statement construction refers to masking entities in the text, replacing signaling entities and non-signaling entities with corresponding mask strings, thereby obtaining a query statement corresponding to the question template, in order to improve the accuracy of template matching. Mask strings include but are not limited to <sig:type> 、 <time>、 <mode> 、 <port> 、 <num>or <value>.in, <sig:type>is signaling, and TYPE is the signaling type. <time>Corresponding time, <mode>Corresponding standard, <port>Corresponding port, <num>Corresponding numbers, <value>Corresponding parameter value.

[0119] Fuzzy matching of signaling constraints refers to obtaining the corresponding question template based on the signaling string of the atomic process clause, thereby obtaining a candidate set of question templates.

[0120] Step 311: Generate a logical expression of the fault judgment text.

[0121] Generating the logical expression for the fault diagnosis text involves filling in the problem template's slots and combining logical expressions. Filling in the problem template's slots involves filling the protocol signaling and non-signaling entities into the slots of the logical expression template corresponding to the problem template. Combining logical expressions involves connecting the logical expressions of multiple sentences based on the logical relationships between the characters in the sentences, creating the logical expression corresponding to the fault diagnosis text.

[0122] Referring to Figure 5, in one example, the fault judgment text is:

[0123] One of the following conditions is met:

[0124] 1. If 4G initiates an SGNB add req and does not receive an ACK within 5 seconds, it sends an SNB rel req with the reason tDCoverall-expiry.

[0125] 2. 5G immediately responds with an SNB ADD ack message. If no response is received from 4G after 20 seconds, the 5G sends an SNB rel require message with the reason value tDCoverall-expiry.

[0126] The logical expression of the fault judgment text is:

[0127] ((time(SGNB_ADD_REQ_ACK)-time(SGNB_ADD_REQ)>5000and(SGNB_REL_REQ:cause==tDCoverall-expiry))

[0128] Or

[0129] ((time(SGNB_RECFG_CMP)-time(SGNB_ADD_REQ_ACK)>20000and(SGNB_REL_REQUIRED:cause==tDCoverall-expiry)).

[0130] Before constructing the method of fault knowledge graph, it is necessary to first establish the correspondence between the problem template and the atomic process. It is introduced below. In an optional embodiment, the method of constructing the fault knowledge graph in the present application also includes: extracting the atomic flow chart from the protocol document; creating an atomic flow chart based on the atomic flow chart of the protocol document, and obtaining a set of fault description sentences from the fault document; obtaining the problem template corresponding to each fault description sentence in the set of fault description sentences; when all the signaling of the target problem template belongs to the signaling set of the target atomic process, the target problem template and the atomic process identifier of the target atomic process are associated. In this way, the correspondence between the problem template and the atomic process can be established, and based on the correspondence, the problem template corresponding to the atomic process can be automatically found.

[0131] In this embodiment, the atomic process graph includes nodes and edges. Nodes include signaling attributes and process attributes. Signaling attributes include at least one of a signaling sender identifier, a signaling receiver identifier, a signaling name, a port type, and a signaling sequence number. Process attributes include an atomic process identifier and a signaling set corresponding to the atomic process. Process attributes may also include an atomic process type. Edges are used to indicate the direction of signaling transmission. The fault description sentence set includes multiple fault description sentences.

[0132] The target question template is any question template. In one example, <time>The question template and question template identifier of REQ-ACK are shown in Figure 6. The question template identifier is <time>REQ-ACK, its corresponding question template includes [send / send / initiate] <sig:req>Pass <time>[Not received / Not received / Not received] <sig:ack>, and [receive / receive / receive] <sig:req>To [Reply / Feedback] <sig:ack>Delay between [less than / no more than / greater than / exceeds] <time>The words in [] are optional keywords, and the words in angle brackets <> are named entities. Signaling entities can also have types, such as <sig:req>Represents a REQ signaling entity.

[0133] If a certain type of atomic process contains all the signaling types involved in the fault, it is considered that the atomic process may have the fault, and the fault template name is added to the attribute information of the atomic process. For example, the atomic process of type "REQ-ACK-REJ" indicates that it contains three types of signaling: REQ, ACK, and REJ. The fault template <time>REQ-ACK involves two types of signaling, REQ and ACK, so <time>REQ-ACK is added to the fault template set associated with this type of atomic process.

[0134] The following describes the method for constructing a fault knowledge graph in the present application. Referring to FIG7 , an embodiment of the method for constructing a fault knowledge graph in the present application includes:

[0135] Step 701: Obtain sentences to be processed according to the fault identification text.

[0136] Specifically, step 701 includes: when the sentence of the fault judgment text does not include logical operation keywords, the sentence of the fault judgment text is used as a sentence to be processed; when the sentence of the fault judgment text includes logical operation keywords, multiple subjects in the sentence of the fault judgment text are determined according to the logical operation keywords, the text obtained after removing the logical operation keywords and multiple subjects from the sentence of the fault judgment text is used as a public expression text, and a sentence to be processed is generated according to the public expression text and each subject. The subject includes but is not limited to the signaling entity. After the sentence is reorganized in this way, the inconsistency between the signaling of the sentence to be processed and the signaling of the atomic process is reduced, which can improve the recognition success rate of the atomic process.

[0137] The logic operation keywords include at least one of the following information: or, either, and, and. When a sentence in the fault identification text includes a logic operation keyword, it indicates that the sentence can be split and reorganized into multiple sentences.

[0138] Referring to Figure 8, in one example, a base station sends an NGAP_PDU_SESSION_RSRC_SETUP_RSP or NGAP_PDU_SESSION_RSRC_MODIFY_RSP message carrying the cause information element "radio-connection-with-ue-lost21." "OR" is a logical operation keyword, and the entities it connects include NGAP_PDU_SESSION_RSRC_SETUP_RSP and NGAP_PDU_SESSION_RSRC_MODIFY_RSP. The public statement includes the following: The base station sends a message carrying the cause information element "radio-connection-with-ue-lost21." Reconstructed Sentences 1 and 2 are generated based on the entities and the public statement.

[0139] Reorganize sentence 1 as follows: The base station sends the NGAP_PDU_SESSION_RSRC_SETUP_RSP message carrying the cause information element radio-connection-with-ue-lost21.

[0140] Reorganized sentence 2 is: The base station sends the NGAP_PDU_SESSION_RSRC_MODIFY_RSP message carrying the cause information element radio-connection-with-ue-lost21.

[0141] Referring to FIG9 , in another example, the fault determination text includes:

[0142] Judgment method: One of the following two conditions is met:

[0143] 1. Check 5G call statistics and check if N.QosFlow.FailEst.UeNoReply is greater than 0.

[0144] 2. Check NG port tracking, and one of the following conditions is met:

[0145] (1) The base station sends an NGAP_PDU_SESSION_RSRC_SETUP_RSP or NGAP_PDU_SESSION_RSRC_MODIFY_RSP message carrying the cause information element as radio-connection-with-ue-lost21;

[0146] (2) After the base station receives the NGAP_INIT_CONTEXT_SETUP_REQ message carrying the PDUSessionID information element, it sends an NGAP_INIT_CONTEXT_SETUP_FAIL message carrying the cause information element of radio-connection-with-ue-lost21.

[0147] According to the hierarchical identifiers and logical relationship words, the following reorganized sentences can be generated:

[0148] s1: Check 5G traffic statistics, N.QosFlow.FailEst.UeNoReply>0;

[0149] s2: The base station sends an NGAP_PDU_SESSION_RSRC_SETUP_RSP message carrying the cause information element radio-connection-with-ue-lost21;

[0150] s3: The base station sends an NGAP_PDU_SESSION_RSRC_MODIFY_RSP message carrying the cause information element as radio-connection-with-ue-lost21;

[0151] s4: The base station receives the NGAP_INIT_CONTEXT_SETUP_REQ message carrying the PDUSessionID information element, and sends the NGAP_INIT_CONTEXT_SETUP_FAIL message carrying the cause information element as radio-connection-with-ue-lost21.

[0152] It should be understood that before step 701 , text preprocessing may be performed to remove illegal characters or redundant spaces.

[0153] Step 702: Obtain the signaling character string of the sentence to be processed.

[0154] Optionally, step 702 includes: when the sentence to be processed includes a keyword from a preset vocabulary, replacing the keyword from the preset vocabulary with an English character string corresponding to the keyword; extracting at least one candidate character string from the replaced sentence, and selecting a candidate character string including a preset signaling type as the signaling character string for the sentence to be processed. This allows the signaling character string for the sentence to be processed to be filtered based on the character string length and the preset signaling type. It should be understood that replacing the keywords from the preset vocabulary with English character strings and filtering the candidate character strings based on character string length are optional steps. Replacing the keywords from the preset vocabulary with English character strings can increase the number of recognizable signaling character strings, while filtering the candidate character strings based on character string length can reduce the number of non-signaling character strings.

[0155] The length of the candidate character string is greater than k, where k can be, but is not limited to, 2. The keywords in the preset vocabulary are Chinese descriptions of signaling. For example, the keyword "SCG add reconfiguration" corresponds to the English string "RRC_CONN_RECFG." The preset signaling types include, but are not limited to, ACK, REQ, REL, or SGNB.

[0156] Optionally, before selecting a candidate string including a preset signaling type as the signaling string for the sentence to be processed, a target candidate string is removed from at least one candidate string. The target candidate string is a string containing a "-", a string beginning with a number, a system name, a network element name, or a port name. This allows removing non-signaling strings before selecting the signaling string, thereby reducing the computational overhead of selecting the signaling string.

[0157] Step 703: Determine the protocol signaling of the sentence to be processed from the atomic flow graph according to the signaling character string of the sentence to be processed.

[0158] Optionally, step 703 includes: selecting a set of candidate protocol signalings from the atomic flow chart according to the signaling type of the signaling string of the sentence to be processed, calculating the fuzzy matching score corresponding to the candidate protocol signaling according to the length of the signaling string, the length of the candidate protocol signaling, and the edit distance between the signaling string and the candidate protocol signaling; and determining the protocol signaling of the sentence to be processed according to the candidate protocol signaling corresponding to the highest fuzzy matching score.

[0159] The edit distance, also known as the Levenshtein distance, describes the minimum number of operations required to convert one string into another, including insertion, deletion, and substitution. The edit distance between the signaling string and the candidate protocol signaling is denoted as ldist(p1, p2), the fuzzy match score is denoted as f(p1, p2), the length of the signaling string is denoted as len(p1), and the length of the candidate protocol signaling is denoted as len(p2). These satisfy the following formula:

[0160] According to this formula, the fuzzy matching score ranges from 0 to 1. A higher fuzzy matching score indicates a greater similarity between the signaling string and the protocol signaling. Because signaling strings often contain irregularities such as abbreviations, typos, and abbreviations, the fuzzy matching score can accurately identify the protocol signaling corresponding to the signaling string, significantly improving signaling recognition accuracy.

[0161] When there are multiple candidate protocol signalings corresponding to the highest fuzzy matching score, one candidate protocol signaling can be selected from the candidate protocol signalings with the highest fuzzy matching score according to a preset matching rule as the protocol signaling of the sentence to be processed.

[0162] The preset matching rules include at least one of the following: the atomic process to which the signaling string belongs is the same as the atomic process to which the candidate protocol signaling belongs, the signaling type of the signaling string and the signaling type of the candidate protocol signaling respond to each other, and the port type of the signaling string and the port type of the candidate protocol signaling are the same.

[0163] If the signaling string and the candidate protocol signaling meet any of the preset matching rules, the match score can be increased by 1. The more rules the signaling string and the candidate protocol signaling meet, the higher the match score. If there are multiple candidate protocol signalings corresponding to the highest fuzzy match score, the accuracy of signaling recognition can be further improved according to the preset matching rules.

[0164] Alternatively, step 703 includes determining the protocol signaling of the signaling to be processed from the atomic flow map based on the signaling string of the sentence to be processed, including: when a previous signaling string of the signaling string of the sentence to be processed matches the first target protocol signaling, determining the protocol signaling of the signaling to be processed as the next protocol signaling of the first target protocol signaling in the atomic flow to which the first target protocol signaling belongs. When the previous signaling string successfully matches the protocol signaling, the protocol signaling of the signaling to be processed can be determined, thereby improving the success rate of signaling identification.

[0165] For example, if the signaling string is "ACK," since ACK is not a protocol signaling, its corresponding protocol signaling cannot be directly determined. Obtain the signaling string preceding ACK in the fault judgment text. For example, the corresponding protocol signaling string is REQ signaling. The signaling string following REQ signaling in the atomic process is used as the protocol signaling corresponding to "ACK."

[0166] For example, the fault diagnosis text includes "5G immediately responded to the SNB ADD ack message. If no 4G response message is received after 20 seconds, an SNB rel message is sent with the reason value tDCoverall-expiry." The signaling string is "response message." Because the "response message" is not a protocol signaling, its corresponding protocol signaling cannot be directly determined. Obtain the signaling string preceding the "response message" in the fault diagnosis text. For example, the corresponding protocol signaling is SGNB_ADD_REQ_ACK. The signaling following SGNB_ADD_REQ_ACK is used as the protocol signaling corresponding to the "response message."

[0167] In another optional embodiment, step 703 includes: determining the protocol signaling of the signaling to be processed from the atomic flow map based on the signaling string of the sentence to be processed, including: when a subsequent signaling string of the signaling string of the sentence to be processed matches the second target protocol signaling, determining that the protocol signaling of the sentence to be processed is the protocol signaling preceding the second target protocol signaling in the atomic flow to which the second target protocol signaling belongs. When the subsequent signaling string successfully matches the protocol signaling, the protocol signaling of the signaling to be processed can be determined, thereby improving the success rate of signaling identification.

[0168] When the previous signaling string fails to match the protocol signaling of the atomic flow map, and / or when the subsequent signaling string fails to match the protocol signaling of the atomic flow map, the target signaling string is searched in the text after the signaling string of the sentence to be processed according to the signaling type of the signaling string of the sentence to be processed; when the target signaling string matches the third target protocol signaling, the protocol signaling of the sentence to be processed is determined to be the third target protocol signaling. This can improve the success rate of finding the protocol signaling. The signaling type of the target signaling string is the same as the signaling type of the signaling string of the sentence to be processed. For example, the signaling type of the signaling string is ACK, and the signaling type of its subsequent signaling is SGNB ACK. According to SGNB ACK, SGNB ADDITION REQUEST ACKNOWLEDGE is found in the atomic process, and the protocol signaling corresponding to ACK is determined to be SGNB ADDITION REQUEST ACKNOWLEDGE.

[0169] It should be noted that when the signaling type of the signaling string is different from the signaling type of the atomic flow chart, the fuzzy matching score between the signaling string and each protocol signaling can also be calculated, and the protocol signaling corresponding to the signaling string can be determined based on the maximum fuzzy matching score. The protocol signaling of the sentence to be processed refers to the protocol signaling corresponding to the signaling string of the sentence to be processed.

[0170] Step 704: Determine the atomic process to which the protocol signaling of the sentence to be processed belongs from the atomic process map.

[0171] Step 705: Generate at least one atomic process clause according to the atomic process and the sentence to be processed.

[0172] When multiple consecutive signalings belong to an atomic flow, the sentence including the multiple signalings is regarded as an atomic flow clause.

[0173] Step 706: Determine the question template corresponding to the atomic process clause.

[0174] Optionally, step 706 includes: replacing the signaling string and non-signaling entity in the atomic process clause with a mask string to obtain a query statement; obtaining a set of candidate question templates corresponding to the atomic process; selecting a subset of candidate question templates from the candidate question template set based on the signaling quantity of the query statement, and for each candidate question template in the candidate question template subset, calculating a fuzzy matching score between the candidate question template and the query statement; and determining the question template corresponding to the atomic process clause as the candidate question template corresponding to the highest matching score. The signaling quantity of each candidate question template in the candidate question template subset is the same as the signaling quantity of the query statement, and the mask string of the query statement is the same as the signaling string corresponding to the question template.

[0175] Refer to Figure 10. In one example, the atomic process clause is: 4G initiates SGN req and does not receive ACK after 5 seconds. The corresponding query statement is: 4G initiates <sig:req>Later <time>Not received <sig:ack>.

[0176] Specifically, the following steps are performed for the candidate question template set:

[0177] Step A: Compare the signaling quantity of the candidate question template with the signaling quantity of the query statement to see if they are the same. If they are different, the template matching fails. If they are the same, execute step B.

[0178] Step B: Obtain the candidate template statement corresponding to each candidate question template.

[0179] In another example, the candidate question template is [send / send] <sig:req>Pass <time>[Not received / Not received] <sig:ack>, the four candidate template statements corresponding to the problem template are respectively:

[0180] The first candidate template statement is: [Send] <sig:req>Pass <time>[Not Received] <sig:ack>。

[0181] The candidate template statement 2 is: [Send] <sig:req>Pass <time>[Not received] <sig:ack>.

[0182] Candidate template statement 3 is: [Send] <sig:req>Pass <time>[Not Received] <sig:ack>.

[0183] Candidate template statement 4 is: [Send] <sig:req>Pass <time>[Not received] <sig:ack>.

[0184] Step C: Calculate the fuzzy matching score between each candidate template sentence and the query sentence. Optionally, the length of the candidate template sentence and the length of the query sentence satisfy the following calculation formula:

[0185] t′ i represents the i-th candidate template statement, and ss represents the query statement. g(t′ i ,ss) represents the fuzzy matching score between the candidate template statement and the query statement, len(t′ i ) represents the length of the i-th candidate template sentence, len(ss) represents the length of the query sentence, ldist(t′ i ,ss) represents the edit distance between the candidate template statement and the query statement.

[0186] Step D: Obtain the maximum value of the fuzzy matching score.

[0187] t′ represents the set of candidate template statements.

[0188] Step E: The question template corresponding to the maximum fuzzy matching score is used as the question template corresponding to the atomic process clause.

[0189] After fuzzy matching, a higher fuzzy matching score indicates a higher accuracy of the found question template.

[0190] Step 707: Generate a logical expression of the atomic process clause according to the logical expression template corresponding to the question template, the protocol signaling corresponding to the atomic process clause, and the non-signaling entity of the atomic process clause.

[0191] Optionally, step 707 includes: filling the protocol signaling corresponding to the atomic process clause and the non-signaling entity of the atomic process clause into the slot of the logical expression template corresponding to the question template to obtain the logical expression of the atomic process clause.

[0192] Refer to Figure 11. In one example, the atomic process clause is: 4G initiates SGN req but does not receive ACK after 5 seconds. The problem template is [send / send / initiate] <sig:req>Pass <time>[Not received / Not received / Not received] <sig:ack>The logical expression template corresponding to this problem template is ((time(?)-(time(?))>?), where each ? corresponds to a slot. The logical expression corresponding to the atomic process clause is (time(SGNB_ADD_REQ_ACK)-(time(SGNB_ADD_REQ))>5000). The signaling entities include SGNB_ADD_REQ_ACK and SGNB_ADD_REQ, and the time value is 5000.

[0193] Step 708: Generate a logical expression of the sentence to be processed according to the logical expression of at least one atomic process clause.

[0194] When there are multiple atomic process clauses, the inter-sentence logical relationship of the multiple atomic process clauses is obtained, and the logical expressions of the multiple atomic process clauses are combined into the logical expression of the sentence to be processed according to the inter-sentence logical relationship of the multiple atomic process clauses. When the delimiter between the multiple atomic process clauses is a comma or a period, the inter-sentence logical relationship of the multiple atomic process clauses is and. When the delimiter between the multiple atomic process clauses includes a logical keyword (for example, and, or, and), the inter-sentence logical relationship of the multiple atomic process clauses corresponds to the logical keyword, the logical relationship corresponding to "and" and "and" is and, and the logical relationship corresponding to "or" is or.

[0195] Step 709: Generate a logical expression of the fault judgment text according to the logical expression of the sentence to be processed.

[0196] Specifically, step 709 includes: when there are multiple sentences to be processed, generating a logical expression of the fault determination text according to the logical relationship between all the sentences to be processed and the logical expressions of the sentences to be processed.

[0197] The inter-sentence logical relationship between any two sentences to be processed includes: and, or. The method for obtaining the inter-sentence logical relationship of the sentences to be processed includes: when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text do not include logical operation keywords, determining the inter-sentence logical relationship of all the sentences to be processed according to the hierarchical identifier; when the fault judgment text does not include a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determining the inter-sentence logical relationship of all the sentences to be processed according to the logical operation keywords; when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determining the inter-sentence logical relationship of all the sentences to be processed according to the hierarchical identifier and the logical operation keywords. When the fault judgment text includes sentences of a hierarchical structure and / or sentences of a parallel structure, the inter-sentence logical relationship of the sentences to be processed can be obtained, and the logical expressions of the sentences to be processed are combined into the logical expression of the fault judgment text according to the inter-sentence logical relationship of the sentences to be processed.

[0198] The fault diagnosis text shown in Figure 8 includes logical relations but does not include hierarchical identifiers, and the logical relation between sentences is 1||2. The fault diagnosis text shown in Figure 9 includes logical relations and hierarchical identifiers, and the logical relation between sentences is s1||((s2||s3||)s4).

[0199] Step 710: Add the logical expression of the fault judgment text to the fault knowledge graph.

[0200] In this embodiment, the corresponding protocol signaling and atomic process can be searched in the atomic process map based on the signaling string of the fault judgment text. Since the protocol signaling and atomic process are standardized objects, the success rate of identifying the signaling string can be improved, thereby improving the success rate of converting the fault judgment text into a logical expression.

[0201] Secondly, the existing method of generating logical expressions relies on grammatical structure analysis and is difficult to determine the logical relationships such as timing, time difference, inclusion, and intersection between parameters. The present application can pre-configure question templates and logical expression templates including the above logical relationships, and establish a correspondence between question templates and atomic processes. In this way, the corresponding question templates and logical expression templates can be found according to the atomic process, thereby generating logical expressions that can represent complex logical relationships.

[0202] Third, network failures often involve relationships between multiple parameters. Existing logical expressions are relational triples, which can only express binary relationships and cannot express logical relationships between more than three parameters. The logical expressions of this application include inter-sentence logical relationships between atomic process clauses and / or inter-sentence logical relationships between sentences to be processed, and can express hierarchical and parallel relationships. Therefore, the logical expressions of this application can represent more complex fault diagnosis methods.

[0203] In another optional embodiment, the method for constructing a fault knowledge graph in the present application also includes: receiving logs sent by network devices; searching for a target logical expression in the fault knowledge graph based on the signaling strings and non-signaling entities included in the logs; and judging whether the logs contain a fault event based on the output result of the target logical expression.

[0204] In this embodiment, after constructing a fault knowledge graph, fault detection can be automatically performed based on the fault knowledge graph, thereby improving fault detection speed. For example, if the output result of the target logical expression is true, the log is determined to contain a fault event; if the output result of the target logical expression is false, the log is determined to not contain a fault event. It should be understood that the correspondence between the output result of the logical expression and the fault event is not limited to the above example and can be set according to actual circumstances, and this application does not impose any restrictions.

[0205] The present application provides a computing device capable of implementing the method of constructing a fault knowledge graph of the present application. In one embodiment, the computing device 1200 includes:

[0206] An acquisition module 1201 is configured to acquire a sentence to be processed based on the fault identification text and acquire a signaling string of the sentence to be processed;

[0207] A first matching module 1202 is configured to determine the protocol signaling of the sentence to be processed from the atomic flow graph according to the signaling string of the sentence to be processed;

[0208] The second matching module 1203 is configured to determine the atomic process to which the protocol signaling of the sentence to be processed belongs from the atomic process map; generate at least one atomic process clause based on the atomic process and the sentence to be processed; and determine the question template corresponding to the atomic process clause.

[0209] Generation module 1204 is used to generate the logical expression of the atomic process clause based on the logical expression template corresponding to the problem template, the protocol signaling corresponding to the atomic process clause and the non-signaling entity of the atomic process clause; generate the logical expression of the sentence to be processed based on the logical expression of at least one atomic process clause; generate the logical expression of the fault judgment text based on the logical expression of the sentence to be processed; and add the logical expression of the fault judgment text to the fault knowledge graph.

[0210] In an optional embodiment, the acquisition module 1201 is specifically used to treat the sentences of the fault judgment text as sentences to be processed when the sentences of the fault judgment text do not include logical operation keywords; when the sentences of the fault judgment text include logical operation keywords, determine multiple subjects in the sentences of the fault judgment text according to the logical operation keywords; and generate a sentence to be processed based on the common expression text and each subject.

[0211] In another optional embodiment, the acquisition module 1201 is specifically used to replace the keywords in the preset vocabulary with English strings corresponding to the keywords when the sentence to be processed includes keywords in the preset vocabulary; extract at least one candidate string from the replaced sentence, and select the candidate string including the preset signaling type as the signaling string of the sentence to be processed.

[0212] Furthermore, in another optional embodiment, the acquisition module 1201 is further configured to remove a target candidate string from at least one candidate string, where the target candidate string is a string containing "-", a string starting with a number, a system name, a network element name, or a port name.

[0213] In another optional embodiment, the first matching module 1202 is specifically used to select a set of candidate protocol signalings from the atomic flow chart according to the signaling type of the signaling string of the sentence to be processed; calculate the fuzzy matching score corresponding to the candidate protocol signaling according to the length of the signaling string of the sentence to be processed, the length of the candidate protocol signaling, and the edit distance between the signaling string of the sentence to be processed and the candidate protocol signaling; and determine the protocol signaling of the sentence to be processed according to the candidate protocol signaling corresponding to the highest fuzzy matching score.

[0214] Furthermore, in another optional embodiment, the first matching module 1202 is specifically used to select a candidate protocol signaling from the candidate protocol signaling with the highest fuzzy matching score as the protocol signaling of the sentence to be processed according to a preset matching rule when there are multiple candidate protocol signalings corresponding to the highest fuzzy matching score.

[0215] In another optional embodiment, the first matching module 1202 is specifically used to determine that the protocol signaling of the signaling to be processed is the subsequent protocol signaling of the first target protocol signaling in the atomic process to which the first target protocol signaling belongs when the previous signaling string of the signaling string of the sentence to be processed matches the first target protocol signaling.

[0216] In another optional embodiment, the first matching module 1202 is specifically used to determine that the protocol signaling of the sentence to be processed is the previous protocol signaling of the second target protocol signaling in the atomic process to which the second target protocol signaling belongs when the latter signaling string of the signaling string of the sentence to be processed matches the second target protocol signaling.

[0217] In another possible implementation, the first matching module 1202 is specifically used to search for a target signaling string in the text following the signaling string of the sentence to be processed according to the signaling type of the signaling string of the sentence to be processed when the current signaling string fails to match the protocol signaling of the atomic flowchart, and / or when the subsequent signaling string fails to match the protocol signaling of the atomic flowchart, and when the target signaling string matches the third target protocol signaling, determine that the protocol signaling of the sentence to be processed is the third target protocol signaling.

[0218] In another optional embodiment, the second matching module 1203 is specifically used to replace the signaling strings and non-signaling entities in the atomic process clause with mask strings to obtain a query statement; obtain a set of candidate question templates corresponding to the atomic process; select a subset of candidate question templates from the candidate question template set according to the signaling quantity of the query statement, and for each candidate question template in the candidate question template subset, calculate the fuzzy matching score between the candidate question template and the query statement; determine that the question template corresponding to the atomic process clause is the candidate question template corresponding to the highest matching score.

[0219] In another optional embodiment, the generation module 1204 is specifically used to fill the protocol signaling corresponding to the atomic process clause and the non-signaling entity of the atomic process clause into the slot of the logical expression template corresponding to the problem template to obtain the logical expression of the atomic process clause.

[0220] In another optional embodiment, the generation module 1204 is specifically used to obtain the inter-sentence logical relationship of multiple atomic flow clauses when there are multiple atomic flow clauses; and generate the logical expression of the sentence to be processed based on the inter-sentence logical relationship of multiple atomic flow clauses and the logical expressions of multiple atomic flow clauses.

[0221] In another optional embodiment, the acquisition module 1201 is further used to determine the inter-sentence logical relationship of all sentences to be processed according to the hierarchical identifier when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text do not include logical operation keywords; when the fault judgment text does not include a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determine the inter-sentence logical relationship of all sentences to be processed according to the logical operation keywords; when the fault judgment text includes a hierarchical identifier and the sentences of the fault judgment text include logical operation keywords, determine the inter-sentence logical relationship of all sentences to be processed according to the hierarchical identifier and the logical operation keywords;

[0222] The generating module 1204 is specifically configured to generate a logical expression of the fault determination text according to the inter-sentence logical relations of all the sentences to be processed and the logical expressions of the sentences to be processed.

[0223] In another optional embodiment, the computing device 1200 of the present application also includes a first creation module, a second creation module and an association module; the first creation module is used to extract an atomic flow chart from a protocol document; create an atomic flow chart based on the atomic flow chart of the protocol document; the second creation module is used to obtain a set of fault description sentences from a fault document; obtain the problem template corresponding to each fault description sentence in the set of fault description sentences; the association module is used to associate the target problem template with the atomic flow identifier of the target atomic flow when all signaling of the target problem template belongs to the signaling set of the target atomic flow.

[0224] In another optional embodiment, the computing device 1200 of the present application also includes a receiving module and a fault detection module; the receiving module is used to receive logs sent by network equipment; the fault detection module is used to search for a target logical expression in the fault knowledge graph based on the signaling strings and non-signaling entities included in the log; and judge whether the log contains a fault event based on the output result of the target logical expression.

[0225] In order to realize the functions in the above-mentioned embodiments, the computing device includes hardware structures and / or software modules corresponding to each function. It should be easy for those skilled in the art to realize that, in combination with the units and method steps of each example described in the embodiments disclosed in this application, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application scenario and design constraints of the technical solution. The explanation of terms in Figure 12, the steps executed by each module and the beneficial effects can be found in the relevant description in the embodiment shown in Figure 7.

[0226] As shown in Figure 13, computing device 1300 includes a processor 1310 and an interface circuit 1320. Processor 1310 and interface circuit 1320 are coupled to each other. It will be appreciated that interface circuit 1320 may be a transceiver or an input / output interface. Optionally, computing device 1300 may also include a memory 1330 for storing instructions executed by processor 1310, input data required by processor 1310 to execute instructions, or data generated by processor 1310 after executing instructions.

[0227] When the computing device 1300 is used to implement the method shown in FIG. 7 , the processor 1310 is used to implement the functions of the acquisition unit 1201 , the first matching unit 1202 , the second matching unit 1203 , and the generation unit 1204 .

[0228] When the computing device is a chip used in a network device, the chip can perform the functions of the computing device in the above method embodiment. When the computing device is a module used in a base station, the base station module implements the functions of the computing device in the above method embodiment. The base station module here can be a baseband chip of the base station, or it can be a DU or other module. The DU here can be a DU in an open radio access network (O-RAN) architecture.

[0229] The processor in the embodiments of the present application may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0230] The method steps in the embodiments of the present application can be implemented in hardware or in software instructions that can be executed by a processor. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disk, mobile hard disk, CD-ROM or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. The storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC. In addition, the ASIC can be located in a base station or a terminal. The processor and storage medium can also exist in a base station or a terminal as discrete components.

[0231] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are performed in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, or other programmable device. The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video disk; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or nonvolatile storage medium, or may include both volatile and nonvolatile types of storage media.

[0232] In the various embodiments of the present application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0233] In this application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In the text description of this application, the character " / " generally indicates that the previous and next associated objects are in an "or" relationship; in the formula of this application, the character " / " indicates that the previous and next associated objects are in a "division" relationship. "Including at least one of A, B and C" can mean: including A; including B; including C; including A and B; including A and C; including B and C; including A, B and C.

[0234] It is understood that the various numbers used in the embodiments of this application are merely for ease of description and are not intended to limit the scope of the embodiments of this application. The order of the sequence numbers of the above-mentioned processes does not necessarily imply a specific order of execution; the order of execution of the processes should be determined by their functions and inherent logic.< / sig:ack> < / time> < / sig:req> < / sig:ack> < / time> < / sig:req> < / sig:ack> < / time> < / sig:req> < / sig:ack> < / time> < / sig:req> < / sig:ack> < / time> < / sig:req> < / sig:ack> < / time> < / sig:req> < / sig:ack> < / time> < / sig:req> < / time> < / time> < / sig:req> < / time> < / sig:ack> < / sig:req> < / sig:ack> < / time> < / sig:req> < / time> < / time> < / value> < / num> < / port> < / mode> < / time> < / sig:type> < / value> < / num> < / port> < / mode> < / time> < / sig:type> < / time> < / sig:ack> < / sig:req> < / sig:ack> < / time> < / sig:req>

Claims

1. A method for constructing a fault knowledge graph, characterized in that: The method is applied to a fault detection scenario in the field of wireless communications, and the method includes: Obtain sentences to be processed according to the fault identification text; Obtaining a signaling string of the sentence to be processed; Determining the protocol signaling of the sentence to be processed from the atomic flow chart according to the signaling character string of the sentence to be processed; Determine the atomic process to which the protocol signaling of the sentence to be processed belongs from the atomic process map; Generate at least one atomic process clause according to the atomic process and the sentence to be processed; Determine the problem template corresponding to the atomic process clause; Generate a logical expression of the atomic flow clause according to the logical expression template corresponding to the question template, the protocol signaling corresponding to the atomic flow clause, and the non-signaling entity of the atomic flow clause; Generate a logical expression of the sentence to be processed according to the logical expression of the at least one atomic process clause; Generate a logical expression of a fault determination text according to the logical expression of the sentence to be processed; The logical expression of the fault discrimination text is added to the fault knowledge graph.

2. The method according to claim 1, characterized in that: The method of obtaining the sentence to be processed according to the fault identification text includes: When the sentence of the fault determination text does not include a logical operation keyword, the sentence of the fault determination text is used as a sentence to be processed; When the sentence of the fault judgment text includes the logical operation keyword, multiple subjects in the sentence of the fault judgment text are determined according to the logical operation keyword; a sentence to be processed is generated according to the common expression text and each subject, and the common expression text is the text obtained by removing the logical operation keyword and the multiple subjects from the sentence of the fault judgment text.

3. The method according to claim 1, characterized in that The step of obtaining the signaling string of the sentence to be processed includes: When the sentence to be processed includes a keyword in a preset vocabulary, replacing the keyword in the preset vocabulary with an English character string corresponding to the keyword; Extracting at least one candidate character string from the replaced sentence, wherein the length of the candidate character string is greater than a preset character string length; A candidate character string including a preset signaling type is selected as a signaling character string of a sentence to be processed.

4. The method according to claim 3, characterized in that Before selecting a candidate character string including a preset signaling type as a signaling character string of a sentence to be processed, the method further includes: A target candidate character string is removed from the at least one candidate character string, where the target candidate character string is a character string including "-", a character string starting with a number, a system name, a network element name, or a port name.

5. The method according to claim 1, characterized in that The determining of the protocol signaling of the sentence to be processed from the atomic flow chart according to the signaling character string of the sentence to be processed comprises: Selecting a candidate protocol signaling set from the atomic flow graph according to the signaling type of the signaling character string of the sentence to be processed, wherein the candidate protocol signaling set includes at least one candidate protocol signaling; Calculate the fuzzy matching score corresponding to the candidate protocol signaling according to the length of the signaling string of the sentence to be processed, the length of the candidate protocol signaling, and the edit distance between the signaling string of the sentence to be processed and the candidate protocol signaling; The protocol signaling of the sentence to be processed is determined according to the candidate protocol signaling corresponding to the highest fuzzy matching score.

6. The method according to claim 5, characterized in that The method of determining the protocol signaling of the sentence to be processed according to the candidate protocol signaling corresponding to the highest fuzzy matching score includes: When there are multiple candidate protocol signalings corresponding to the highest fuzzy matching score, a candidate protocol signaling is selected from the candidate protocol signalings with the highest fuzzy matching score according to a preset matching rule as the protocol signaling of the sentence to be processed, and the preset matching rule includes at least one of the following rules: the atomic process to which the signaling string belongs is the same as the atomic process to which the candidate protocol signaling belongs, the signaling type of the signaling string and the signaling type of the candidate protocol signaling respond to each other, and the port type of the signaling string is the same as the port type of the candidate protocol signaling.

7. The method according to claim 1, characterized in that The step of determining the protocol signaling of the signaling to be processed from the atomic flow chart according to the signaling character string of the sentence to be processed comprises: When a previous signaling string of the signaling string of the sentence to be processed matches the first target protocol signaling, determining that the protocol signaling of the signaling to be processed is a subsequent protocol signaling of the first target protocol signaling in the atomic flow to which the first target protocol signaling belongs; or, When the subsequent signaling string of the signaling string of the sentence to be processed matches the second target protocol signaling, the protocol signaling of the sentence to be processed is determined to be the previous protocol signaling of the second target protocol signaling in the atomic process to which the second target protocol signaling belongs.

8. The method according to claim 7, characterized in that The method further comprises: When the previous signaling string fails to match the protocol signaling of the atomic flow map, and / or when the subsequent signaling string fails to match the protocol signaling of the atomic flow map, searching for a target signaling string in the target text according to the signaling type of the signaling string of the sentence to be processed, the target text is the text after the signaling string of the sentence to be processed; When the target signaling character string matches the third target protocol signaling, it is determined that the protocol signaling of the sentence to be processed is the third target protocol signaling.

9. The method according to claim 1, characterized in that: The problem template for determining the atomic process clause corresponding to the problem template includes: Replacing the signaling string and the non-signaling entity in the atomic flow clause with the mask string to obtain a query statement; Obtain a set of candidate question templates corresponding to the atomic process; Selecting a subset of candidate question templates from the candidate question template set according to the signaling quantity of the query statement, wherein the signaling quantity of each candidate question template in the subset of candidate question templates is the same as the signaling quantity of the query statement; For each candidate question template in the candidate question template subset, calculating a fuzzy matching score between the candidate question template and the query statement; The question template corresponding to the atomic process clause is determined to be the candidate question template corresponding to the highest matching score.

10. The method according to claim 1, characterized in that The step of generating the logical expression of the atomic flow clause according to the logical expression template corresponding to the question template, the protocol signaling corresponding to the atomic flow clause, and the non-signaling entity of the atomic flow clause comprises: The protocol signaling corresponding to the atomic flow clause and the non-signaling entities of the atomic flow clause are filled in the slots of the logical expression template corresponding to the question template to obtain the logical expression of the atomic flow clause.

11. The method according to claim 1, characterized in that: The step of generating the logical expression of the sentence to be processed according to the logical expression of the at least one atomic process clause comprises: When there are multiple atomic flow clauses, obtaining the logical relationship between the multiple atomic flow clauses; The logical expression of the sentence to be processed is generated according to the logical relationship between the multiple atomic flow clauses and the logical expressions of the multiple atomic flow clauses.

12. The method according to claim 1, characterized in that The method further comprises: When the fault identification text includes a level identifier and the sentences of the fault identification text do not include a logical operation keyword, determining the inter-sentence logical relationship of all sentences to be processed according to the level identifier; When the fault identification text does not include a level identifier and the sentences of the fault identification text include logic operation keywords, determining the inter-sentence logical relationship of all sentences to be processed according to the logic operation keywords; When the fault identification text includes a level identifier and the sentences of the fault identification text include logic operation keywords, determining the inter-sentence logical relationship of all sentences to be processed according to the level identifier and the logic operation keywords; The logical expression for generating the fault determination text according to the logical expression of the sentence to be processed includes: The logical expression of the fault determination text is generated according to the inter-sentence logical relations of all the sentences to be processed and the logical expressions of the sentences to be processed.

13. The method according to any one of claims 1 to 12, characterized in that The method further comprises: Extract atomic flow charts from protocol documents; Creating an atomic flow chart according to the atomic flow chart of the protocol document, wherein the atomic flow chart includes an atomic flow identifier and a signaling set corresponding to the atomic flow; Obtain a set of fault description sentences from the fault document; Obtaining a question template corresponding to each fault description sentence in the fault description sentence set; When all the signaling of the target problem template belongs to the signaling set corresponding to the target atomic flow, the target problem template is associated with the atomic flow identifier of the target atomic flow, and the target problem template is any one of the problem templates.

14. The method according to any one of claims 1 to 12, characterized in that The method further comprises: Receive logs sent by network devices; Searching for a target logical expression in the fault knowledge graph according to the signaling character strings and non-signaling entities included in the log; Whether the log contains a fault event is determined according to the output result of the target logical expression.

15. A computing device comprising means for executing the method of any one of claims 1 to 14.

16. A computing device, characterized in that: The invention comprises a processor and an interface circuit, wherein the interface circuit is used to receive signals from other computing devices and transmit them to the processor or send signals from the processor to other computing devices, and the processor is used to implement the method as claimed in any one of claims 1 to 14 through a logic circuit or executing code instructions.

17. A computer-readable storage medium, characterized in that: The storage medium stores a computer program or instruction, and when the computer program or instruction is executed by a computing device, the method according to any one of claims 1 to 14 is implemented.

18. A computer program product, characterized in that When the computer program product is run on a computer, the computer is caused to execute the method according to any one of claims 1 to 14.

Images