Digital system problem processing method and device

Abstract

The invention discloses a digital system problem processing method and device. The method comprises the following steps: performing intention recognition and entity recognition on a target question text input by a target terminal to obtain a corresponding target question intention and a plurality of key entities; a knowledge entry set identical to the target question intention is searched in a preset operation and maintenance knowledge graph through the reverse index, multi-dimensional attribute information of all knowledge entries in the knowledge entry set is searched through the forward index, and all the knowledge entries in the knowledge entry set comprise at least one key entity of the target question text; sorting a plurality of knowledge entries in the knowledge entry set by taking the frequency of the key entities in the inverted index and the multi-dimensional attribute information in the forward index as sorting factors; and feeding back the sorted knowledge item set to the target terminal. According to the method and the device, the technical problem of low problem solving efficiency caused by manually inquiring the solution of the operation and maintenance problems of the digital system is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data processing, in particular to a digital system problem processing method and device. BACKGROUND

[0002] In the daily operation and maintenance of a digital system, users and maintenance personnel face a series of challenges, which mainly result from the limitations of current technical solutions. For example, a traditional digital system operation manual exists in the form of static text. When a user encounters a specific operation problem, the user needs to search for a solution in the manual through keyword search. This method is not only inefficient, but also makes it difficult to capture semantic information, resulting in time-consuming and inaccurate problem positioning. In addition, for complex fault troubleshooting in system maintenance, existing technologies often rely on the experience of professional maintenance personnel, and lack a unified and intelligent decision support system. In particular, when dealing with cross-domain problems, cumbersome communication and cooperation between different professionals are required, which increases the maintenance cost and reduces the response speed.

[0003] At present, no effective solution has been proposed to solve the above problems. SUMMARY

[0004] The embodiments of the present application provide a digital system problem processing method and device to at least solve the technical problem that the solution to the operation and maintenance problems of a digital system is manually queried by a person, resulting in low problem solving efficiency.

[0005] According to an aspect of an embodiment of the present application, a digital system problem processing method is provided, including: performing intent recognition and entity recognition on a target problem text input by a target terminal to obtain a corresponding target question intent and a plurality of key entities, wherein the type of the target problem text at least includes an operation problem or a fault problem of a digital system; finding a knowledge item set same as the target question intent in a preset operation and maintenance knowledge graph through an inverted index, and finding multi-dimensional attribute information of each knowledge item in the knowledge item set through a forward index, wherein each knowledge item in the knowledge item set contains at least one key entity of the target problem text, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree, and a user satisfaction score; taking the frequency of the key entity in the inverted index and the multi-dimensional attribute information in the forward index as a sorting factor to sort a plurality of knowledge items in the knowledge item set; and feeding back the sorted knowledge item set to the target terminal.

[0006] Optionally, the target question text input by the target terminal is subjected to intent recognition and entity recognition to obtain a corresponding target question intent and a plurality of key entities, including: obtaining the target question text input by the target terminal; performing a first preprocessing operation on the target question text, wherein the first preprocessing operation at least includes: removing noise characters, unifying text formats, processing abbreviations, and synonym replacement; encoding the preprocessed target question text to obtain a corresponding text embedding vector, and analyzing the text embedding vector using an intent recognition model to obtain the probability that the target question text belongs to each of a plurality of preset question intents, and taking the question intent with the highest probability as the target question intent corresponding to the target question text, wherein the intent recognition model is obtained by training a deep learning model using a plurality of sets of training sample data, and each set of training sample data includes: a text embedding vector corresponding to a question text and a question intent corresponding to the question text, and the question intent includes at least one of the following: an operation guidance intent, a fault troubleshooting intent, a performance optimization intent, and a system state query intent; analyzing the preprocessed target question text using a named entity recognition model to obtain a plurality of key entities related to the system operation and maintenance field within the target question text.

[0007] Optionally, the construction process of the operation and maintenance knowledge graph includes: obtaining a plurality of multi-modal knowledge data related to the system operation and maintenance field, wherein the multi-modal knowledge data includes at least one of the following: system operation logs, system operation and maintenance documents, system operation manuals, system operation video tutorials, and system configuration parameters; for each multi-modal knowledge data, performing a second preprocessing operation on the multi-modal knowledge data, wherein the second preprocessing operation at least includes: converting image data into text data using optical character recognition technology, converting audio and video data into text data using automatic speech recognition technology, and performing a first preprocessing operation on the text data; analyzing the preprocessed multi-modal knowledge data using a named entity recognition model to obtain a plurality of entities related to the system operation and maintenance field within the multi-modal knowledge data, and extracting semantic relationships between the entities using a preset operation and maintenance field knowledge rule, wherein the plurality of entities at least include: system component state entities, business process entities, error code entities, exception type entities, technical parameter optimization entities, operation process entities, and operation and maintenance step entities; extracting semantic relationships between the entities using a preset operation and maintenance field knowledge rule; constructing an operation and maintenance knowledge graph by taking each entity as a graph node and taking the semantic relationships between the entities as edges, wherein the operation and maintenance knowledge graph includes a plurality of knowledge entries composed of a plurality of graph nodes and a plurality of associated edges, which are operation and maintenance knowledge containing questions and solution step processes or configuration parameters.

[0008] Optionally, the same knowledge item set as the target question intention is found in the preset operation and maintenance knowledge graph through the inverted index, including: finding a candidate knowledge item set related to the target question text in the operation and maintenance knowledge graph through the inverted index, wherein each initial knowledge item in the candidate knowledge item set includes at least one key entity of the target question text; constructing a structured query statement according to the target question intention, the plurality of key entities and the candidate knowledge item set, and querying the candidate knowledge item set using the structured query statement to obtain the knowledge item set.

[0009] Optionally, the structured query statement is constructed according to the target question intention, the plurality of key entities and the candidate knowledge item set, including: determining a target query statement template corresponding to the target question intention from a preset template library, wherein the template library at least includes: a first query statement template for realizing an operation guidance intention, a second query statement template for realizing a fault troubleshooting intention, and a third query statement template for realizing a performance optimization intention; embedding the plurality of key entities and the candidate knowledge item set into the target query statement template as query variables and query constraints to obtain the structured query statement.

[0010] Optionally, the sorted knowledge item set is fed back to the target terminal, including: determining the data content of each knowledge item in the sorted knowledge item set; in the case that the data content is a configuration parameter, feeding back the knowledge item to the target terminal in a text format; in the case that the data content is a solution step process, feeding back the knowledge item to the target terminal in a text format or a video format.

[0011] Optionally, after the sorted knowledge item set is fed back to the target terminal, the method further includes: obtaining a feedback evaluation of any knowledge item in the knowledge item set by the target terminal, wherein the feedback evaluation at least includes: a satisfaction score and an optimization suggestion; in the case that the satisfaction score of the knowledge item is not higher than a preset threshold, modifying the knowledge item in the operation and maintenance knowledge graph according to the optimization suggestion.

[0012] According to another aspect of the embodiments of the present application, a digital system problem processing apparatus is also provided, comprising: an identification module configured to perform intent recognition and entity recognition on a target question text input by a target terminal to obtain a target question intent and a plurality of key entities, wherein the type of the target question text at least includes an operation question or a fault question of a digital system; a query module configured to find a same knowledge item set as the target question intent in a preset operation and maintenance knowledge graph through an inverted index, and find multi-dimensional attribute information of each knowledge item in the knowledge item set through a forward index, wherein each knowledge item in the knowledge item set contains at least one key entity of the target question text, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree and a user satisfaction score; a sorting module configured to take the frequency of the key entity in the inverted index and the multi-dimensional attribute information in the forward index as sorting factors to sort the plurality of knowledge items in the knowledge item set; and a feedback module configured to feed back the sorted knowledge item set to the target terminal.

[0013] According to another aspect of the embodiments of the present application, a computer program product is also provided, comprising: a computer program, wherein the computer program is executed by a processor to implement the digital system problem processing method described above.

[0014] According to another aspect of the embodiments of the present application, an electronic device is also provided, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the digital system problem processing method described above through the computer program.

[0015] In the embodiments of the present application, the target question text input by the target terminal is subjected to intent recognition and entity recognition, the knowledge item set matched with the target question intent and the multi-dimensional attribute information of each knowledge item in the set are found in the preset operation and maintenance knowledge graph through the inverted index and the forward index, and the knowledge items are sorted by using the frequency of the key entity contained in the knowledge item and the multi-dimensional attribute information, so as to feed back the sorted knowledge item set to the target terminal. The technical effect that the most relevant and most credible operation and maintenance knowledge items are quickly and accurately retrieved and sorted according to the question text is achieved, the purposes of improving the problem solving efficiency, enhancing the credibility of the solution and improving the user operation experience are achieved, and the technical problem that the solution of the operation and maintenance problem of the digital system is manually queried by a person, resulting in low problem solving efficiency is solved. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are included to provide a further understanding of the present application, form a part of the present application and illustrate the illustrative embodiments of the present application and the explanation of the present application, and do not constitute improper limitations on the present application. In the drawings:

[0017] Figure 1 is a flow diagram of an optional digital system problem processing method according to an embodiment of the present application;

[0018] Figure 2 is a structural diagram of an optional digital system problem processing device according to an embodiment of the present application;

[0019] Figure 3 is a structural diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the personnel in the art better understand the scheme of the present application, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor should belong to the protection scope of the present application.

[0021] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] Embodiment 1

[0023] According to an embodiment of the present application, a digital system problem processing method is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that herein.

[0024] Figure 1 is a flow diagram of a digital system problem processing method according to an embodiment of the present application, as shown in Figure 1 the method comprises the following steps:

[0025] Step S102, performing intent recognition and entity recognition on the target question text input by the target terminal to obtain a corresponding target question intent and a plurality of key entities. The type of the target question text at least includes an operation question or a fault question of a digital system.

[0026] Step S104, finding a knowledge item set same as the target question intent in the preset operation and maintenance knowledge graph through inverted indexing, and finding multi-dimensional attribute information of each knowledge item in the knowledge item set through forward indexing. At least one key entity of the target question text is contained in each knowledge item in the knowledge item set, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree, and a user satisfaction score.

[0027] Step S106, taking the frequency of the key entity in the inverted indexing and the multi-dimensional attribute information in the forward indexing as a sorting factor to sort a plurality of knowledge items in the knowledge item set.

[0028] Step S108, feeding back the sorted knowledge item set to the target terminal.

[0029] Based on the scheme defined in the above steps S102 to S108, it can be known that in the embodiments of the present application, by performing intent recognition and entity recognition on the target question text input by the target terminal, the knowledge item set matching the target question intent and the multi-dimensional attribute information of each knowledge item in the set are found in the preset operation and maintenance knowledge graph through inverted indexing and forward indexing, and the knowledge items are sorted by using the frequency of the key entity contained in the knowledge item and the multi-dimensional attribute information, so as to feed back the sorted knowledge item set to the target terminal. The technical effect of quickly and accurately retrieving and sorting the most relevant and most reliable operation and maintenance knowledge items according to the question text is achieved, and the purposes of improving the problem solving efficiency, enhancing the reliability of the solution, and improving the user operation experience are achieved.

[0030] The steps of the digital system question processing method will be described in combination with a specific implementation process.

[0031] As an optional implementation, in the technical scheme provided in the above step S102, the method can include:

[0032] First step: obtaining the target question text input by the target terminal. The target terminal can be a desktop computer, a mobile device, a special terminal, etc. Therefore, the user inputs the operation question or system failure description, such as "prompt password error when logging in" and "how to export a report", through the application interface on the terminal, such as the customer service system, the operation assistance software, and the intelligent operation and maintenance platform, in the form of text box and dialogue window. Alternatively, the intelligent assistance system can automatically collect running alarm texts, such as "CPU utilization > 85%" and "API call returns 404 error".

[0033] Second step: performing first preprocessing operation on the target question text to ensure the data quality for subsequent analysis.

[0034] The first preprocessing operation includes but is not limited to the following: removing noise characters such as punctuation marks and special symbols to ensure the purity and readability of the text; unifying the text format, such as converting the text to lowercase, removing extra spaces, and ensuring the consistency of the text; processing abbreviations by expanding them into complete words through a vocabulary table to improve the accuracy of understanding and matching; using synonym replacement to enhance the diversity and accuracy of semantic expression. The synonym replacement can be achieved by looking up a synonym dictionary or using a pre-trained word embedding model.

[0035] Third step: encoding the preprocessed target question text using a deep learning model (such as BERT model) to obtain the corresponding text embedding vector, and then analyzing the text embedding vector using an intent recognition model to obtain the probability of the target question text belonging to each of the pre-set question intents, and taking the question intent with the highest probability as the target question intent corresponding to the target question text.

[0036] The intent recognition model is obtained by training a deep learning model using multiple sets of training sample data, and each set of training sample data includes the text embedding vector corresponding to the question text and the question intent corresponding to the question text. The question intent includes at least one of the following: operation guidance intent, fault troubleshooting intent, performance optimization intent, and system status query intent.

[0037] Fourth step: analyzing the preprocessed target question text using a named entity recognition model to obtain multiple key entities related to system operation and maintenance in the target question text.

[0038] The named entity recognition model is obtained by training a deep learning model using a large number of system operation and maintenance field text samples annotated with entity information. The deep learning model includes but is not limited to BERT model, BERT+Conditional Random Field (CRF) model, etc.

[0039] For example, if the target question text is "Login prompts password error", the above-mentioned intention recognition model can analyze the text embedding vector corresponding to the question text to obtain the corresponding target question intention "operation guidance intention". This is because the question text mainly asks how to correctly perform the login operation to solve the password error prompt, so it belongs to the guidance demand of the user operation level. At the same time, the above-mentioned named entity recognition model can analyze the preprocessed target question text to obtain the following two key entities: the business process entity is the user login process, and the abnormal type entity is the password error.

[0040] In addition, if the target question text is "The CPU utilization rate of the database server is > 85%", the above-mentioned intention recognition model can analyze the text embedding vector corresponding to the question text to obtain the corresponding target question intention "troubleshooting intention" or "performance optimization intention". This is because the question text mentions that the system performance index exceeds the normal range, which implies that the user or the system wants to know the reason for the high CPU utilization rate and seeks optimization or troubleshooting suggestions. At the same time, the above-mentioned named entity recognition model can analyze the preprocessed target question text to obtain the following two key entities: the system component state entity is the CPU utilization rate of the database server, and the abnormal type entity is the high CPU utilization rate.

[0041] Further, the intelligent auxiliary system can pre-build a structured and highly correlated knowledge base, namely an operation and maintenance knowledge graph, and based on this, the intelligent auxiliary system can provide more accurate, efficient and deep understanding services.

[0042] Optionally, regarding the operation and maintenance knowledge graph, the construction process includes:

[0043] First step: Obtain a plurality of multi-modal knowledge data related to the field of system operation and maintenance. The multi-modal knowledge data includes at least one of the following: system operation log, system operation and maintenance document, system operation manual, system operation video tutorial, and system configuration parameter. These data sources are diverse and can comprehensively cover all aspects of system operation and maintenance.

[0044] Second step: For each multi-modal knowledge data, perform a second preprocessing operation on the multi-modal knowledge data to ensure that the format of all multi-modal knowledge data is uniform, facilitating subsequent analysis and processing. The second preprocessing operation at least includes: converting image data into text data using optical character recognition technology, converting audio and video data into text data using automatic speech recognition technology, and performing a first preprocessing operation on the text data.

[0045] Third step: using a named entity recognition model to analyze the preprocessed multi-modal knowledge data, obtaining a plurality of entities related to the system operation and maintenance field in the multi-modal knowledge data, and using a preset operation and maintenance field knowledge rule to extract semantic relationships between each entity.

[0046] Among them, the plurality of entities at least includes: system component state entity, business process entity, exception type entity, error code entity, technical parameter optimization entity, operation process entity, operation and maintenance step entity, etc. In addition, the extraction of semantic relationships can be assisted by dependency syntax analysis, cross-document relationship extraction, template matching and other technologies to ensure that the constructed graph relationship is clear and logically coherent, and the semantic relationship between the extracted entities can include but is not limited to: influence relationship, dependency relationship, causal relationship, operation relationship, association relationship, etc.

[0047] Fourth step: taking each entity as a graph node and taking the semantic relationship between each entity as an edge to construct an operation and maintenance knowledge graph, wherein the plurality of knowledge entries included in the operation and maintenance knowledge graph are operation and maintenance knowledge containing problems and solution step processes or configuration parameters composed of a plurality of graph nodes and a plurality of associated edges, and each knowledge entry has a corresponding entry ID.

[0048] Therefore, the intelligent auxiliary system uses the semantic understanding ability of the large language model to construct a multi-modal operation and maintenance knowledge base integrating text, image, video and other types of data. This integration enables the system to handle more complex query scenarios and improve the comprehensiveness and practicality of the solution.

[0049] For example, multi-modal knowledge data is collected from various operation and maintenance materials and running logs of enterprise resource planning software (ERP), including: system operation manual PDF files that detail the use methods and common operation problems of ERP modules; system operation and maintenance documents covering troubleshooting steps, system upgrade guidelines, performance optimization suggestions, etc.; operation and maintenance related video tutorials demonstrating the implementation process of specific operations and emergency response drills; real-time system logs recording information about the system running state of the digital system, including system component status, business process execution, technical parameter settings, etc.; configuration parameter files containing various parameters set during system initialization, such as database connection information, network port configuration, etc.; error code library listing all possible system exception error codes and their meanings and common solutions.

[0050] Next, the multi-step preprocessing operation is performed on the collected multi-modal knowledge data, including: using optical character recognition technology to convert the system operation manual PDF file into parseable text data, facilitating subsequent text analysis; using automatic speech recognition technology to convert the audio dialogue in the operation and maintenance related video tutorial into a text script, ensuring that the video content can be processed by text; performing first preprocessing operations such as standardization, cleaning, and word segmentation on the text data to eliminate data noise and unify data formats.

[0051] Then, the preprocessed multi-modal knowledge data is subjected to entity recognition using a named entity recognition model, obtaining a plurality of entities related to the system operation and maintenance field, such as: system component state entities, including real-time state information of database servers, application servers, load balancers, storage systems, network devices, etc., and the attribute information of these entities includes but is not limited to CPU utilization, memory usage, disk space, connection number, etc. index information; business process entities, including user login process, report generation process, data backup process, etc., and the attribute information of these entities includes system components associated with each node of the process, etc.; exception type entities, including system component performance index exceptions, process exceptions, etc., and the attribute information of these entities includes exception type, exception occurrence time, exception impact range, etc.; error code entities, including ERROR_404, WARNING_503, CRITICAL_808, etc., and the attributes of these entities include specific description information of the error code; technical parameter optimization entities, including adjusting database cache size, optimizing network configuration, improving thread pool size, etc., and the attributes of these entities contain technical parameter optimization suggestions and specific operation guidance information; operation process entities, including report export process, user permission setting process, system upgrade process, etc., and the attribute of these entities describes the specific steps and operation interface indications of the operation process; maintenance step entities, including restarting services, replacing hard disks, increasing network bandwidth, adjusting parameter configurations, etc., and the attribute information of these entities contains specific operation steps in the maintenance process.

[0052] Subsequently, the system extracts semantic relationships from the relationships between entities according to the preset system operation and operation and maintenance field knowledge, and establishes the connection between entities. Specifically, there are dependency relationships and operation relationships between system component state entities and other entities. For example, when the state indicator of the "database server" is abnormal, specific operation steps (such as restarting the service and increasing the database cache) may need to be performed to adjust its state, which reflects the operation relationship. There is an association relationship between the business process entity and the exception type entity, and a causal relationship between the technical parameter optimization entity, the operation process entity, and the operation step entity. For example, when the disk space of the database server is insufficient, it will cause the report generation process to respond slowly or fail, and then there is a causal relationship between the "abnormal report generation process" and the "disk space state of the database server". There is also a causal relationship between the error code entity and the system component state entity and the business process entity. For example, the error code ERROR_404 (not found resource) may be caused by the parameter configuration error of the application server, and then there is a causal relationship between "the real-time state information of the application server" and "ERROR_404". There is an operation relationship between the technical parameter optimization entity and the technical parameter entity, wherein the technical parameter entity includes the maximum number of database connections, server port configuration, network bandwidth, and application thread number, and the technical parameter optimization entity can have a causal relationship with the system component state entity, the business process entity, and the error code entity. There is an operation relationship between the operation process entity and the system component state entity, that is, the influence of a specific operation on the system component state. The operation process entity can also have an association with the operation step entity, that is, the operation step that needs to be executed under a specific operation process to support or guarantee the normal operation of the process. There is an operation relationship between the operation step entity and the system component state entity, the business process entity, and the technical parameter entity, that is, by executing a specific operation step to adjust the system component state, repair the business process exception, and optimize the technical parameter.

[0053] Finally, the above-identified various entities are taken as graph nodes, and the semantic relationship between each entity is taken as an edge, to construct an operation and maintenance knowledge graph of a graph data structure, wherein each knowledge entry in the knowledge graph is actually a subgraph structure containing multiple graph nodes (entities) and multiple edges (entity relationship). Such a subgraph can completely describe a problem and its solution step flow or configuration parameters. For example, when the problem text is "the CPU utilization of the database server exceeds 85%", the question intention and key entities of the text can retrieve the following subgraph from the knowledge graph: starting from the "CPU utilization of the database server" node, connecting to the "database performance degradation" node through the "influence relationship" edge, connecting to the "database connection number is too high" node through the "causal relationship" edge, and finally connecting to the "adjust the database connection number configuration" node through the "operation relationship" edge. This subgraph constitutes a complete knowledge entry, which describes the cause, influence and solution steps of the problem.

[0054] On the basis of the above operation and maintenance knowledge graph, the intelligent auxiliary system can use an inverted index to find a knowledge entry set identical to the target question intention in the preset operation and maintenance knowledge graph. The inverted index is a high-efficiency data retrieval mechanism, which is different from the way of retrieving entities according to document identification information in the forward index. The inverted index is a reverse record of key entities appearing in which knowledge entries, so that all knowledge entries containing the entity can be quickly located according to the key entity. Therefore, the system first locates all knowledge entries containing each key entity through the inverted index, and then determines which knowledge entry set is most matched with the target question intention, to finally obtain the knowledge entry set identical to the target question intention, and each knowledge entry in the knowledge entry set contains at least one key entity of the target problem text.

[0055] Optionally, the intelligent auxiliary system can find the knowledge entry set identical to the target question intention from the above operation and maintenance knowledge graph according to the following steps, including:

[0056] First step: find a candidate knowledge entry set related to the target problem text in the operation and maintenance knowledge graph through the inverted index.

[0057] In the context of operating and maintaining the knowledge graph, the inverted index records the mapping relationship between the entities contained in the knowledge entries and the entry IDs, rather than the mapping of the traditional document IDs and the list of words. The purpose of this is to quickly locate all knowledge entries that mention at least one key entity when receiving the target problem text, ensuring that the inverted index becomes a quick query bridge from entities to knowledge entries. Therefore, the intelligent auxiliary system can use the pre-constructed inverted index to query the knowledge entries related to the key entities, so that querying any single key entity or key entity combination can directly find the knowledge entries containing them, avoiding traversal or full-text search in the entire knowledge graph, greatly reducing the computational complexity and response time. Through the query of the inverted index, the intelligent auxiliary system obtains a candidate knowledge entry set, and each initial knowledge entry in the candidate knowledge entry set includes at least one key entity of the target problem text.

[0058] Since the inverted index can quickly generate a candidate set, but it may contain some entries that do not completely match the target question intent, the intelligent auxiliary system can further optimize and filter the candidate knowledge entry set according to the relationship between the target question intent and the key entities, to ensure that the knowledge entry set returned to the user is the most closely matched to the user's problem requirements, and also excludes those entries that contain key entities but are irrelevant or have low relevance to the problem.

[0059] Second step: constructing a structured query statement according to the target question intent, the plurality of key entities, and the candidate knowledge entry set, and querying the candidate knowledge entry set using the structured query statement to obtain the knowledge entry set.

[0060] Specifically, the structured query statement can be constructed according to the following steps: first, determine the target query statement template corresponding to the target question intent from the pre-set template library, wherein the template library includes a plurality of query statement templates, and each query statement template is classified and customized for different question intents. Specifically, the template library at least includes the following three types of query statement templates: a first query statement template for implementing an operation guidance intent (such as querying specific operation steps, etc.), a second query statement template for implementing a troubleshooting intent (such as querying troubleshooting guidelines for specific error codes or system component abnormal states, etc.), and a third query statement template for implementing a performance optimization intent (such as retrieving optimization opinions or configuration parameters related to system component states, etc.); then embed the plurality of key entities and the candidate knowledge entry set into the target query statement template as query variables and query constraints to obtain the structured query statement. Therefore, the query statement obtained by the above method follows the syntax standards of graph databases (such as Cypher, Gremlin, etc.), and can accurately specify the query path, relationship type, and entity attribute.

[0061] Next, the intelligent auxiliary system can submit the structured query statement to the graph database engine for execution and obtain the set of knowledge entries directly related to the questioning intention and key entities fed back by the graph database engine. In this way, the templated structured query statement can be effectively utilized in combination with the target questioning intention and key entity information to find the required knowledge entries in the operation and maintenance knowledge graph with extremely high efficiency and pertinence, greatly improving the response quality and user experience.

[0062] For example, if the target question text is "prompt for password error when logging in", the inverted index in the operation and maintenance knowledge graph is first used for efficient query to locate the knowledge entries containing at least one of the key entities "user login process" and "password error" identified in the question text, to obtain a set of candidate knowledge entries. Then, according to the questioning intention "operation guidance intention" of the question text, a first query statement template matching the intention is found from the preset template library, such as "MATCH(n:OperationStep)-[:RELATION]->(m:Error)WHERE m.name=errorNameRETURN n". Then, the key entities identified from the question text and the set of candidate knowledge entries are embedded in the first query statement template, and the m.name m.name variable is replaced with "password error" to construct a specific structured query statement, such as the Cypher query statement: "MATCH(n:OperationStep)-[:RELATION]->(m:Error{name:'password error'})RETURN n". Finally, the structured query statement is fed back to the graph database engine, and the query result returned by the graph database engine is received, and the query result is parsed to extract the set of knowledge entries associated with the "operation guidance intention" and containing the "user login process" and "password error" key entities.

[0063] Further, the system finds the multi-dimensional attribute information of each knowledge entry in the set of knowledge entries through the forward index. The forward index is used to store the attribute information of each knowledge entry in the operation and maintenance knowledge graph, such as the knowledge update time (i.e. the time of the last modification or verification), the authority degree (i.e. the source and the professionalism of the writer of the knowledge entry, to ensure the reliability of the knowledge entry), the user satisfaction score (i.e. the feedback of the user on the actual application of the knowledge entry), etc. These information are added to the forward index when the knowledge entry is created or updated, ensuring the complete storage of multi-dimensional information related to knowledge.

[0064] Therefore, the intelligent auxiliary system can take the frequency of the key entities in the inverted index (i.e., the number of occurrences of each key entity in each knowledge item, the higher the number of occurrences, the greater the relevance of the knowledge item to the target problem text, and vice versa) and the multi-dimensional attribute information in the forward index as the sorting factors to sort the multiple knowledge items in the knowledge item set. The sorting process can use a descending sorting method, i.e., the knowledge item with a higher score is ranked higher.

[0065] Therefore, the intelligent auxiliary system can convert the user problem or system error into a precise query for the operation and maintenance knowledge base, which is much better than the traditional keyword search, can quickly locate the problem, reduce invalid queries, and greatly shorten the problem solving time.

[0066] Finally, the intelligent auxiliary system can feed back the sorted knowledge item set to the target terminal, ensuring that the knowledge items with high relevance, strong authority, good user feedback, and recent updates are presented first, thereby improving the efficiency of problem solving and user experience.

[0067] As an optional implementation, in the technical solution provided in step S108, the intelligent auxiliary system can feed back the sorted knowledge item set to the target terminal in the following manner:

[0068] First, determine the data content of each knowledge item in the sorted knowledge item set.

[0069] In the case of configuration parameters as the data content, the intelligent auxiliary system can feed back the knowledge item to the target terminal in a text format. The text format can be JSON, XML, or a simple text list format to present the configuration parameters, ensuring that the target terminal (user or automated script) can easily read and apply these parameters.

[0070] In the case of solution step procedures as the data content, the knowledge item is fed back to the target terminal in a text format or a video format. For example, if the solution step procedure is relatively simple and clear, the intelligent auxiliary system can feed back the solution step procedure guide in a text format to facilitate quick reference by the target terminal. If the solution step procedure involves multiple interface operations or complex operation procedures, the system can provide a video tutorial to assist users in understanding the operation procedure through visual demonstration.

[0071] In the case of standardized operation and maintenance as the data content, instructions corresponding to the standardized operation and maintenance can be generated and fed back to the target terminal, where the standardized operation and maintenance includes but is not limited to automatic restart of services, automatic execution of database backup, automatic cleaning of cache, etc.

[0072] It should be noted that the intelligent auxiliary system can also adapt to different needs and preferences of the user, and set a dynamic adaptation and optimization mechanism. For example, if the user prefers video format when receiving the solution step guide, the system will gradually adjust its output strategy according to user feedback and increase the proportion of video feedback. If the network status of the target terminal is not good, the system will preferentially select the text format to reduce the data transmission amount and ensure the rapid delivery of knowledge items.

[0073] In addition, after the intelligent auxiliary system feeds back the sorted knowledge item set to the target terminal, it can also obtain the feedback evaluation of any knowledge item in the knowledge item set by the target terminal, wherein the feedback evaluation at least includes a satisfaction score and an optimization suggestion. The satisfaction score can be in numerical form, such as a five-star rating system, and the optimization suggestion can be specific improvement points or additional information needs. If the satisfaction score of the knowledge item is higher than a preset threshold value, the system considers that the knowledge item provides an effective solution for the target terminal, and therefore maintains the knowledge item in the operation and maintenance knowledge graph unchanged and continues to participate in the subsequent query process as part of the graph. If the satisfaction score of the knowledge item is not higher than the preset threshold value, the system will first review the optimization suggestion provided by the target terminal. If the optimization suggestion is reasonable and can substantially improve the quality of the knowledge item, the system will modify or reconstruct the knowledge item in the operation and maintenance knowledge graph according to the optimization suggestion, wherein the modification includes correcting incorrect information, supplementing missing steps, introducing new parameters, etc., to enhance the practicality and reliability of the knowledge item.

[0074] Therefore, in the above-mentioned digital problem processing method, the intelligent auxiliary system covers the entire workflow from problem identification, scheme matching to execution feedback, forming a complete closed loop. This closed loop mechanism ensures the continuous optimization and iteration of the solution and the real-time inclusion of user feedback, improving the continuous effectiveness of system maintenance and service.

[0075] Embodiment 2

[0076] According to the embodiments of the present application, a digital system problem processing apparatus for implementing the digital system problem processing method in Embodiment 1 is also provided, as shown in Figure 2 The digital system problem processing apparatus at least includes an identification module 22, a query module 24, a sorting module 26 and a feedback module 28, wherein:

[0077] The identification module 22 is configured to perform intent recognition and entity recognition on the target problem text input by the target terminal to obtain the corresponding target question intent and multiple key entities, wherein the type of the target problem text at least includes an operation problem or a fault problem of the digital system;

[0078] The query module 24 is configured to search for a knowledge item set identical to the target question intention in the preset operation and maintenance knowledge graph through an inverted index, and search for multi-dimensional attribute information of each knowledge item in the knowledge item set through a forward index, wherein each knowledge item in the knowledge item set contains at least one key entity of the target question text, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree, and a user satisfaction score.

[0079] The sorting module 26 is configured to sort the plurality of knowledge items in the knowledge item set by taking the frequency of the key entity in the inverted index and the multi-dimensional attribute information in the forward index as sorting factors.

[0080] The feedback module 28 is configured to feed back the sorted knowledge item set to the target terminal.

[0081] The functions of the modules of the digital system question processing device will be described in detail in combination with a specific implementation process.

[0082] Optionally, the identification module 22 can implement intention recognition and entity recognition on the target question text input by the target terminal according to the following steps:

[0083] First step: obtaining the target question text input by the target terminal.

[0084] Second step: performing a first preprocessing operation on the target question text, so as to ensure the data quality for subsequent analysis. The first preprocessing operation includes but is not limited to: removing noise characters such as punctuation marks and special symbols, so as to ensure the purity and readability of the text; unifying the text format, such as converting the text to lowercase, removing extra spaces, and ensuring the consistency of the text; processing abbreviations, that is, expanding them to complete words through a vocabulary table, so as to improve the accuracy of understanding and matching; using synonym replacement to enhance the diversity and accuracy of semantic expression. The synonym replacement can be realized by searching a synonym dictionary or using a pre-trained word embedding model.

[0085] Third step: encoding the preprocessed target question text by using a deep learning model (such as a BERT model), obtaining a corresponding text embedding vector, and then analyzing the text embedding vector by using an intention recognition model, obtaining the probability of the target question text belonging to each of the preset question intentions, and taking the question intention with the highest probability as the target question intention corresponding to the target question text. The intention recognition model is obtained by training a deep learning model by using a plurality of training sample data, and each set of training sample data includes a text embedding vector corresponding to a question text and a question intention corresponding to the question text, and the question intention includes at least one of the following: an operation guidance intention, a fault diagnosis intention, a performance optimization intention, and a system state query intention.

[0086] Fourth step: using a named entity recognition model to analyze the preprocessed target problem text, obtaining multiple key entities in the target problem text related to the system operation and maintenance field. The named entity recognition model is obtained by training a deep learning model using a large number of system operation and maintenance field text samples labeled with entity information, and the deep learning model includes but is not limited to BERT model, BERT+Conditional Random Field (CRF) model, etc.

[0087] Further, the device further includes a graph construction module for constructing a structured and highly correlated knowledge base, i.e., an operation and maintenance knowledge graph, and based on this, the intelligent auxiliary system provides more accurate, efficient and deep understanding services.

[0088] Optionally, the graph construction module can construct the operation and maintenance knowledge graph according to the following steps, including:

[0089] First step: obtaining multiple multi-modal knowledge data related to the system operation and maintenance field. The multi-modal knowledge data includes at least one of the following: system operation log, system operation and maintenance document, system operation manual, system operation video tutorial, system configuration parameter. These data sources are diverse and can comprehensively cover all aspects of system operation and maintenance.

[0090] Second step: for each multi-modal knowledge data, performing a second preprocessing operation on the multi-modal knowledge data to ensure that the format of all multi-modal knowledge data is uniform, facilitating subsequent analysis and processing. The second preprocessing operation at least includes: converting image data into text data using optical character recognition technology, converting audio and video data into text data using automatic speech recognition technology, and performing a first preprocessing operation on the text data.

[0091] Third step: using a named entity recognition model to analyze the preprocessed multi-modal knowledge data, obtaining multiple entities in the multi-modal knowledge data related to the system operation and maintenance field, and determining the entity attributes of each entity; using a pre-set operation and maintenance field knowledge rule to extract the semantic relationship between each entity.

[0092] Among the multiple entities, at least include: system component state entity, business process entity, exception type entity, error code entity, technical parameter optimization entity, operation process entity, operation and maintenance step entity, etc. In addition, the extraction of semantic relationships can be assisted by dependency syntax analysis, cross-document relationship extraction, template matching and other technologies to ensure that the constructed graph relationship is clear and logically coherent, and the semantic relationship between the extracted entities can include but is not limited to: influence relationship, dependency relationship, causal relationship, operation relationship, and association relationship.

[0093] Fourth step: taking each entity as a graph node and taking the semantic relationship between each entity as an edge, an operation and maintenance knowledge graph is constructed, wherein the plurality of knowledge entries included in the operation and maintenance knowledge graph are composed of a plurality of graph nodes and a plurality of edges associated therewith, and the operation and maintenance knowledge includes a problem and a solution step flow or a configuration parameter.

[0094] Therefore, the graph construction module utilizes the semantic understanding capability of the large language model to construct a multi-modal operation and maintenance knowledge base integrating text, image, video and other types, and such fusion enables the device to handle more complex query scenarios and improve the comprehensiveness and practicality of the solution.

[0095] Further, the query module 24 can use an inverted index to find a knowledge entry set same as the target question intention in the operation and maintenance knowledge graph constructed by the graph construction module in advance. The inverted index is an efficient data retrieval mechanism, unlike the way of retrieving entities by document identifier information in the forward index, the inverted index records which knowledge entries the key entities appear in reversely, so that all knowledge entries containing the entity can be quickly located according to the key entity. Therefore, the system first locates all knowledge entries containing each key entity through the inverted index, and then determines which knowledge entry set is most matched with the target question intention, and finally obtains the knowledge entry set same as the target question intention, and each knowledge entry in the knowledge entry set contains at least one key entity of the target question text.

[0096] Optionally, the query module 24 can find the knowledge entry set same as the target question intention from the above operation and maintenance knowledge graph according to the following steps, including:

[0097] First step: find a candidate knowledge entry set related to the target question text in the operation and maintenance knowledge graph through the inverted index.

[0098] Second step: construct a structured query statement according to the target question intention, the plurality of key entities and the candidate knowledge entry set, and use the structured query statement to query the candidate knowledge entry set to obtain the knowledge entry set.

[0099] Specifically, regarding the structured query statement, the query module 24 can be constructed in the following steps: first, determine the target query statement template corresponding to the target question intention from the preset template library, wherein the template library includes a plurality of query statement templates, and each query statement template is classified and customized for different question intentions. Specifically, the template library at least includes the following three types of query statement templates: the first query statement template for implementing the operation guidance intention (such as querying specific operation steps, etc.), the second query statement template for implementing the troubleshooting intention (such as querying troubleshooting guidelines for specific error codes or system component abnormal state, etc.), and the third query statement template for implementing the performance optimization intention (such as retrieving optimization opinions or configuration parameters related to system component state, etc.); then embed a plurality of key entities and candidate knowledge entries into the target query statement template as query variables and query constraints to obtain a structured query statement.

[0100] Next, the query module 24 can submit the structured query statement to the graph database engine for execution and obtain the knowledge entry set directly related to the question intention and the key entity fed back by the graph database engine. In this way, the templated structured query statement can be effectively utilized in combination with the target question intention and key entity information to find the required knowledge entries in the operation and maintenance knowledge graph with extremely high efficiency and pertinence, greatly improving the response quality and user experience.

[0101] Further, the query module 24 finds the multi-dimensional attribute information of each knowledge entry in the knowledge entry set through the forward index. The forward index is used to store the attribute information of each knowledge entry in the operation and maintenance knowledge graph, such as knowledge update time (i.e., the last time it is modified or verified), authority degree (i.e., the source and professionalism of the knowledge entry writer, ensuring the reliability of the knowledge entry), user satisfaction score (i.e., user feedback on the actual application of the knowledge entry), etc. These information are added to the forward index when the knowledge entry is created or updated, ensuring the complete storage of multi-dimensional information related to knowledge.

[0102] Next, the sorting module 26 can use the frequency of the key entity in the inverted index (i.e., the number of occurrences of each key entity in each knowledge entry, the higher the number of occurrences, the greater the relevance of the knowledge entry to the target question text, and vice versa) and the multi-dimensional attribute information in the forward index as sorting factors to sort the plurality of knowledge entries in the knowledge entry set.

[0103] Finally, the feedback module 28 can feed back the sorted knowledge entry set to the target terminal, ensuring that the knowledge entries with high relevance, strong authority, good user feedback, and recent updates are presented first, thereby improving the efficiency of problem solving and user experience.

[0104] Optionally, the feedback module 28 can feed the ranked knowledge item set to the target terminal in the following manner, including:

[0105] First, determine the data content of each knowledge item in the ranked knowledge item set.

[0106] In the case of data content being configuration parameters, the intelligent assistant system can feed the knowledge item to the target terminal in a text format. The text format can be JSON, XML, or a simple text list format to present the configuration parameters, ensuring that the target terminal (user or automated script) can easily read and apply these parameters.

[0107] In the case of data content being a solution step process, the knowledge item is fed to the target terminal in a text format or a video format. For example, if the solution step process is relatively simple and clear, the intelligent assistant system can feed the solution step process guide through a text format to facilitate quick reference by the target terminal; if the solution step process involves multiple interface operations or complex operation processes, the system can provide a video tutorial to assist the user in understanding the operation process through visual demonstration.

[0108] In the case of data content being a standardized operation and maintenance operation, instructions corresponding to the standardized operation and maintenance operation can be generated and fed to the target terminal, where the standardized operation and maintenance operation includes but is not limited to automatic restart of service, automatic execution of database backup, automatic cleaning of cache, etc.

[0109] In addition, the digital system problem processing device also includes a feedback module, which can obtain feedback evaluation of the target terminal on any knowledge item in the knowledge item set, where the feedback evaluation at least includes a satisfaction score and an optimization suggestion. The satisfaction score can be in numerical form, such as a five-star rating system, while the optimization suggestion can be a specific improvement point or additional information requirement. In the case where the satisfaction score of the knowledge item is higher than a preset threshold, the system considers that the knowledge item provides an effective solution for the target terminal, and therefore maintains the knowledge item in the operation and maintenance knowledge graph unchanged, continuing to participate in the subsequent query process as part of the graph. In the case where the satisfaction score of the knowledge item is not higher than the preset threshold, the system first reviews the optimization suggestion provided by the target terminal. If the optimization suggestion is reasonable and can substantially improve the quality of the knowledge item, the system will modify or reconstruct the knowledge item in the operation and maintenance knowledge graph according to the optimization suggestion, where the modification includes correcting incorrect information, supplementing missing steps, introducing new parameters, etc., to enhance the practicality and reliability of the knowledge item.

[0110] It should be noted that the modules in the digital system problem processing apparatus in the embodiments of the present application correspond one by one to the implementation steps of the digital system problem processing method in Embodiment 1. Since Embodiment 1 has been described in detail, the details not embodied in this embodiment can be referred to Embodiment 1, and will not be described in detail here.

[0111] Embodiment 3

[0112] According to the embodiments of the present application, a computer program product is also provided, which includes a computer program. When the computer program is executed by a processor, the digital system problem processing method in Embodiment 1 is implemented.

[0113] According to the embodiments of the present application, a non-volatile storage medium is also provided, which includes a stored computer program. The device in which the non-volatile storage medium is located executes the digital system problem processing method in Embodiment 1 by running the computer program.

[0114] According to the embodiments of the present application, a processor is also provided, which is used to run a computer program. When the computer program is run, the digital system problem processing method in Embodiment 1 is executed.

[0115] According to the embodiments of the present application, an electronic device is also provided, which includes a memory and a processor. The memory stores a computer program, and the processor is configured to execute the digital system problem processing method in Embodiment 1 by the computer program.

[0116] Specifically, when the computer program is run, the following steps are implemented: performing intent recognition and entity recognition on a target question text input by a target terminal to obtain a corresponding target question intent and a plurality of key entities, wherein the type of the target question text at least includes an operation question or a fault question of a digital system; finding a knowledge item set same as the target question intent in a preset operation and maintenance knowledge graph through an inverted index, and finding multi-dimensional attribute information of each knowledge item in the knowledge item set through a forward index, wherein each knowledge item in the knowledge item set contains at least one key entity of the target question text, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree, and a user satisfaction score; taking the frequency of the key entity in the inverted index and the multi-dimensional attribute information in the forward index as a sorting factor to sort a plurality of knowledge items in the knowledge item set; and feeding back the sorted knowledge item set to the target terminal.

[0117] As an optional implementation, the electronic device can exist in the form of a mobile terminal, a computer terminal, or a similar computing device. Figure 3A hardware structure block diagram of an electronic device for implementing the digital system problem processing method is shown. As shown in Figure 3 The electronic device 30 can include one or more processors 302 (the processor 302 can include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 304 for storing data, and a transmission device 306 for communication functions. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports in the BUS bus), a network interface, a power supply and / or a camera. Those skilled in the art can understand that Figure 3 The structure shown is only schematic, which does not limit the structure of the above-mentioned electronic device. For example, the electronic device 30 can include more or fewer components than those shown in Figure 3 or have a different configuration than that shown in Figure 3 .

[0118] It should be noted that the one or more processors 302 and / or other data processing circuits described above can be referred to herein as "data processing circuits" in general. The data processing circuit can be embodied in whole or in part as software, hardware, firmware or any combination thereof. In addition, the data processing circuit can be a single independent processing module, or all or part of any one of the other elements combined into the electronic device 30. As referred to in the embodiments of the present application, the data processing circuit serves as a processor control (for example, selection of a variable resistance terminal path connected to an interface).

[0119] The memory 304 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the digital system problem processing method in the embodiments of the present application. The processor 302 executes various functional applications and data processing by running the software programs and modules stored in the memory 304, that is, implements the vulnerability detection method of the application program described above. The memory 304 can include a high-speed random access memory, and can also include a non-volatile memory such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 304 can further include a memory remotely disposed with respect to the processor 302, which can be connected to the electronic device 30 through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0120] The transmission device 306 is configured to receive or send data via a network. The network can include a wireless network provided by a communication provider of the electronic device 30. In one embodiment, the transmission device 306 includes a network interface controller (NIC) that can be connected to other network devices through a base station to communicate with the Internet. In one embodiment, the transmission device 306 can be a radio frequency (RF) module that is configured to communicate with the Internet wirelessly.

[0121] The display can be a liquid crystal display (LCD) that is touch screen type, for example, which can enable a user to interact with a user interface of the electronic device 30.

[0122] The above-mentioned embodiment numbers are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0123] In the above-mentioned embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0124] In the several embodiments provided by the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the unit embodiment described above is only schematic. For example, the division of the units can be a logical function division, and there can be another division manner in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, unit or module, and can be electrical or other forms.

[0125] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of units. Part or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0126] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0127] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0128] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should be considered as the protection scope of the present application.

Claims

1. A method for handling a system problem in digitization, characterized by, The application comprises the following steps: Intention recognition and entity recognition are performed on the target question text input by the target terminal to obtain a corresponding target question intention and multiple key entities, wherein the type of the target question text at least includes an operation question or a fault question of a digital system; A knowledge item set identical to the target question intention is searched for in a preset operation and maintenance knowledge graph through inverted indexing, and multi-dimensional attribute information of each knowledge item in the knowledge item set is searched for through forward indexing, wherein at least one key entity of the target question text is contained in each knowledge item in the knowledge item set, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree and a user satisfaction score; The frequency of the key entity in the inverted indexing and the multi-dimensional attribute information in the forward indexing are taken as sorting factors to sort multiple knowledge items in the knowledge item set; The sorted knowledge item set is fed back to the target terminal.

2. The method of claim 1, wherein, Intention recognition and entity recognition are performed on the target question text input by the target terminal to obtain a corresponding target question intention and multiple key entities, comprising the following steps: A target question text input by the target terminal is acquired; First preprocessing operations are performed on the target question text, wherein the first preprocessing operations at least include removing noise characters, unifying text formats, processing abbreviations and synonym replacement; The preprocessed target question text is encoded to obtain a corresponding text embedding vector, and an intention recognition model is used to analyze the text embedding vector to obtain the probability that the target question text belongs to each of multiple preset question intentions, and the question intention with the highest probability is taken as the target question intention corresponding to the target question text, wherein the intention recognition model is obtained by training a deep learning model using multiple sets of training sample data, and each set of training sample data includes a text embedding vector corresponding to a question text and a question intention corresponding to the question text, and the question intention at least includes one of the following: an operation guidance intention, a fault diagnosis intention, a performance optimization intention and a system state query intention; A named entity recognition model is used to analyze the preprocessed target question text to obtain multiple key entities related to the system operation and maintenance field in the target question text.

3. The method of claim 1, wherein, The construction process of the operation and maintenance knowledge graph comprises the following steps: Multiple multi-modal knowledge data related to the system operation and maintenance field are acquired, wherein the multi-modal knowledge data at least include one of the following: system running logs, system operation and maintenance documents, system operation manuals, system operation video tutorials and system configuration parameters; For each multi-modal knowledge data, second preprocessing operations are performed on the multi-modal knowledge data, wherein the second preprocessing operations at least include converting image data into text data using optical character recognition technology, converting audio and video data into text data using automatic speech recognition technology, and performing first preprocessing operations on the text data; The preprocessed multi-modal knowledge data is analyzed by using a named entity recognition model to obtain a plurality of entities related to the system operation and maintenance field in the multi-modal knowledge data, and a preset operation and maintenance field knowledge rule is used to extract semantic relationships between the entities, wherein the plurality of entities at least include: system component state entities, business process entities, error code entities, exception type entities, technical parameter optimization entities, operation process entities, and operation and maintenance step entities; and the semantic relationships between the entities are extracted by using the preset operation and maintenance field knowledge rule. Each of the entities is taken as a graph node, and the semantic relationships between the entities are taken as edges to construct the operation and maintenance knowledge graph, wherein a plurality of knowledge entries included in the operation and maintenance knowledge graph are operation and maintenance knowledge including questions and solution step processes or configuration parameters composed of a plurality of graph nodes and a plurality of associated edges.

4. The method of claim 1, wherein, A same knowledge entry set as the target question intention is found in the preset operation and maintenance knowledge graph through an inverted index, including: A candidate knowledge entry set related to the target question text is found in the operation and maintenance knowledge graph through an inverted index, wherein each initial knowledge entry in the candidate knowledge entry set includes at least one key entity of the target question text; A structured query statement is constructed according to the target question intention, the plurality of key entities, and the candidate knowledge entry set, and the structured query statement is used to query the candidate knowledge entry set to obtain the knowledge entry set.

5. The method of claim 4, wherein, A structured query statement is constructed according to the target question intention, the plurality of key entities, and the candidate knowledge entry set, including: A target query statement template corresponding to the target question intention is determined from a preset template library, wherein the template library at least includes: a first query statement template for implementing an operation guidance intention, a second query statement template for implementing a fault troubleshooting intention, and a third query statement template for implementing a performance optimization intention; The plurality of key entities and the candidate knowledge entry set are embedded into the target query statement template as query variables and query constraints to obtain the structured query statement.

6. The method of claim 1, wherein, The sorted knowledge entry set is fed back to the target terminal, including: The data content of each knowledge entry in the sorted knowledge entry set is determined; In the case that the data content is a configuration parameter, the knowledge entry is fed back to the target terminal in a text format; In the case that the data content is a solution step process, the knowledge entry is fed back to the target terminal in a text format or a video format.

7. The method of claim 1, wherein, After the sorted knowledge entry set is fed back to the target terminal, the method further includes: A feedback evaluation of any knowledge entry in the knowledge entry set by the target terminal is obtained, wherein the feedback evaluation at least includes: a satisfaction score and an optimization suggestion; In the case that the satisfaction score of the knowledge entry is not higher than a preset threshold value, the knowledge entry in the operation and maintenance knowledge graph is modified according to the optimization suggestion.

8. A digital system problem processing apparatus characterized by comprising: including: An identification module is configured to perform intent recognition and entity recognition on target question text input by a target terminal to obtain a corresponding target question intent and a plurality of key entities, wherein the target question text is at least one of an operation question or a fault question of a digital system; A query module is configured to find a knowledge item set identical to the target question intent in a preset operation and maintenance knowledge graph through an inverted index and find multi-dimensional attribute information of each knowledge item in the knowledge item set through a forward index, wherein each knowledge item in the knowledge item set contains at least one key entity of the target question text, and the multi-dimensional attribute information at least includes a knowledge update time, an authority degree, and a user satisfaction score; An ordering module is configured to take the frequency of the key entity in the inverted index and the multi-dimensional attribute information in the forward index as ordering factors to order a plurality of knowledge items in the knowledge item set; A feedback module is configured to feed back the ordered knowledge item set to the target terminal.

9. A computer program product, characterised in that, The computer program is executed by a processor to implement the digital system question processing method in any one of claims 1 to 7. The memory stores a computer program, and the processor is configured to execute the digital system question processing method in any one of claims 1 to 7 through the computer program.

10. An electronic device, comprising: ​ ​

Images