Document verification method and system based on MCP technology

Abstract

The invention discloses a document verification method and system based on an MCP technology, and the method comprises the following steps: analyzing a technical document through a natural language processing technology, carrying out the atomization decomposition of operation steps in the technical document into a plurality of basic steps, and extracting an expected result of each basic step; constructing an interactive verification scene based on the basic steps; in the interactive verification scene, according to whether the MCP service corresponding to the basic step is found or not, dynamically selecting to be actually executed by the MCP service or to be simulated by the large language model to execute so as to obtain an actual execution result; and comparing the actual execution result with the expected result to generate a verification result. According to the technical document verification method and device, full-process automation of technical document verification is achieved, the verification dimension extends to the operability level from the grammar level, the verification efficiency, accuracy and universality are remarkably improved, and meanwhile the resource cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of computer software technology, and in particular to a document verification method and system based on MCP technology. Background Technology

[0002] In enterprise R&D, the quality of technical documents such as "Software Installation Guides" and "System Configuration Manuals" is crucial, as their accuracy and operability directly impact team collaboration efficiency and project progress speed. Currently, the verification of technical documents is mainly achieved through manual review, format checking tools, and automated test scripts.

[0003] In the process of developing the existing technology, the inventors discovered that: Human review requires specific knowledge and personal experience to provide suggestions, making it inefficient and susceptible to subjective biases from technical personnel. Formatting tools can only identify surface-level issues and cannot verify the logical correctness or practical executability of the document content. Automated test scripts are costly to build and maintain, lack versatility, and struggle to fully reproduce complex user scenarios. Both formatting tools and automated test scripts offer single-source verification with low accuracy.

[0004] Based on the above, this application provides an interactive technical document verification scheme with high verification efficiency, high accuracy, and wide applicability, thereby systematically solving the problems of low verification efficiency, poor accuracy, and limited applicability in the prior art. Summary of the Invention

[0005] This application provides an interactive technical document verification scheme with high verification efficiency, high accuracy, and wide applicability, thereby systematically solving the problems of low verification efficiency, poor accuracy, and limited applicability in the prior art.

[0006] Specifically, a document verification method based on MCP technology includes the following steps: Document parsing steps: The technical document is parsed using natural language processing technology, the operation steps in the technical document are atomically decomposed into multiple basic steps, and the expected result of each basic step is extracted; Verification environment generation steps: Based on the aforementioned basic steps, construct an interactive verification scenario; Verification execution steps: In the interactive verification scenario, depending on whether the MCP service corresponding to the basic steps is found, the system dynamically selects whether to actually execute the steps using the MCP service or to simulate them using the large language model, in order to obtain the actual execution results. Result feedback step: Compare the actual execution result with the expected result to generate a verification result.

[0007] Furthermore, the atomization decomposition includes: The operation steps are broken down to the granularity of a single command or a single operation, wherein each of the basic steps corresponds to a single command or a single operation.

[0008] Furthermore, the verification execution step includes: The basic steps are semantically understood and rewritten using a large language model to match the corresponding MCP service.

[0009] Furthermore, the verification execution step is simulated by a large language model, including: Based on the internal knowledge graph pre-set by the large language model, the execution path of the inference command is simulated to generate interactive output containing success prompts, error messages, or system path structures.

[0010] Furthermore, the verification execution step is actually performed by the MCP service, including: A standardized request is constructed using the MCP client and sent to the corresponding MCP server; The MCP server executes the specific system tasks corresponding to the operation steps and returns the actual execution results; The MCP server is pre-registered and configured according to functional modules.

[0011] Furthermore, the functional modules include at least one of the following: a system operation module, a network communication module, a graphical interface module, and a third-party tool integration module.

[0012] Furthermore, the interactive verification scenario provides a graphical user interface for visually displaying to the user the execution process of the operation steps, the actual execution results, and the comparison with the expected results.

[0013] Furthermore, the result feedback step includes: After each step is executed, the actual execution results are compared with the expected results in real time. When the comparison results are inconsistent, an error log is recorded in real time and the user is notified.

[0014] Furthermore, the verification results include a verification report, which contains document improvement suggestions generated based on the comparison results and error logs.

[0015] This application also provides a document verification system based on MCP technology, including: The document parsing module is used to parse technical documents using natural language processing technology, atomize the operation steps in the technical documents into multiple basic steps, and extract the expected results of each basic step. The verification environment construction module is used to construct an interactive verification scenario based on the aforementioned basic steps. An automated execution module is used in the interactive verification scenario to dynamically select whether to actually execute the MCP service or simulate the execution by the large language model based on whether an MCP service corresponding to the basic steps is found, so as to obtain the actual execution result. The feedback analysis module is used to compare the actual execution results with the expected results and generate verification results.

[0016] The technical solution provided in this application has at least the following beneficial effects: By automatically parsing documents and atomically decomposing operation steps using natural language processing technology, and combining the automated execution engine of MCP service and LLM simulation, the entire process of technical document verification has been automated, greatly improving verification efficiency. By extending the verification dimension from the format and syntax level to the operability level of the document content, and through the comparison mechanism between the actual results and the expected results, it is possible to accurately discover deep-seated problems such as infeasible steps and incorrect order, which significantly improves the accuracy of verification. By combining real execution of MCP services with LLM simulation, we can obtain feedback from real environments and cope with complex environment configurations, breaking the environment dependence of traditional automation scripts and greatly enhancing the system's adaptability. Using large language models for logic-level simulation eliminates the need to create a real physical or virtual environment for each verification task, significantly reducing computational resource consumption and environment maintenance costs. By providing a visual demonstration of the execution process and real-time result comparison through interactive verification scenarios, the verification process becomes transparent, greatly improving user experience and problem diagnosis efficiency. This invention comprehensively improves the technical level of technical document verification from multiple dimensions such as efficiency, accuracy, universality, cost, and user experience. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a document verification method based on MCP technology provided in this application embodiment.

[0018] Figure 2 This is a schematic diagram of a document verification system based on MCP technology provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Please refer to Figure 1 This application discloses a document verification method based on MCP technology, including the following steps: S110: Document parsing steps: The technical document is parsed using natural language processing technology, the operation steps in the technical document are atomized and decomposed into multiple basic steps, and the expected results of each basic step are extracted.

[0021] Specifically, after receiving the natural language processing (NLP) documents uploaded by users, the system parses them using NLP techniques. First, it extracts the core operational steps from the document, then performs atomized decomposition, breaking down the complex operational steps into multiple basic steps.

[0022] Atomization decomposition is the process of breaking down the operation steps into the granularity of a single command or single operation, wherein each basic step corresponds to a single command or single operation, that is, each basic step represents an indivisible single command or single operation and has a clear expected result.

[0023] For example, taking "downloading Eclipse" as the core operation step, it can be atomized into: Basic Step 1: Open the Firefox browser. Expected Result: The browser window will open successfully on the desktop.

[0024] Basic Step 2: Enter "https: / / www.eclipse.org / downloads / " in your browser's address bar. Expected Result: The official Eclipse website will be displayed.

[0025] Basic Step 3: Click the Download button on the page and select the installation package that matches your local computer's CPU architecture. Expected Result: Successfully locate the download dropdown and select the correct version.

[0026] Basic Step 4: Double-click the downloaded installation package file. Expected Result: The installer will launch, and a new Eclipse icon will be generated on the desktop.

[0027] S120: Verification Environment Generation Steps: Based on the basic steps described above, construct an interactive verification scenario.

[0028] The interactive verification scenario is dynamically created based on the basic step sequence generated in step one.

[0029] Furthermore, the interactive verification scenario provides a graphical user interface to visually demonstrate to the user the execution process of the operation steps, the actual execution results, and the comparison with the expected results.

[0030] S130: Verification Execution Steps: In the interactive verification scenario, depending on whether an MCP service corresponding to the basic steps is found, dynamically select whether to actually execute the steps using the MCP service or simulate them using the large language model to obtain the actual execution results.

[0031] Specifically, MCP, in this application, refers to the Model Context Protocol, which is an open protocol that allows applications to interact securely and in a standardized manner with tools and services surrounding the Large Language Model (LLM).

[0032] In the context of this invention, an MCP service refers to a modular functional unit encapsulated and registered according to this protocol for performing specific system tasks. Each MCP service provides one or more standardized operation interfaces, such as: System operation service: providing the ability to perform system-level tasks such as "installing software," "creating folders," and "running commands." Network communication service: providing network interaction functions such as "API calls," "web page navigation," and "DNS resolution." Graphical interface service: providing the ability to simulate user interaction with a graphical interface, such as "clicking buttons," "entering text," and "reading screen elements." Third-party tool service: providing operation interfaces for specific third-party tools (such as Eclipse, MySQL, Docker, etc.).

[0033] First, the step descriptions are semantically understood and rewritten using a large language model, and then mapped to standardized MCP service calls. For example, "Open Firefox browser" is mapped to the "Launch application" service of the graphical interface module. "Enter URL" is mapped to the "Web navigation" service of the network communication module.

[0034] Then, the execution path is dynamically selected based on the matching results: When the corresponding MCP service is found, the MCP service performs the actual execution, including: constructing a standardized request through the MCP client and sending it to the corresponding MCP server; the MCP server executes the specific system task corresponding to the operation steps and returns the actual execution result; the MCP server is pre-registered and orchestrated according to functional modules. Functional modules include system operation modules, network communication modules, graphical interface modules, and third-party tool integration modules, etc. After the server executes the specific system task, it returns the actual execution result. When the corresponding MCP service cannot be found, the system automatically triggers the simulation execution function of the large language model, including: inferring the execution path of the command based on the internal knowledge graph preset by the large language model, and simulating the generation of interactive output containing success prompts, error messages or system path structures.

[0035] For example, for the command `mvn clean install`, in an environment without an MCP service, a large model might simulate the following output: [INFO] Scanning for projects... [ERROR] Could not find pom.xml in current directory; This simulated output is not the actual result of the operation, but it is sufficient to advance the process and verify the document logic in a verification environment.

[0036] S140: Result feedback step: Compare the actual execution result with the expected result to generate a verification result.

[0037] Specifically, after each step is executed, the actual execution result is compared with the expected result in real time; when the comparison result is inconsistent, an error log is recorded in real time and the user is notified.

[0038] The actual execution results are compared with the expected results in real time. For example, in the "Open Firefox browser" step, the browser process status returned by the MCP service is compared with the expected result of "browser window opened successfully".

[0039] When the comparison results are inconsistent, the system immediately records a detailed error log, including the step number, error type, and the difference between the actual result and the expected result.

[0040] After all basic steps are completed, a verification result is generated. This verification result includes a verification report containing document improvement suggestions based on the comparison results and error logs. For example, the report might state: "In step 3, the download link is invalid; it is recommended to update to the latest download address" or "Step 4 lacks a description of the prerequisite requirements for system administrator privileges."

[0041] Please refer to Figure 2 This application also provides a document verification system based on MCP technology, including: The document parsing module 10 is used to parse technical documents using natural language processing technology, atomize the operation steps in the technical documents into multiple basic steps, and extract the expected results of each basic step.

[0042] Verification environment construction module 20 is used to construct an interactive verification scenario based on the basic steps described above.

[0043] The automated execution module 30 is used in the interactive verification scenario to dynamically select whether to actually execute the MCP service or simulate the execution by the large language model based on whether the MCP service corresponding to the basic steps is found, so as to obtain the actual execution result.

[0044] The feedback analysis module 40 is used to compare the actual execution result with the expected result and generate a verification result.

[0045] The system modules are connected sequentially: the output of the document parsing module 10 is connected to the input of the verification environment construction module 20; the output of the verification environment construction module 20 is connected to the input of the automated execution module 30; and the output of the automated execution module 30 is connected to the input of the feedback analysis module 40. The automated execution module 30 includes an MCP service invocation unit and a large language model simulation unit, dynamically selecting the execution path based on the MCP service matching results. The MCP service invocation unit is connected to a cluster of MCP servers pre-registered and orchestrated according to functional modules.

[0046] This application provides a computer device, characterized in that it includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a document verification method based on MCP technology.

[0047] This application also provides a computer-readable storage medium storing computer-readable instructions, which, when executed by a computer's processor, cause the computer to perform a method in a document verification method based on MCP technology.

[0048] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0049] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A document verification method based on MCP technology, characterized in that, Includes the following steps: Document parsing steps: The technical document is parsed using natural language processing technology, the operation steps in the technical document are atomically decomposed into multiple basic steps, and the expected result of each basic step is extracted; Verification environment generation steps: Based on the aforementioned basic steps, construct an interactive verification scenario; Verification execution steps: In the interactive verification scenario, depending on whether the MCP service corresponding to the basic steps is found, the system dynamically selects whether to actually execute the steps using the MCP service or to simulate them using the large language model, in order to obtain the actual execution results. Result feedback step: Compare the actual execution result with the expected result to generate a verification result.

2. The method according to claim 1, characterized in that, The atomization decomposition includes: The operation steps are broken down to the granularity of a single command or a single operation, wherein each of the basic steps corresponds to a single command or a single operation.

3. The method according to claim 1, characterized in that, The verification execution steps include: The basic steps are semantically understood and rewritten using a large language model to match the corresponding MCP service.

4. The method according to claim 1, characterized in that, The verification execution step is simulated by a large language model and includes: Based on the internal knowledge graph pre-set by the large language model, the execution path of the inference command is simulated to generate interactive output containing success prompts, error messages, or system path structures.

5. The method according to claim 1, characterized in that, The verification execution step is actually executed by the MCP service, including: A standardized request is constructed using the MCP client and sent to the corresponding MCP server; The MCP server executes the specific system tasks corresponding to the operation steps and returns the actual execution results; The MCP server is pre-registered and configured according to functional modules.

6. The method according to claim 5, characterized in that, The functional modules include at least one of the following: system operation module, network communication module, graphical interface module, and third-party tool integration module.

7. The method according to claim 1, characterized in that, The interactive verification scenario provides a graphical user interface to visually demonstrate to the user the execution process of the operation steps, the actual execution results, and the comparison with the expected results.

8. The method according to claim 1, characterized in that, The result feedback step includes: After each step is executed, the actual execution results are compared with the expected results in real time. When the comparison results are inconsistent, an error log is recorded in real time and the user is notified.

9. The method according to claim 1, characterized in that, The verification results include a verification report, which contains document improvement suggestions generated based on the comparison results and error logs.

10. A document verification system based on MCP technology, characterized in that, include: The document parsing module is used to parse technical documents using natural language processing technology, atomize the operation steps in the technical documents into multiple basic steps, and extract the expected results of each basic step. The verification environment construction module is used to construct an interactive verification scenario based on the aforementioned basic steps. An automated execution module is used in the interactive verification scenario to dynamically select whether to actually execute the MCP service or simulate the execution by the large language model based on whether an MCP service corresponding to the basic steps is found, so as to obtain the actual execution result. The feedback analysis module is used to compare the actual execution results with the expected results and generate verification results.

Images