Work order automatic processing method, device, storage medium and program product combined with semantic expansion

Abstract

Embodiments of the present application provide a work order automatic processing method and device combined with semantic expansion, a storage medium and a program product, relating to the technical field of data processing. The method comprises: obtaining an original work order description; performing semantic expansion processing on the original work order description to generate a question set containing at least one extended question; performing semantic matching on the question set and a historical solution knowledge base according to a preset similarity threshold to obtain a matching result; in response to the matching result indicating a matching success, taking the matched historical solution as a preliminary solution; in response to the matching result indicating a matching failure, collecting real-time running data related to the original work order description from a correlation system to generate a preliminary solution; receiving a first instruction from an audit interface; and outputting a final solution according to the first instruction. The method of the present application increases the retrieval coverage of the problem solution and improves the overall efficiency of complex work order processing.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to an automatic work order processing method, device, storage medium, and program product that incorporates semantic extension. Background Technology

[0002] In IT operations, customer service, and technical support, ticketing systems are core tools for handling user technical issues and requests. When users encounter technical problems such as server performance anomalies or application errors while using a system or service, they typically need to submit a service request with a description of the problem through a ticketing system.

[0003] In current implementations, the follow-up processing of such work orders heavily relies on manual operation. Specifically, after receiving a work order, support personnel need to rely on their personal experience to understand the problem and manually log into multiple distributed systems (such as server terminals, log monitoring centers, databases, source code repositories, etc.) to retrieve information and collect data, thereby analyzing the root cause of the problem and formulating a solution.

[0004] However, the above implementation method has the problem of being difficult to scale up and efficiently process massive or complex work order requests. Summary of the Invention

[0005] This application provides a method, device, storage medium, and program product for automatically processing work orders by incorporating semantic extension, in order to solve the technical problem of efficiency in processing massive or complex work order requests.

[0006] In a first aspect, embodiments of this application provide an automatic work order processing method incorporating semantic extension, the method comprising:

[0007] Obtain the original work order description;

[0008] The original work order description is semantically expanded to generate a set of questions containing at least one expanded question.

[0009] Based on a preset similarity threshold, semantic matching is performed between the question set and the historical solution knowledge base to obtain the matching results;

[0010] In response to a successful match result, the matched historical solutions are used as preliminary solutions.

[0011] In response to a matching result indicating a matching failure, real-time operational data related to the original work order description is collected from at least one associated system based on the original work order description. The original work order description and the real-time operational data are then input into a pre-trained analysis model to generate a preliminary solution.

[0012] Receive a first instruction from the review interface for the preliminary solution, the first instruction including at least one of confirmation, modification and regeneration instructions;

[0013] Based on the first instruction, output the final solution.

[0014] In this embodiment, by introducing an automated workflow that combines semantic expansion, intelligent retrieval and matching, multi-source data fusion analysis, and mandatory manual review, the problems of reliance on human experience in traditional methods and low matching rates in simple retrieval systems are solved, thereby improving the efficiency and accuracy of work order processing. Specifically, semantic expansion expands the scope of retrieval by mapping the original fuzzy description to multiple technically equivalent expressions, thereby increasing the reuse rate of the historical knowledge base and reducing matching failures caused by expression differences. When faced with new problems without historical solutions, the system automatically collects multi-source data and uses a pre-trained model for analysis and reasoning, achieving automated root cause localization of complex problems. The mandatory manual review process forms a human-machine collaborative decision-making loop, ensuring the efficiency of automated processing while improving the reliability and business security of the final output solution.

[0015] In one possible implementation, the original work order description undergoes semantic expansion processing, including:

[0016] A pre-trained semantic model is used to process the original work order description to generate at least one extended question covering multiple semantic types, including at least one of synonym substitution, cause deduction, and scene visualization.

[0017] In this implementation, by generating extended questions covering multiple semantic types (such as synonym substitution, causal deduction, and scene visualization), the vague original description is mapped into a multi-dimensional set of semantic expressions. This technique broadens the semantic coverage of retrieval and improves the relevance and accuracy of search results, thereby reducing matching failures caused by differences in expression. This allows more work orders to obtain solutions during the knowledge base retrieval stage, effectively improving the automation level and response efficiency of the processing flow.

[0018] One possible implementation involves using a pre-trained semantic model to process the original work order description, and also includes:

[0019] Obtain contextual information associated with the original work order description;

[0020] Based on the original work order description and context information, at least one extended question is generated through a semantic model; wherein, the context information includes at least one of the following: device hardware parameters, system version information, and operation log fragments.

[0021] In this implementation, by introducing and integrating contextual information, the semantic expansion process is upgraded from processing isolated problem texts to interpreting the problem-environment complex. This enhances the technical relevance and scenario adaptability of the generated expanded questions, enabling subsequent semantic matching to more directly and accurately locate solutions that occurred in the same or similar technical environments in the past, thereby improving the accuracy and efficiency of handling complex technical problems.

[0022] In one possible implementation, the method further includes:

[0023] Based on the matching results and the first instruction, generate feedback information related to the problem type described in the original work order;

[0024] Based on the feedback information, the similarity threshold used for semantic matching of similar question types is updated.

[0025] In this implementation, by introducing a feedback loop mechanism, the system's semantic matching module can automatically adapt to the semantic complexity of different problem types and the distribution characteristics of the historical knowledge base as the number of work orders processed accumulates, making the matching strategy more targeted, thereby improving the accuracy of the first retrieval and the overall intelligence level of the process.

[0026] In one possible implementation, real-time operational data related to the original work order description is collected from at least one associated system, including:

[0027] Based on the context information associated with the original work order description, select a combination of target interfaces from a pre-configured set of application interfaces;

[0028] Data collection is performed by calling the target interface combination; if the call fails, a predefined exception handling strategy is executed, which includes enabling a backup interface or retrying the operation.

[0029] In this implementation, the aforementioned mechanism enables the data acquisition module to obtain critical operational data to the maximum extent possible, even in complex and heterogeneous operational environments and when faced with unavoidable network jitter or temporary service unavailability. This provides a stable and reliable data input foundation for subsequent analysis models, thereby improving the availability of the entire analysis path and the quality of the final generated solution.

[0030] In one possible implementation, after outputting the final solution, the following is also included:

[0031] The final solution is submitted to the work order management platform to close or update the work order status corresponding to the original work order description; the work order management platform is an external business system used to manage the lifecycle of work orders.

[0032] In this implementation, through the aforementioned technical steps, the system can automatically transform the solutions derived from analysis into actual operational instructions in external business systems. This achieves end-to-end automation from problem analysis to business status updates, avoiding the efficiency losses and operational risks caused by manual repetitive operations and information synchronization between two systems. Ultimately, this design allows the processing flow to be embedded into the enterprise's existing management system, forming a complete automated closed loop, improving work order processing efficiency while also enhancing the standardization of the operation and maintenance process.

[0033] Secondly, this application provides a work order processing system, comprising:

[0034] The semantic extension module is used to perform semantic extension processing on the original work order description to generate a set of questions containing at least one extended question.

[0035] The semantic matching module is used to semantically match the problem set with the historical solution knowledge base;

[0036] The data acquisition and analysis module is used to collect real-time running data from at least one related system based on the original work order description when the semantic matching module fails to match, and to generate a preliminary solution based on the original work order description and the real-time running data; wherein, the preliminary solution is a historical solution matched by the semantic matching module, or a solution generated by the data acquisition and analysis module;

[0037] The review interaction module is used to output the preliminary solution to the review interface and receive the first instruction from the review interface regarding the preliminary solution.

[0038] The output module is used to output the final solution based on the first instruction.

[0039] Thirdly, this application provides an electronic device, including: a processor and a memory communicatively connected to the processor;

[0040] The memory stores the instructions that the computer executes;

[0041] The processor executes computer-executable instructions stored in memory to implement any of the methods of the first aspect.

[0042] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method of any one of the first aspects.

[0043] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method of any one of the first aspects. Attached Figure Description

[0044] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0045] Figure 1 A flowchart illustrating an automatic work order processing method incorporating semantic extension, provided as an embodiment of this application;

[0046] Figure 2 This is a schematic diagram of the overall architecture and processing flow of a work order processing system provided in an embodiment of this application.

[0047] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0048] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0049] It should be noted that the automatic work order processing method, device, storage medium, and program product with semantic extension provided in this application can be used in the field of data processing technology, or in any field other than data processing technology. This application does not limit the application field of the automatic work order processing method, device, storage medium, and program product with semantic extension.

[0050] This application can be applied to business scenarios requiring frequent handling of complex technical issues, such as enterprise IT operations, customer service, and technical support. Specifically, it can be widely used in organizations with high levels of IT infrastructure, such as financial institutions, internet companies, and telecommunications operators, to handle various technical problem tickets generated in daily operations. Specific application scenarios include, but are not limited to, the following:

[0051] IT infrastructure operation and maintenance scenario: When enterprise servers, network devices or cloud platforms experience performance abnormalities (such as abnormal spikes in CPU utilization, memory leaks, or surges in network latency), operation and maintenance personnel can use this application system to process relevant alarm work orders and automatically correlate and analyze system logs, monitoring indicators and configuration change records.

[0052] Software application technical support scenario: When end users report problems such as "slow interface loading" or "functional errors" during the operation of enterprise-level software systems, technical support personnel can use the system in this application to automatically correlate and analyze application logs, database query records, and code version information.

[0053] Database management scenario: When the database experiences connection pool exhaustion, a surge in slow queries, or data synchronization anomalies, the database administrator can use this application system to process relevant work orders and automatically analyze database performance indicators, SQL execution plans, and lock wait information.

[0054] Production environment fault diagnosis scenario: When production failures occur in business systems such as e-commerce and online transactions (e.g., payment failure, order abnormality), the operations and maintenance team can quickly process emergency work orders through this application system, while simultaneously analyzing business logs, system monitoring data, and recent code deployment records.

[0055] Security incident response scenario: When the security team receives a ticket for a suspected security vulnerability or attack, the system in this application will automatically correlate and analyze security logs, network traffic data, and system vulnerability scan reports.

[0056] In the aforementioned application scenarios, the system of this application can effectively handle problem descriptions that are vague, incomplete, or lack accurate technical terminology. Through intelligent data collection and analysis processes, it assists technical personnel in quickly locating the root cause of the problem and generating solutions.

[0057] For the above application scenarios, existing technical solutions can typically be implemented in the following two ways:

[0058] In one possible implementation, ticket processing could rely on a manual workflow. Upon receiving a technical issue ticket from a user, support staff would first manually search a pre-defined knowledge base (such as a wiki or other document system) based on the problem description, attempting to match existing historical solutions. If no directly matching solution was found, support staff would manually log into multiple relevant target systems (including server terminals, log monitoring centers, database management systems, source code repositories, etc.) to collect multi-source heterogeneous data such as runtime logs, performance metrics, configuration information, and recent code change records. Based on their personal experience, they would then conduct comprehensive analysis, logical reasoning, and problem attribution on the collected data to ultimately formulate a solution and implement the repair operation.

[0059] In another possible implementation, work order processing can be automated using a retrieval-augmented generation (RAG) technique. This approach first converts historical work orders and their corresponding solution texts into vectorized representations, building a searchable vector knowledge base. When a new work order is submitted, the system converts the original problem description into a query vector, performs a similarity matching search in the vector knowledge base, and finally returns the historical solution text with the highest matching degree as the recommendation result. The core of this approach lies in establishing a direct retrieval and matching path from the original problem description to the corresponding solution.

[0060] However, the above implementation method has the following technical problems:

[0061] First, the traditional manual processing model relies excessively on the experience and manual operation of technical personnel, requiring repeated logins to multiple independent systems to collect and compare scattered data. This non-standardized processing flow not only results in long work order response cycles and low processing efficiency, but also makes it difficult to meet the requirements for processing speed and consistency in scenarios with large-scale concurrent work orders.

[0062] Secondly, current automation solutions based on RAG technology often only perform shallow semantic matching on the original work order description, lacking a deep understanding of the problem's essence and semantic generalization capabilities. Because they fail to effectively expand the expression variations of the original problem and related technical scenarios, the system has limited retrieval coverage when faced with actual work orders that are vaguely worded or use non-standard terminology, making it difficult to accurately associate them with truly applicable historical solutions.

[0063] Furthermore, existing automated systems generally lack necessary reliability assurance mechanisms, particularly the absence of mandatory human review at critical decision-making nodes. This can lead to algorithm-generated solutions being deployed directly without sufficient contextual verification. If errors exist due to semantic misunderstandings or insufficient data integrity, it will directly impact the final quality of work order processing and operational security.

[0064] The automatic work order processing method combined with semantic expansion provided in this application aims to solve the aforementioned technical problems of existing technologies. By constructing an intelligent work order processing workflow, it achieves a systematic optimization of the existing processing mode.

[0065] The core of this method lies in introducing semantic understanding and expansion mechanisms, which can perform in-depth analysis and generalization processing on the original work order descriptions submitted by users to generate a variety of synonyms and related technical scenario variations, thereby broadening the retrieval boundaries and improving the matching success rate and coverage with the historical solution knowledge base.

[0066] Furthermore, this method incorporates an automated data acquisition and intelligent analysis module. This module identifies the necessary data sources based on the work order content and completes data acquisition and correlation between multiple systems through pre-defined interface protocols, replacing the traditional manual login process. Based on the collected multi-source heterogeneous data, the system can perform automated problem analysis and logical reasoning to assist in locating potential root causes.

[0067] To ensure the reliability of the output, this method introduces a mandatory human review constraint mechanism at key decision-making nodes. The preliminary solution generated by the system is not immediately effective but must undergo interactive confirmation or correction by designated personnel. This design, while retaining the efficiency advantages of automated processing, provides the ultimate guarantee for the accuracy and feasibility of the solution through necessary human-machine collaboration.

[0068] Overall, this method combines semantic expansion, automated processing, and human-machine collaboration to achieve a synergistic improvement in the efficiency, coverage, and reliability of the work order processing workflow.

[0069] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0070] First, combine Figure 1 This application introduces an automatic work order processing method incorporating semantic expansion, as provided in its embodiments. Figure 1 A flowchart illustrating an automatic work order processing method incorporating semantic expansion, as provided in this application embodiment, is shown below. Figure 1 As shown, the method includes:

[0071] S101. Obtain the original work order description.

[0072] Specifically, the ticket processing system (hereinafter referred to as the system) receives original tickets submitted by users from the upstream customer service platform or operation and maintenance monitoring system. The ticket contains the problem described by the user in natural language, such as "the production server is extremely slow to respond around 10:00 AM every day", and may be accompanied by relevant contextual information, such as the server IP address, the time of occurrence, and error code snippets.

[0073] S102. Perform semantic expansion processing on the original work order description to generate a set of questions containing at least one expanded question.

[0074] Specifically, the system invokes the semantic extension module (also known as the problem generalization module), which uses a pre-trained semantic model to perform deep semantic analysis on the original work order description. The core of semantic extension processing lies in capturing key entities and states in the original description through the model's deep contextual understanding capabilities, and generating variants based on semantic associations.

[0075] In one possible implementation, a bidirectional encoder representations from transformers (BERT) model based on the Transformer architecture could be used, and this model could be combined with a built-in industry thesaurus and scene rules for generation. Those skilled in the art will also understand that other pre-trained language models with similar capabilities or solutions combining domain knowledge graphs are also suitable for implementing this step. Exemplary examples include, for instance, enhanced representation through knowledge integration (ERNIE) and a robustly optimized BERT approach (RoBERTa).

[0076] Through this process, the system can automatically generate multiple semantically equivalent or related extended questions. For example, for "slow server response," it may generate multiple extended questions such as "high server CPU utilization causing latency," "application service processing timeout during peak business periods," and "slow database queries causing front-end lag." These extended questions together constitute a set of questions, aiming to cover the same core issue from different angles and technical expressions, thereby improving the hit rate of subsequent searches.

[0077] S103. Based on the preset similarity threshold, perform semantic matching between the problem set and the historical solution knowledge base to obtain the matching results.

[0078] Specifically, the system transforms each extended question in the question set into a high-dimensional vector and performs vector similarity retrieval in the historical solution knowledge base. The construction of the historical solution knowledge base includes the process of cleaning and vectorizing historical work order texts (e.g., using models such as Sentence-BERT) and building an efficient vector index (e.g., using the Faiss library).

[0079] The system calculates the cosine similarity between each extended question and an entry in the knowledge base, and takes the highest score as the score for this match. The system pre-sets a similarity threshold (e.g., 85%). It should be understood that this threshold can be set differently depending on the question type or matching history. If the highest score is greater than or equal to the threshold, the matching result is indicated as "match successful"; otherwise, it is indicated as "match failed".

[0080] S104. In response to the matching result indicating a successful match, the matched historical solution is used as the preliminary solution.

[0081] Specifically, when a match is successful, the system directly retrieves the complete solution corresponding to the historical problem with the highest matching degree from the historical solution knowledge base. This solution typically includes root cause analysis, processing steps, and verification methods. For example, if a historical solution for "high CPU utilization" is matched, its content might be "check the scheduled task and optimize the JVM heap memory parameters of the Java application." The system directly extracts this historical solution as the initial solution for this processing, without initiating subsequent data collection and analysis processes, thus achieving rapid reuse of known high-frequency problems.

[0082] S105. In response to the matching result indicating a matching failure, based on the original work order description, collect real-time operational data related to the original work order description from at least one associated system, and input the original work order description and real-time operational data into a pre-trained analysis model to generate a preliminary solution.

[0083] Specifically, when a match fails, it indicates that the current issue may be a new problem or a complex, compounded problem. The system first automatically invokes pre-configured API interfaces based on the context (e.g., server identifier, application name) in the original work order description to collect recent real-time operational data from multiple related systems (e.g., server performance monitoring platforms, log aggregation systems, database management platforms). For example, it collects CPU, memory, and disk I / O metrics for a specified server during the problem period, retrieves relevant application error logs, and queries slow SQL records in the database.

[0084] Subsequently, the system inputs the structured original work order description along with the collected multi-source real-time operational data into a pre-trained analysis model. This analysis model refers to a large language model that has been pre-trained on massive amounts of general text and code data and can be further fine-tuned on operation logs and work order data. The input to this model is usually a prompt text composed of work order descriptions and structured operational data (in JSON or a specific template format). The model performs comprehensive analysis, logical reasoning, and root cause localization on the input information, following a predefined structured format such as "root cause-step-risk," and ultimately generates a structured preliminary solution. This solution typically includes: the inferred core root cause of the problem, specific operational steps, suggested commands or configuration modifications, and potential risk warnings. For example, the generated solution might be: "The root cause is that the database connection pool is exhausted. Step 1: Log in to the database server and execute the command SHOW PROCESSLIST; check the number of connections. Step 2: Adjust the maxPoolSize parameter in the application configuration file from 50 to 100 and restart the application."

[0085] S106. Receive the first instruction from the review interface for the preliminary solution.

[0086] Specifically, the system pushes the generated preliminary solution (whether from historical matching or model generation) to the review interface. The review interface is a mandatory interactive interface; the system will lock the subsequent process within this interface until it receives a valid first instruction. This interface is typically presented to the designated technical support engineer as a mandatory pop-up or highlighted notification. The engineer must review the solution and take explicit action from the options provided on the interface.

[0087] The first instruction includes: "Confirm" (agree to the solution), "Modify" (edit and adjust the existing solution), or "Regenerate" (requires the system to regenerate the solution based on the same or new information). The process can only continue after the system receives one of these explicit instructions. This ensures that any generated solution must be manually reviewed and confirmed before being deployed to the production environment.

[0088] S107. Output the final solution according to the first instruction.

[0089] Specifically, the system executes corresponding operations based on the first received instruction: if the instruction is "Confirm," the preliminary solution is directly output as the final solution; if the instruction is "Modify," the engineer's modified version is used as the final solution; if the instruction is "Regenerate," a new round of data collection and analysis may be triggered, or the system may return to step S102 to try other semantic expansion paths to generate a new solution. The final solution will be automatically submitted to the work order processing system to reply to the user or guide the execution of automated repair scripts, thereby completing the entire work order processing flow.

[0090] The work order automatic processing method combined with semantic expansion provided in this application solves the problems of reliance on human experience and low matching rate of simple retrieval systems in traditional methods by introducing an automated workflow that combines semantic expansion, intelligent retrieval and matching, multi-source data fusion analysis, and mandatory manual review. This improves the efficiency and accuracy of work order processing. Specifically, semantic expansion expands the retrieval scope by mapping the original fuzzy description to multiple technically equivalent expressions, thereby increasing the reuse rate of the historical knowledge base and reducing matching failures caused by expression differences. When faced with new problems without historical solutions, the system automatically collects multi-source data and uses a pre-trained model for analysis and reasoning, achieving automated root cause localization of complex problems. The mandatory manual review process forms a human-machine collaborative decision-making loop, ensuring the efficiency of automated processing while improving the reliability and business security of the final output solution.

[0091] Optionally, this embodiment further elaborates on the semantic expansion process, which is a key step in performing in-depth processing on the original work order description to generate diversified expanded questions.

[0092] Specifically, semantic expansion processing is implemented through a pre-trained semantic model. This model possesses powerful contextual semantic understanding and generation capabilities, and the system inputs the acquired original work order description into this model.

[0093] The model first performs semantic analysis and encoding on the original description to understand its core problem entities, states, and implicit technical context. Then, based on this understanding, the model generates expanded question types covering multiple preset semantic types, rather than simply replacing synonyms. These semantic types aim to interpret the same problem from different technical dimensions, including the following:

[0094] Synonym substitution type: focuses on describing the same phenomenon using different technical terms or common expressions. For example, expanding "service unavailable" to "service downtime" or "access timeout".

[0095] Cause deduction type: Focuses on transforming the description of a phenomenon into a speculative description of the potential root cause based on common technical causal chains. For example, expanding "slow page loading" to "excessive front-end static resources cause slow loading" or "third-party API response delays slow down the overall page rendering".

[0096] Scenario concretization type: This focuses on adding context or boundary conditions to the problem, making its description more specific and scenario-based. For example, "database error" can be expanded with context to "primary key conflict error occurred during batch data import" or "inconsistent data query caused by read / write split slave database synchronization delay".

[0097] In practice, the system can generate multiple extended questions covering one or more of the above semantic types. For example, for the original description "application crash", it may simultaneously generate synonym replacement type "program terminates abnormally", cause inference type "memory access out of bounds causes crash", and scenario visualization type "application crashes during high-concurrency user login stress test".

[0098] Ultimately, the system outputs a set of questions containing at least one extended question format. This design enables the system to improve the matching success rate with substantively similar or related questions in the historical solution knowledge base, even if the original work order description is vague, colloquial, or incomplete, through multi-dimensional and multi-type semantic expansion. This provides a foundation for subsequent intelligent retrieval.

[0099] By generating extended questions covering multiple semantic types (such as synonym substitution, causal deduction, and scenario visualization), the ambiguous original description is mapped into a multi-dimensional set of semantic expressions. This technique broadens the semantic coverage of retrieval and improves the relevance and accuracy of search results, thereby reducing matching failures caused by differences in expression. This allows more work orders to obtain solutions during the knowledge base retrieval stage, effectively improving the automation level and response efficiency of the processing flow.

[0100] Optionally, this embodiment is a further optimization of the semantic expansion processing method, specifically illustrating how to generate more targeted expanded questions by incorporating contextual information.

[0101] In practice, before performing semantic expansion, the system obtains contextual information associated with the original work order description. This contextual information is typically structured or semi-structured data automatically extracted from the configuration management database (CMDB), monitoring system, or log platform associated with the work order. Specific types include, but are not limited to:

[0102] Device hardware parameters: such as the server model where the problem occurred (e.g., "Intel Xeon Gold 6338"), number of CPU cores, and memory capacity.

[0103] System version information: such as operating system version (e.g., "CentOS 7.9"), middleware version (e.g., "Tomcat 9.0.58"), and application version number.

[0104] Runtime log snippets: Critical error logs or warning messages captured from the relevant system before and after the time the problem occurred. For example, a runtime log snippet might be "ORA-12154:TNS:could not resolve the connectidentifier".

[0105] After obtaining the above context information, the system merges the original work order description with the relevant context information and uses them together as input to the pre-trained semantic model. For example, the input received by the model might be: "Work order description: Database connection failed. Context: Server model is Dell R740, database version is Oracle 19c, log fragment shows ORA-12154."

[0106] The semantic model performs deep understanding and generation based on this enhanced input information, enabling it to identify key entities (e.g., Oracle 19c) and states (e.g., ORA-12154) in the context, and semantically associate these identified contents with the original question (e.g., connection failure). Based on this, the model generates extended questions that incorporate these specific contextual features, thereby achieving more accurate scenario visualization and causal deduction.

[0107] For example, given the above input, the model might generate the following highly contextualized extended questions:

[0108] "Oracle 19c database encountered a TNS connection identifier resolution failure on a Dell R740 server."

[0109] "Database service connection error caused by ORA-12154".

[0110] By introducing and integrating contextual information, the semantic expansion process is upgraded from processing isolated problem texts to interpreting the problem-environment complex. This enhances the technical relevance and scenario adaptability of the generated expanded questions, enabling subsequent semantic matching to more directly and accurately locate solutions that occurred in the same or similar technical environments in the past, thereby improving the accuracy and efficiency of handling complex technical problems.

[0111] Optionally, this embodiment is a further optimization of the aforementioned automatic work order processing method, aiming to achieve system self-adaptation capabilities. Its core lies in the system's ability to learn and adjust based on the actual results of each processing step.

[0112] Specifically, after completing a work order processing loop (i.e., from semantic matching to manual review and output of the final solution), the system will initiate a feedback learning process. This process includes the following steps:

[0113] Step 1: Generating Feedback Information

[0114] The system extracts two key data points from this processing: the semantic matching result (successful or failed match) and the first manual review instruction (confirm, modify, or regenerate). Based on these two data points, combined with the problem type described in the original work order (e.g., categorized as "database connection failure," "application performance bottleneck," etc.), the system generates structured feedback information. This feedback information is essentially a score of the matching quality and the effectiveness of the solution. Specifically, it may include the following:

[0115] If a match is successful and the first instruction is confirmation, positive feedback is generated, indicating that the current threshold is appropriate for this type of problem.

[0116] If a match fails, but the solution subsequently generated by the analysis model is manually confirmed, negative feedback is generated, indicating that the current threshold may be too high, causing valid historical solutions that should have been matched to be missed.

[0117] If the matching is successful or unsuccessful, the first instruction is to modify or regenerate. Then, feedback information indicating that the matched or generated scheme is not completely accurate and may require a comprehensive evaluation of the threshold.

[0118] Step 2: Dynamic Update of Thresholds

[0119] The system integrates the above feedback information into the historical feedback sequence for the corresponding question type. Based on the accumulated feedback data, an optimization algorithm (e.g., a statistical sliding window averaging method or a lightweight online learning model) is used to dynamically adjust the similarity threshold used for semantic matching of that question type. The core logic is:

[0120] For frequently occurring problem types with good feedback, the system may maintain or slightly increase the threshold to achieve more accurate matching and reduce false positives.

[0121] For question types with high matching failure rates or frequent manual modifications, the system may proactively lower the threshold to broaden the search scope and attempt to obtain more potentially relevant historical solutions, thus avoiding triggering complex analysis processes due to overly strict requirements.

[0122] By introducing a feedback loop mechanism, the system's semantic matching module can automatically adapt to the semantic complexity of different problem types and the distribution characteristics of the historical knowledge base as the number of processed work orders accumulates, making the matching strategy more targeted, thereby improving the accuracy of the first retrieval and the overall intelligence level of the process.

[0123] Optionally, this embodiment further describes the reliability implementation mechanism of the data acquisition step, focusing on how to select the data source and ensure the fault tolerance of the acquisition process.

[0124] After a matching failure triggers the data collection process, the system will execute an interface routing and fault-tolerant collection process:

[0125] The system first parses the original work order description and its associated contextual information (such as work order tags, system identifiers, and error keywords). Based on this information, the system selects the optimal combination of target interfaces from a pre-configured application programming interface (API) registry.

[0126] This selection follows a predefined strategy. For example, if the ticket involves slow database queries, the combination of the database performance monitoring API and the slow query log retrieval API will be prioritized. If the ticket description mentions an application crash, the error log collection API for that application and the resource monitoring API for the server it resides on will be prioritized, based on the application name in the context.

[0127] This context-based dynamic selection improves the accuracy and efficiency of data collection, avoids making a large number of invalid calls to unrelated systems, and thus reduces system load and collection latency.

[0128] After selecting the target interface combination, the system calls these interfaces sequentially or in parallel to collect real-time operational data. During this process, the system has a built-in comprehensive exception handling mechanism to handle various faults, which may include the following:

[0129] Interface call failure determination: When an interface call times out, returns an error status code, or has an abnormal data format, the system determines that the call has failed.

[0130] Exception handling strategy execution: If a call fails, the system will not halt the entire data collection process but will immediately trigger a predefined exception handling strategy. This strategy primarily includes two core approaches: one is to enable a backup interface, where the system automatically switches to a pre-configured backup interface for the same data source (e.g., a different version of the API, or another node in a load balancer) to re-initiate the call. The other is to perform a retry operation, where for recoverable errors such as temporary network failures, the system automatically retries the original interface a limited number of times after a short delay.

[0131] Through the above mechanism, the data acquisition module can still acquire critical operational data to the maximum extent possible, even in the face of complex and heterogeneous operation and maintenance environments and unavoidable network jitter or temporary service unavailability. This provides a stable and reliable data input foundation for subsequent analysis models, thereby improving the availability of the entire analysis path and the quality of the final generated solution.

[0132] Optionally, this embodiment supplements the final output stage of the automatic work order processing method, clarifying how the processing flow of this solution can be integrated in a closed loop with the enterprise's existing operation and maintenance management system.

[0133] It should be noted that the work order management platform mentioned in this embodiment refers to a core business system (such as an enterprise's self-developed IT service management tool) that is independently deployed within an enterprise and used to receive, dispatch, track and manage user service requests. This work order management platform is different from the work order processing system protected in this application that performs semantic expansion, matching and analysis.

[0134] After the final solution is output according to the first instruction from the review interaction module, the process defined in this embodiment continues to perform a key operation: automatically submitting the final solution to the work order management platform. The system completes this submission through a pre-configured integration interface (such as a RESTful API). The submitted content not only includes the complete text of the solution, but also usually includes structured instructions or metadata to drive downstream business processes.

[0135] The purpose of this operation is to finalize the status of the original work order that triggered this automated processing flow and synchronize it with the business logic, based on the approved final solution. This is specifically manifested in the following two business actions:

[0136] Closing a work order: If the final solution indicates that the problem has been completely resolved, the system will instruct the work order management platform to update the status field of the work order to "Resolved" or "Closed", thereby formally ending the lifecycle of the service request at the business level.

[0137] Update work order status: If the final solution is a phased processing step or requires subsequent manual verification, the system will instruct the work order management platform to update the work order status to "processing", "pending verification", etc., or add a record of this processing to the work order log, so as to drive the work order to enter the next stage in the established operation and maintenance process.

[0138] Through the aforementioned technical steps, the system can automatically transform the solutions derived from analysis into actual operational instructions in external business systems. This achieves end-to-end automation from problem analysis to business status updates, avoiding the efficiency losses and operational risks caused by manual repetitive operations and information synchronization between two systems. Ultimately, this design allows the processing flow to be embedded into the enterprise's existing management system, forming a complete automated closed loop, improving work order processing efficiency while also enhancing the standardization of the operation and maintenance process.

[0139] The above embodiments have described in detail the key steps of the automatic work order processing method combined with semantic expansion provided by the embodiments of this application from the perspective of method steps. To more intuitively and completely demonstrate the overall architecture of the technical solution of this application and the collaborative workflow between various modules, the following will combine... Figure 2 A systematic introduction will be provided. Figure 2 This is a schematic diagram of the overall architecture and processing flow of a work order processing system provided in an embodiment of this application.

[0140] like Figure 2 As shown, this system architecture corresponds to the aforementioned method embodiments, fully realizing a closed-loop process from work order input to final response. The system architecture specifically includes the following modules and processes:

[0141] Work order issue input module: As the system entry point, it receives original work order issues submitted by external work order management platforms or users, and automatically extracts the contextual information contained therein (such as server model, operating system version, error log fragments, etc.).

[0142] Question generalization module (corresponding to semantic expansion module): This module uses a pre-trained semantic model (such as BERT) and combines it with a domain knowledge base to perform deep semantic analysis and expansion on the original question, generating an equivalent question set covering multiple semantic types (such as synonym substitution, causal inference, and scene visualization) to improve the coverage of subsequent retrieval.

[0143] The RAG knowledge base query module (corresponding to the semantic matching module) connects to a vectorized knowledge base built on historical work orders. It uses the question set output by the question generalization module as query input and performs semantic similarity matching. Based on a preset or dynamically adjusted similarity threshold, it determines whether the match is successful or unsuccessful, and accordingly decides the process branch.

[0144] Data Acquisition Module (corresponding to some functions of the Data Acquisition and Analysis Module): This module is triggered when the knowledge base fails to match. Based on the work order context information, it calls various pre-configured API interfaces to automatically collect real-time, structured runtime data from related systems (such as servers, log centers, and databases).

[0145] The logic processing module (corresponding to another function of the data acquisition and analysis module): This module integrates a pre-trained analysis model (such as a large language model), receives structured data from the data acquisition module, and combines it with the original work order description to perform comprehensive analysis, reasoning and root cause localization, and finally generate a structured preliminary solution.

[0146] Manual confirmation stage (corresponding to the review interaction module): This stage involves submitting the preliminary solution (whether from knowledge base matching or generated by the logic organization module) to support personnel for final decision-making through a mandatory pop-up review interface. Support personnel must perform one of the following actions: "Confirm," "Modify," or "Regenerate." The system advances the process accordingly, ensuring the mandatory nature and reliable closed-loop of manual review.

[0147] Responding to a work order (corresponding output module): After manual confirmation, the system outputs the final solution and executes a response to the work order. This operation submits the solution to an external work order management platform to close or update the status of the corresponding work order, thereby completing the end-to-end automated process from intelligent analysis to business closure.

[0148] like Figure 2 As shown, the modules work together in sequence, and through the dual-branch judgment logic of "matching successfully" and "matching failed", they form an efficient, reliable and self-learning work order processing system.

[0149] The electronic device provided in this application embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0150] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the methods in any of the above method embodiments.

[0151] All or part of the steps in the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof.

[0152] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

[0156] In this application, the term "comprising" and its variations can refer to non-limiting inclusion; the term "or" and its variations can refer to "and / or". The terms "first", "second", etc., in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0157] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0158] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0159] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0160] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0161] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0162] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for automatically processing work orders combining semantic expansion, characterized in that, The method includes: Obtain the original work order description; The original work order description is semantically expanded to generate a set of questions containing at least one expanded question. Based on a preset similarity threshold, the set of questions is semantically matched with a historical solution knowledge base to obtain the matching results; In response to the matching result indicating a successful match, the matched historical solution is used as the preliminary solution; In response to the matching result indicating a matching failure, based on the original work order description, real-time operational data related to the original work order description is collected from at least one associated system, and the original work order description and the real-time operational data are input into a pre-trained analysis model to generate a preliminary solution. Receive a first instruction from the review interface for the preliminary solution, the first instruction including at least one of confirmation, modification, and regeneration instructions; Based on the first instruction, output the final solution.

2. The method according to claim 1, characterized in that, The semantic expansion processing of the original work order description includes: A pre-trained semantic model is used to process the original work order description to generate at least one extended question covering multiple semantic types; wherein, the multiple semantic types include at least one of synonym substitution, cause inference and scene visualization.

3. The method according to claim 2, characterized in that, The process of using a pre-trained semantic model to process the original work order description also includes: Obtain the context information associated with the original work order description; Based on the original work order description and the context information, the semantic model generates at least one extended question; wherein the context information includes at least one of device hardware parameters, system version information, and operation log fragments.

4. The method according to any one of claims 1-3, characterized in that, The method further includes: Based on the matching result and the first instruction, generate feedback information related to the problem type to which the original work order description belongs; Based on the feedback information, the similarity threshold used for semantic matching of similar question types is updated.

5. The method according to claim 1, characterized in that, The process of collecting real-time operational data related to the original work order description from at least one associated system includes: Based on the context information associated with the original work order description, a target interface combination is selected from a plurality of pre-configured application interfaces; Data collection is performed by calling the target interface combination; if the call fails, a predefined exception handling strategy is executed, which includes enabling a backup interface or performing a retry operation.

6. The method according to claim 1, characterized in that, Following the output of the final solution, the following is also included: The final solution is submitted to the work order management platform to close or update the work order status corresponding to the original work order description; wherein, the work order management platform is an external business system used to manage the lifecycle of work orders.

7. A work order processing system, characterized in that, The system includes: The semantic extension module is used to perform semantic extension processing on the original work order description to generate a set of questions containing at least one extended question. The semantic matching module is used to perform semantic matching between the set of questions and the historical solution knowledge base; The data acquisition and analysis module is used to collect real-time operation data from at least one related system based on the original work order description when the semantic matching module fails to match, and to generate a preliminary solution based on the original work order description and the real-time operation data; wherein, the preliminary solution is a historical solution matched by the semantic matching module, or a solution generated by the data acquisition and analysis module; The review interaction module is used to output the preliminary solution to the review interface and receive the first instruction from the review interface for the preliminary solution. The output module is used to output the final solution according to the first instruction.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

Images