Intelligent auxiliary bid evaluation method and system for electric power bid invitation major

Abstract

The invention discloses an intelligent auxiliary bid evaluation method and system for an electric power bid invitation specialty, and the method comprises the steps: obtaining uploaded purchase information, and creating purchase batches based on configured basic parameters and purchase review rules, so as to generate a review task list with a unique batch number; and based on the review task list and the configured basic parameters, extracting context information of each review task to construct context feature vectors of the review tasks. And matching the context feature vector with a preset negative rule to obtain a negative suggestion list corresponding to each review task in the review task list. And binding the negative suggestion list with the corresponding bidding task, and pushing the bidding task bound with the negative suggestion list to the user terminal. According to the technical scheme, the user terminal receives the operation of the user on each negative suggestion so as to generate the final negative suggestion, and the final negative suggestion is stored and archived, so that the negative suggestion generation efficiency is improved, the negative expression is standardized, and the work intensity of experts is reduced.

Description

Technical Field

[0001] The present invention relates to the field of electric power engineering technical services, and in particular to an intelligent auxiliary bid evaluation method and system for electric power bidding. Background Art

[0002] As the scale of procurement of power industry engineering projects continues to expand, power bidding and evaluation work presents the characteristics of large workload, complex standards and tight time. In order to ensure the fairness and compliance of the evaluation work, the tenderer needs to organize evaluation experts with professional backgrounds to conduct a comprehensive evaluation of the bidding documents in terms of technology, business and qualifications. At present, there are the following prominent problems in the actual evaluation process in the field of power engineering: (1) The evaluation criteria are complex and the application of veto clauses is not uniform: Since the bidding documents corresponding to different batches and bidding sections have different document structures and inconsistent content expressions, the evaluation experts often rely on their personal experience in making judgments, which may easily lead to situations such as unclear reference clauses in veto opinions and misjudgment of applicable conditions.

[0003] (2) The text of the veto opinion is not standardized and the burden of report review is heavy: the current bid evaluation system mostly uses semi-structured input, and the bid evaluation experts use free writing to fill in the veto opinion, resulting in inconsistent wording style and confusing logical structure.

[0004] (3) Fragmented management of veto rules and lack of dynamic adaptation capabilities: The bidding unit has not formed a standardized rule base, and the bid evaluation system is unable to perform dynamic matching and personalized recommendations based on specific contexts, making it difficult to effectively reuse existing knowledge.

[0005] (4) The review process lacks guidance and the experts’ operation efficiency is low: Faced with a large number of bidders and complex terms, the review experts are inefficient in reviewing, judging and writing opinions one by one in the system, which easily leads to risks such as fatigue review, wrong rejection or missed rejection.

[0006] To sum up, there is an urgent need to design an intelligent assisted bid evaluation method and system suitable for power bidding business scenarios, which can intelligently match structured veto rules on the basis of fully understanding procurement requirements and bid evaluation context, automatically generate standardized veto opinion templates, guide experts to complete the review tasks efficiently, and improve the standardization, accuracy and information level of bid evaluation work. Summary of the invention

[0007] The present invention discloses an intelligent assisted bid evaluation method and system for electric power bidding, aiming to solve technical problems in existing electric power bidding and evaluation work, such as non-standard application of veto clauses, low efficiency in filling out evaluation opinions, and uneven quality of bid evaluation report generation.

[0008] In a first aspect, the present disclosure provides an intelligent assisted bid evaluation method for power bidding, the method comprising: Obtain the uploaded procurement information, and create procurement batches based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number; Based on the review task list and the configured basic parameters, extract the context information of each review task to construct a context feature vector of the review task; Match the context feature vector with the preset veto rules to obtain a list of veto suggestions corresponding to each review task in the review task list; Bind the list of veto suggestions to the corresponding bidding tasks, and push the bidding tasks with the bound list of negative suggestions to the user terminal; Receive the operations of the user on each veto suggestion through the user terminal to generate a final veto opinion, and store and archive the final veto opinion.

[0009] Further, the obtaining the uploaded procurement information, and creating procurement batches based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number includes: Obtain the procurement requirement form, the judge information form, and the procurement document. The procurement information includes the procurement requirement form, the judge information form, and the procurement document; wherein, the procurement requirement form contains multiple section information, the judge information form contains the personal information of multiple expert judges, and the procurement document contains multiple veto clauses; Configure the basic parameters based on the procurement information; the basic parameters include procurement category, evaluation model version, veto format, and judge authority.

[0010] Further, the obtaining the uploaded procurement information, and creating procurement batches based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number further includes: Parse the procurement document to obtain the procurement review rules, and create the procurement batches based on the procurement review rules; Determine the allocated batch numbers for each batch based on the procurement batches, and generate a subcontracting structure list, review task items, and an initial veto clause library based on the allocated batch numbers; Write the subcontracting structure list, review task items, and the initial veto clause library into the context initialization parameter table to generate the review task list.

[0011] Further, the extracting the context information of each review task based on the review task list and the configured basic parameters to construct a context feature vector of the review task includes: Extract structured information from the configured basic parameters; the structured information includes at least batch number, package number in the subcontracting structure list, equipment type, qualification clause number, maximum bid price limit, qualification type of the bidding unit, identity information of the evaluation experts, and bidding unit number; Perform semantic analysis on the unstructured text paragraphs in the bidding documents through natural language parsing and keyword extraction techniques to identify the technical test requirements corresponding to the equipment type and the guiding language from the subjective judgments of the evaluation experts from the bidding documents; Extract keyword tags from the technical test requirements and guiding language, and perform standardization processing on the keyword tags to obtain semantic signal fields; Integrate the semantic signal fields, keyword tags, and structured information, and encode to generate the context feature vector of the evaluation task.

[0012] Further, match the context feature vector with the preset veto rules to obtain a list of veto suggestions corresponding to each evaluation task in the evaluation task list, including: Extract the structured set of veto rules from the veto rule database; each veto rule in the set of veto rules includes evaluation standard category, rule label, reference clause or document basis, veto scenario type, and veto template; Call the preset semantic similarity matching algorithm to compare the keyword fields labeled with keyword tags in the context feature vector with the veto rules to obtain the veto rule entries with a similarity exceeding the first threshold; Based on the identity information of the evaluation experts, call the semantic similarity matching algorithm to screen out the veto rule entries with a technical standard similarity exceeding the second threshold corresponding to the identity information of the evaluation experts; Sort the screened veto rule entries according to the similarity to generate the list of veto suggestions corresponding to the scenarios of each evaluation task; Among them, the list of veto suggestions includes veto clause reference, veto reason, parameter fields, and text suggestion paragraphs.

[0013] Further, bind the list of veto suggestions to the corresponding bidding task, and push the bidding task bound with the list of negative suggestions to the user terminal, including: Obtain the bidding task to be bound, and determine the list of veto suggestions with a matching degree exceeding the third threshold with the evaluation task based on the bidding task; Bind the list of veto suggestions to the bidding task through the task binding engine to obtain the binding result, and write the binding result into the task table in JSON format; Push the list of veto suggestions to other user terminals of technical review experts corresponding to the review tasks through the user terminal.

[0014] Further, receiving, by the user terminal, operations of the user on each veto suggestion to generate a final veto opinion, and storing and archiving the final veto opinion, includes: Obtaining, by the user terminal, interaction information between the review expert and the operation panel; the interaction information includes adopting a veto suggestion, modifying a veto suggestion, and ignoring a veto suggestion; Obtaining a final status record of each veto suggestion based on the interaction information, and converting the final status record into structured fields to generate the final veto opinion; the structured fields include the original allocation number of the veto suggestion, interaction information, interaction content, citation basis, parameter filling value, timestamp, and identity information of the review expert.

[0015] In a second aspect, the present disclosure provides an intelligent auxiliary bid evaluation system for the power bidding profession, which is used to implement the intelligent auxiliary bid evaluation method for the power bidding profession described in the first aspect. The system includes: A review task list generation module, configured to obtain uploaded procurement information, and create a procurement batch based on configured basic parameters and procurement review rules to generate a review task list with a unique batch number; A feature vector construction module, configured to extract context information of each review task based on the review task list and configured basic parameters to construct a context feature vector of the review task; A veto rule matching module, configured to match the context feature vector with a preset veto rule to obtain a list of veto suggestions corresponding to each review task in the review task list; A binding and pushing module, configured to bind the list of veto suggestions to the corresponding bidding task, and push the bidding task bound with the list of negative suggestions to the user terminal; A veto opinion generation module, configured to receive, by the user terminal, operations of the user on each veto suggestion to generate a final veto opinion, and store and archive the final veto opinion.

[0016] In a third aspect, the present disclosure provides an electronic device, including: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method described in the first aspect of the present disclosure.

[0017] Fourthly, the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the method described in the first aspect of the present disclosure when executed by a processor.

[0018] The beneficial effects of the present disclosure are as follows. Compared with the prior art, the present disclosure has the following advantages: (1) Improve the bid evaluation efficiency and the operation experience of experts. By automatically generating a visual list of rejection suggestions, a human-computer interaction mode of "one-key insertion + parameter filling + quick confirmation" is realized, significantly reducing the manual writing burden of experts and saving the evaluation time when facing the intensive evaluation tasks of a large number of bidders.

[0019] (2) Ensure the standardization and compliance of the citation of rejection clauses. The present invention uses "review context feature vectors" as the retrieval key to accurately call and match clauses from the rejection rule library, and automatically attach the citation of regulations and the source of procurement documents, effectively avoiding the phenomena of incorrect citation and missed citation, and improving the auditability and legal support of rejection opinions.

[0020] (3) Build a knowledge precipitation and reuse mechanism for rejection rules. All rejection opinions are stored through structured fields, realizing the automatic accumulation of historical opinions, rule application frequencies and expert usage preferences, providing a basis for model recommendations for subsequent similar projects, and supporting the precipitation and continuous optimization of standard clauses at the organizational level.

[0021] (4) Enhance the visual supervision and process controllability of the evaluation process. The system records in real time the generation source, operation path and expert confirmation status of each rejection opinion, and has the data governance ability of "traceable, statable and analyzable throughout the process", providing a data support basis for bidding agents, supervision groups and legal personnel.

[0022] (5) Have cross-project adaptability and multi-rule compatibility capabilities. It supports dynamically adjusting the priority of rejection rules and the template structure according to context information such as different procurement categories, bid evaluation stages, and evaluation dimensions, and is applicable to large-scale, high-frequency, and multi-type power procurement bid evaluation operations, with good versatility and scalability.

[0023] (6) The present invention constructs the context feature vector C as a context graph with a directed topological structure, and explicitly models the dependency relationships among its nodes (fields / terms). Correspondingly, the veto rule should also have its own logical premise chain and applicable boundary, which are also expressed as rule topological paths. When performing matching, the system should introduce a "topological consistency comparison mechanism for the context graph and the rule graph", that is, compare whether the execution paths, stage orders, and preconditions of the two are consistent. If there are conflicts (such as the context being executed before the rule expects, or missing preconditions), the automatic push of the veto suggestion should be suspended, and it should be transferred to an expert for manual confirmation through a system prompt. The boundary between AI and expert judgment is determined, enhancing the robustness of the system.

[0024] Generally speaking, the intelligent auxiliary bid evaluation method and system for the power bidding profession provided by the present invention construct a review context feature vector, perform semantic matching with a structured veto rule library, generate a list of veto suggestions, and push them to the expert review terminal, supporting one-click adoption, semantic modification, and parameter completion to form a final veto opinion. The final opinion is stored in a structured manner and automatically archived into the preliminary review report, realizing the standardized adaptation of veto clauses, the efficient guidance of expert operations, and the traceable management of the entire review process. It is applicable to large-scale power procurement bid evaluation scenarios and has good adaptability and promotion value. Brief Description of the Drawings

[0025] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure.

[0026] Figure 1 It is a schematic flow chart of the intelligent auxiliary bid evaluation method for the power bidding profession provided by the present disclosure; Figure 2 It is a schematic structural diagram of the intelligent auxiliary bid evaluation system for the power bidding profession provided by the present disclosure; Through the above accompanying drawings, the clear embodiments of the present disclosure have been shown, and there will be more detailed descriptions hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present disclosure in any way, but to explain the concept of the present disclosure to those skilled in the art by referring to specific embodiments. Detailed Embodiments

[0027] The following further describes the present disclosure with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present disclosure and should not be used to limit the protection scope of the present disclosure.

[0028] Figure 1 A schematic flow chart of an intelligent auxiliary bid evaluation method for the power bidding profession provided by an embodiment of the present disclosure. See Figure 1, the following specifically elaborates on each step in combination with this embodiment.

[0029] S100. Obtain the uploaded procurement information, and create a procurement batch based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number.

[0030] In the present invention, it first supports the bidding agency to upload procurement-related information, including the list of requirements, the list of bidders, the list of reviewers, and procurement documents, etc. Through structured parsing, it extracts section information, technical parameters, review terms, and rejection rule entries; subsequently, the user sets basic parameters in the system, such as procurement category, evaluation model, expert role permissions, and key review fields (such as delivery date, upper price limit, qualification requirements, etc.); the system combines the set parameters with the parsing results, automatically creates a procurement batch according to the procurement review rules, generates a unique batch number, initializes the subcontracting structure, generates a review task list, and completes the expert task assignment, realizing the integrated management from procurement information collection to batch configuration, and providing a standardized starting point for the subsequent review process and the intelligent generation of rejection opinions.

[0031] In a typical implementation scenario, a provincial power company of the State Grid plans to carry out the "centralized procurement of 10kV cable accessories" and entrusts the intelligent assisted bid evaluation system provided by the present invention for review management. The system realizes the full-process automatic configuration from information collection to batch initialization.

[0032] First, the bidding agency uploads the procurement documents through the batch creation module of the system, including: Procurement requirement form (.xls): It contains multiple section information, such as section number, equipment name, specification model, quantity, delivery date and other fields. For example: section number: SG2025 - B05 - 001; equipment name: cold shrink cable terminal; quantity: 320 sets; delivery date: before September 30, 2025. Reviewer information form (.csv): It contains expert name, affiliated unit, professional field, whether it is the team leader, etc. Procurement document (.docx / .pdf): The system automatically parses the review method therein and extracts the rejection clauses, such as: "If the valid product type test report is not provided, its bid should be rejected", "If the bid price is higher than 10% of the highest price limit, it should be rejected".

[0033] Based on the uploaded file, the system automatically performs the following basic parameter setting operations: Set the procurement category to "Power Transmission and Distribution Materials"; Set the evaluation model version to "JS-DL-PINGBIAO-V4.0"; Set the rejection method to "Rule Structured Matching + Expert Confirmation"; Configure the reviewer permissions to "Group Review + Write Opinions by Role"; Enable system fields such as "Quotation Parameters", "Delivery Date Parameters", and "Qualification Certificate Number" as referenceable parameters. After uploading the above information and setting the parameters, the system automatically creates a procurement batch based on the parsed review rules and batch structure, assigns the batch number DL2025-05-B1, and generates the following resources: Subcontract structure list (each bid section corresponds to a package number); 96 review task items, which are automatically grouped by expert type and package number by the system; Initial rejection clause library, written to the veto_report_rule table; Context initialization parameter table, for constructing the review context feature vector C later.

[0034] S200. Extract the context information of each review task based on the review task list and the configured basic parameters to construct the context feature vector of the review task.

[0035] In a certain centralized procurement project of "10kV Cold Shrink Cable Terminals", after the system completes the creation of the procurement batch, it enters step S200, that is, "Extract the context information of each review task based on the review task list and the configured basic parameters to construct the context feature vector of the review task". This step is automatically executed by the system, aiming to integrate and encode the structured and unstructured information related to the current review task into a standardized context vector C for subsequent intelligent rejection rule matching.

[0036] First, the system extracts structured information from the configuration data of the procurement batch, including fields such as batch number, package number, equipment type, qualification clause number cited in the bidding document, maximum price limit of the package, required qualification type, role category of the current review expert, and bidder unit number. For example, for this review task, the structured information collected by the system includes: batch number "DL2025-05-B1", package number "P01", equipment type "Cold Shrink Terminal Head", corresponding qualification clause number "QL-04", maximum bid price limit of 215,000 yuan, requiring the bidder to provide a type test report certified by CNAS, the current review expert is a technical review expert, and the current review task is for the bidder "JS-Supplier-xxx".

[0037] Secondly, through natural language parsing and keyword extraction technologies, the system performs semantic analysis on unstructured text paragraphs such as the evaluation methods and technical requirements in the tender documents. The system identifies the technical test requirements related to the equipment from the documents, such as "need to pass the thermal aging test and voltage breakdown test", and the guiding words for the subjective judgment of the evaluation experts, such as "the judges can judge the authenticity of the test data based on past performance". The system extracts keyword tags based on this information, such as "thermal aging", "voltage breakdown", "historical performance", "authenticity verification", etc., and standardizes them into internal semantic signal fields uniformly.

[0038] Finally, the system integrates the above structured fields and semantic tags uniformly, and encodes to generate the context feature vector C corresponding to the current review task. This feature vector contains the unique identifier of the procurement project, equipment type, review package information, expert identity, price constraint, qualification requirements, key test identifier, whether historical performance verification is involved, and relevant keyword tags. This vector will be used as the input of the semantic matching module in the next step to drive the system to complete the targeted rejection suggestion push.

[0039] Through this step, the system successfully converts the context background of the review task into a standardized vector format, realizes the structured modeling of the review scenario, and provides complete and targeted context support for the subsequent rejection rule matching. This effectively avoids the inefficient operations of experts missing key information and manually consulting tender documents, and enhances the context relevance and accuracy of the rejection suggestion generation.

[0040] S300. Match the context feature vector with the preset rejection rules to obtain a list of rejection suggestions corresponding to each review task in the review task list.

[0041] In a specific embodiment, the system first extracts a structured set of rejection rules from the rejection rule database. Each rejection rule consists of the following elements: the corresponding review standard category (such as qualification review, business review, technical review), rule label (such as "test report not provided", "price limit exceeded", "qualification certificate does not match", etc.), legal provisions or procurement document basis cited, description of the rejection situation, standardized rejection statement template, and the equipment types, procurement categories or evaluation models to which this rejection clause applies.

[0042] The system uses a preset semantic similarity matching algorithm to compare each keyword field in the context vector C with the labels, keywords, and applicable conditions of the rejection rule entries one by one. For example, if the context feature vector C contains fields such as "equipment type is cold shrink terminal head", "need to provide a CNAS-certified test report", and "price ceiling is 215,000 yuan", the system will retrieve the following highly matching rule entries in the rejection rule set: 1. Bidders who fail to provide test reports with CNAS certification shall be rejected. 2. Bids with a tender offer higher than 10% of the maximum price limit for the section shall be rejected. 3. Bidders who fail to provide factory certificates of conformity for the corresponding product models shall be rejected.

[0043] The matching algorithm not only compares keyword tags but also considers the applicability of the bid evaluation roles. If the current review expert is a "technical review expert", the system will only push rejection recommendation entries related to technical standards; for legal review experts or commercial review experts, corresponding legal or price-related rejection rules will be pushed to avoid interference from irrelevant rules.

[0044] After screening and sorting, the system finally generates a "Rejection Recommendation List" that matches the current review task scenario. The content of this list includes: reference to rejection clauses, description of rejection reasons, relevant parameter fields (such as unfiled test report numbers, over-limit amounts, etc.), and recommended text insertion paragraphs.

[0045] This rejection recommendation list is then bound to the specific review task, and the system pushes it to the corresponding review expert interface for viewing, reference, or modification.

[0046] Through this step, the system completes an intelligent closed-loop from context semantic modeling to precise retrieval of rejection rules, effectively improving the efficiency and standardization of generating rejection opinions during the bid evaluation process, reducing the probability of experts missing key information, and ensuring the legality and context consistency of rule references.

[0047] S400. Bind the rejection recommendation list to the corresponding bidding task and push the bidding task with the bound rejection recommendation list to the user terminal.

[0048] In a specific embodiment, in this embodiment, the system has generated a rejection recommendation list that highly matches the current review context for a certain bidding task (bidding unit number JS-Supplier-xxx, package number P01) in the "10kV Cold Shrink Cable Terminal Procurement Project". This list contains the following typical content: 1. Rejection clause situation: Failure to provide a product type test report with CNAS certification. 2. Source of rejection basis: Article 4.2.3 of the "Qualification Review Measures". 3. Recommended rejection statement template: After verification, this bidder fails to provide a CNAS-certified type test report for the "cold shrink terminal head" product, which does not meet the requirements of qualification clause 4.2.3. It is recommended to reject. 4. Parameter fields: Test report number (parameter is empty), review expert remarks column (supports filling in text). 5. Applicable Expert Role: Technical Review Expert.

[0049] The system binds the list of veto suggestions to the bidding task DL2025-05-B1-P01-JS-Supplier-xxx through the "Task Binding Engine" and writes the binding result into the task table veto_report_judge_task in JSON format. The corresponding fields include task_id, judge_user_id, veto_suggestion_list, is_pushed (marked as "true"), etc.

[0050] Subsequently, the system actively pushes the above list of veto suggestions to the technical review experts of the current task through the expert review terminal (i.e., the bid evaluation expert interface). The expert interface adopts a modular layout, presenting the veto suggestions in the "Recommended Veto Column". The system supports the following interactive functions: 1. One-click Citation: After the expert clicks, the content of the suggestion can be directly filled into the review opinion. 2. Parameter Completion: If the suggestion template contains parameter fields (such as test report numbers), they can be filled in manually or automatically extracted by the system. 3. Explanation of Veto Reasons: The system reserves a text box to support experts in making supplementary explanations. 4. Attachment Upload: If experts need to provide supporting materials for the basis of veto (such as screenshots, comparison documents, etc.), they can upload them through the terminal. 5. Adoption / Ignoring Mark: Experts can mark whether each suggestion is adopted, which is convenient for the system to record the basis for judgment and the chain of responsibility.

[0051] This mechanism ensures that the veto suggestions are ready at the start stage of the review task and are submitted to experts for decision-making, avoiding the situations of missed or incorrect vetoes by experts due to lack of information, and effectively improving the standardization, closed-loop nature, and efficiency of the review process.

[0052] Through this step, the veto suggestions have achieved a closed-loop distribution from system generation to expert interaction, forming a complete mapping relationship of "task - suggestion - expert", laying a foundation for the finalization and archiving of the next-step veto opinions.

[0053] S500. Receive the operations of users on each veto suggestion through the user terminal to generate the final veto opinion, and store and archive the final veto opinion.

[0054] Taking the review task of the JS-Supplier-xxx tendering unit by technical review experts in the "DL2025-05-B1" batch as an example. After the expert opens the terminal interface and enters the "Recommendation and Rejection Suggestions" panel, the first thing seen is three rejection suggestions pushed by the system according to the context. The system defaults to attaching three operation options of "Adopt", "Modify", and "Ignore" after each suggestion.

[0055] In this implementation, the expert selects to adopt the first rejection suggestion: "CNAS-certified type test report not provided", and completes the test report number field in the structured template provided by the system (entering "not provided" when it is blank), and manually enters in the remarks column: "Upon verification, Article 4.2.3 of the tender document clearly requires, and no report upload record is found through automatic comparison by the system." Next, for the second suggestion "The quotation is 10% higher than the maximum price limit", the expert believes it should be more rigorous, so clicks the "Modify" operation. The system automatically expands the semantic editor, and the expert modifies the rejection statement from "Rejection should be made if it is 10% higher than the price limit" to "The tender quotation is 236,800 yuan, which has exceeded the maximum price limit of 215,000 yuan, not meeting the requirements for quotation validity, and it is recommended to be rejected", and quotes Section 5.1 of the "Quotation Validity Judgment" clause in the tender document.

[0056] For the third suggestion "Factory inspection certificate not provided", the expert finds that the unit has actually submitted a "Batch Certificate" PDF in the attachment materials, so clicks the "Ignore" button and adds a note: "The bidder has uploaded the certificate PDF, which is considered to meet the clause requirements and will not be rejected."

[0057] After all the above interactions are completed, the system records the final status of each rejection suggestion as a structured field, including information such as the original suggestion number, handling method (adopt / modify / ignore), expert input content, citation basis, parameter filling value, editing timestamp, review expert ID, etc. The finally generated rejection opinion is encapsulated into a standard JSON structure and written back to the fields review_comments and review_comments_params in the database table veto_report_detail, and marked as "finalized".

[0058] Through this step, the rejection opinion is transformed from the initial suggestion generated by the system into a final text with legality, personalized correction, and auditability through the expert's decision-making behavior, realizing a complete closed-loop from suggestion generation to expert finalization. This process not only retains the expert's subjective judgment space but also improves the consistency, standardization, and accountability of the rejection text.

[0059] In a specific embodiment, after the expert confirms the completion of the review task and clicks the "Submit Review" button, the system submits all the processed negative opinions to the backend database in the form of structured data. The specific storage operations include the following: The system stores each final negative opinion in the veto_report_detail table. The fields include: item_id: the unique number of the veto item; report_id: the corresponding tender evaluation grouping report number; bidder_name: the name of the bidder; veto_item: the situation of the veto clause; veto_regulations: the legal / purchasing document clauses cited; review_comments: the content of the final negative statement; review_comments_params: the parameters involved in the statement (such as numerical values, numbers, etc.); comment_user_id: the user ID of the filling expert; confirmed_time: the final confirmation submission time.

[0060] At the same time, the system marks the status of this review task as "submitted" in the veto_report_judge_task table and locks the record processed by this expert to prevent post-event tampering.

[0061] In the subsequent report generation process, the system triggers the "Report Arrangement Engine" to automatically aggregate the negative opinions submitted by all reviewers according to the report_id. The "List of Veto Items" section reserved in the report template will automatically call the above data content and classify and merge it according to the package number and bidder number.

[0062] In the generated preliminary review report, each negative opinion will be inserted in the following standard structure: veto item number, veto clause name (automatically extracted by the system, such as "test report not provided"), veto description statement (the complete negative opinion confirmed by the expert), basis of the veto clause (automatically attached with the corresponding tender clause number), name or number of the veto expert (displayed or not depending on the permission).

[0063] The preliminary review report document generated by the system supports export to Word (.docx) or PDF format, and at the same time supports synchronizing the structured review data back to the project main system for online review by roles such as supervision, legal, and team leader. Each negative opinion can be traced back to the original processing record, with a complete timestamp and chain of responsibility.

[0064] Through this step, a closed-loop process is achieved, starting from the generation of expert veto opinions, to the structured storage of data, and then to the automatic archiving of review reports. This not only improves the efficiency of preparing bid evaluation reports, but also significantly enhances the accuracy and consistency of the citation of veto clauses, helps reduce human risks such as missing filling, incorrect filling, and non-standard language, and meets the requirements of review standardization and compliance in large-scale power procurement scenarios.

[0065] Furthermore, for an intelligent auxiliary bid evaluation method for the power bidding profession disclosed in this embodiment, the matching method of the veto rule is generated based on the semantic matching similarity between the context vector and the rule label.

[0066] In a specific embodiment, to achieve the precise recommendation of veto opinions during the review process, the system establishes a semantic association between the context features and the veto rule labels based on the context semantic analysis technology. Its core mechanism is as follows: The system constructs a context vector of the review task and performs semantic matching with the semantic labels of each rule in the veto rule library, and filters out the veto recommendation content most relevant to the current review scenario according to the similarity level.

[0067] Specifically, when the review task is started, the system first analyzes the information included in the bidding task, such as the bidder name, the category of the product bid, the list of attachment files, whether the key terms are met, etc., and combines the identity of the review expert (such as technical expert or business expert) and the project procurement requirements to extract a series of representative context keywords and phrases. These keywords are encoded by the semantic analysis module to form a set of semantic vector collections, collectively referred to as "context vectors".

[0068] At the same time, for each rule in the veto rule library maintained in the system, when it is entered, it is accompanied by text description contents such as manually written labels, clause citations, and applicable scopes. After being analyzed by the language model, the corresponding semantic labels are generated. The semantic labels not only include keywords, but also descriptions of typical applicable scenarios, summary phrases of historical cases, etc., to ensure that they have strong context expression capabilities.

[0069] In the specific matching process, the system compares the above-mentioned "context vectors" with the "rule semantic labels" one by one. The comparison process is completed through the semantic knowledge network constructed by the language model, which can understand that there is an equivalent or near-synonymous relationship between "CNAS certification report not uploaded" and "missing third-party test file". The system scores the matching result between each rule and the current context, and the higher the score value, the higher the semantic fit degree between the rule and the current review task.

[0070] The system is set with a matching threshold line. For example, only the veto rules with a semantic score above 70% are retained, and they are sorted according to the matching scores. Finally, the system selects several highly relevant veto rules from the rule library to form a list of veto suggestions and pushes them to the reviewer's terminal.

[0071] For example, in the review task of a certain cable terminal equipment, the system identifies that the tender document lacks a CNAS certification report, the quotation is higher than the maximum price limit, and the product model is not marked. After combining these three types of context information, the system accurately matches three rules in the veto rule library: one is "no product type test report is provided", the second is "the quotation exceeding the price limit cannot pass", and the third is "the technical parameters are inconsistent and it is rejected". These three rules are screened into the suggestion list by the system due to their high matching degree and are displayed to the experts in the order of relevance.

[0072] The semantic matching mechanism of this embodiment abandons the traditional "keyword inclusion" fuzzy retrieval method and improves the accuracy, scenario adaptability and intelligence of veto rule recommendation through a combination of context understanding and tag parsing, ensuring the professionalism and compliance of review opinions and effectively avoiding the risk of wrong rejection or missed rejection by experts due to neglecting certain elements.

[0073] Furthermore, for the intelligent assisted bid evaluation method for the power bidding profession disclosed in this embodiment, the list of veto suggestions is expressed in a structured content, including but not limited to the reference clause section, the veto statement section, and the parameter filling section.

[0074] In a specific embodiment, to improve the understanding efficiency and quick operation ability of reviewers for veto suggestions, the system organizes and displays each veto suggestion in a structured content manner. Each suggestion consists of three basic component sections, namely the "reference clause section", the "veto statement section", and the "parameter filling section". The system encapsulates them uniformly to form a "structured list of veto suggestions" and pushes it to the reviewer's terminal interface for presentation.

[0075] First of all, the "reference clause section" is mainly used to clarify the basis of regulations or procurement documents for this veto suggestion. The system automatically extracts the corresponding regulation article number, the chapter title and the abstract of the clause content of the bidding document. The content of this section usually includes the legal source, scope of application and judgment criteria of the clause. For example, for a certain suggestion "no test report issued by a third party is provided", the content of its reference clause section may be "According to Article 4.3.1 of the 'Procurement Document', all cable terminals must provide a type test report issued by an institution with CNAS qualification." Secondly, the "veto statement segment" is the standardized veto description corresponding to this piece of advice. This segment of content is automatically generated by the system according to the standard veto statements in the rule library and can also be modified or supplemented by experts later. The veto statement uses an objective and clear expression method and has the characteristic of being directly quotable by the report. For example: "Upon investigation, the bidder did not provide a CNAS type test report in the response document, which does not meet the requirements of Article 4.3.1 of the procurement document. It is recommended to veto." Thirdly, the "parameter filling segment" is used to supplement specific context variables, such as report numbers, document page numbers, tender offer values, etc. This segment usually contains multiple preset parameter fields, which can be automatically filled by the system according to context data or allow experts to manually enter. For example, for veto items related to quotations, the parameter filling segment may include "ceiling price: 220,000 yuan", "tender offer: 243,600 yuan", "exceeding percentage: 10.7%".

[0076] In addition, other fields can also be extended in the structured list, such as "applicable clause type" (such as eligibility, compliance, scoring), "matching confidence score" (used to indicate the degree of matching of this piece of advice with the current context), "historical adoption frequency" (used to show the usage frequency of this piece of advice in historical tasks), etc.

[0077] When experts browse the structured veto advice list in the terminal interface, they can quickly make a decision on whether to adopt based on the structured paragraphs. The system interface supports one-key citation (inserting the entire structured content into the review opinion letter), field modification (such as manually revising clause citations or parameter values), and adding remarks (supplementary explanations for specific judgments). The entire operation process does not require manual page-by-page search for articles in the procurement document, repeated writing of veto texts, or copying of parameters, improving the review efficiency and text standardization.

[0078] Through the structured expression method of this embodiment, the system realizes the efficient conversion of veto advice from the original rules to actionable review opinions, significantly improves the review consistency, text accuracy, and user experience, and reduces the risk of disputes caused by vague expressions or citation errors of the bid evaluation experts.

[0079] Furthermore, for the intelligent auxiliary bid evaluation method for the power bidding profession disclosed in this embodiment, the review expert terminal interface supports functions such as one-key citation, text insertion, attachment upload, scoring confirmation, and veto explanation.

[0080] In a specific embodiment, to improve the operation efficiency and decision-making accuracy of review experts during the preliminary review process, the system provides a highly integrated review expert terminal interface. This interface has function modules such as one-key citation, text insertion, attachment upload, scoring confirmation, and veto explanation, which can significantly improve the generation speed and standardization of veto opinions.

[0081] First, the one-key citation function means that after the system pushes the list of veto suggestions to the reviewer's interface, the reviewer can quickly insert the selected structured veto suggestions as a whole into the text area of the current review opinion through the "Insert" button on the interface. This operation does not require manual copying and pasting. The system automatically completes the clause citation, splicing of veto statements and parameters, ensuring uniform format specifications.

[0082] Second, the text insertion function allows reviewers to make supplementary explanations for specific case situations based on structured suggestions. For example, reviewers can add at the end of the veto statement "After verification, the bidder does indeed lack this report on page 15 of the quotation document, and it is impossible to judge technical consistency". This function supports inserting custom paragraphs, term phrases, sentence templates, etc., and maintains the consistency of the language through automatic grammar checking.

[0083] Third, the attachment upload function mainly serves for evidence supplementation and retention of review bases in complex situations. Reviewers can upload materials such as key pages, screenshots, and review records related to the veto opinion to the system through the "Upload Attachment" button. The system will automatically generate a unique number for each attachment and bind it to the corresponding review opinion, supporting retrospective review during the later report review stage.

[0084] In addition, the scoring confirmation function is used to quickly check the "Whether Passed" and "Scoring Value" fields during the review process. After the veto suggestion is generated, reviewers can confirm the status of this bidding item (such as "Veto" or "Retain") through the scoring module and input or confirm the value in the scoring box provided by the system. This function ensures the consistency between the veto and scoring results, avoiding omissions or contradictory expressions.

[0085] Finally, the veto explanation function provides an optional field area for reviewers to make written explanations when dealing with difficult clauses or ambiguous cases. For example, reviewers can add "Although the bidder submitted a test report, it did not bear the CNAS logo, and according to the regulations, it should also be regarded as invalid". The system supports automatically classifying and tagging this explanation, and it can be reused for subsequent projects.

[0086] The entire reviewer terminal interface adopts a modular design, supporting mouse clicks and keyboard shortcut operations. The system automatically records the timestamp, modification history, and operator identity of each veto action, forming a complete audit trail. In addition, the interface can be linked with the batch task management module. When the grouping is locked or the task is submitted, the system will automatically prompt the reviewer to complete the confirmation to avoid omissions.

[0087] Through the above function integration, the system has significantly improved the operation efficiency and expression consistency in the bid evaluation process, enabling reviewers to maintain accurate judgment, standardized text, and well-founded operations in a high-intensity and multi-task review environment, thus enhancing the professional level and digital quality of the entire tender review process.

[0088] Furthermore, for the intelligent auxiliary bid evaluation method for the power bidding profession disclosed in this embodiment, the final rejection opinions are stored using structured fields.

[0089] In a specific embodiment, to achieve standardized management and efficient reuse of review data, the system stores the final rejection opinions confirmed by review experts in the form of structured fields. This structured storage mode not only supports the standardized archiving of opinions but also supports subsequent retrieval analysis, statistical classification, and automatic report generation.

[0090] Specifically, after the experts confirm and submit the final rejection opinions, the system parses them into multiple concrete fields and stores them in the "Rejection Report Detail Opinion Table". The structured fields mainly include, but are not limited to, the following: 1. Rule source field: Records the regulations, procurement document clause numbers, and brief descriptions on which the rejection opinion is based. For example: The regulation source is Article 4.3.1 of the "Bidding and Procurement Review Rules", "Technical parameter inconsistencies shall not pass".

[0091] 2. Rejection statement field: Stores the standardized rejection statements confirmed by experts, usually including judgment logic and conclusion expression. Such as: "The bidder's quotation is higher than the maximum price limit, which does not conform to the procurement document regulations, and it is recommended to reject." 3. Parameter field: Used to record the variable values involved in this rejection judgment, such as "Maximum price limit: 3.5 million yuan", "Bid quotation: 3.72 million yuan", "Exceeding limit range: 6.3%".

[0092] 4. Operator field: Records the identity information of the final confirmor of this rejection opinion, including expert ID, expert role (such as technical expert or business expert), etc.

[0093] 5. Attachment identification field: If this rejection opinion is accompanied by screenshots, supplementary materials, or third-party documents, their file paths and numbers are stored in the database together to ensure traceability and display in the later stage.

[0094] 6. Rejection reason description field: Supplementary description information filled in by experts when necessary. For example: "Although the bidder submitted an inspection report, but the detection parameters A and B are missing, so it is not accepted."

[0095] 7. Rejection status field: Used to mark the status of this opinion, such as "Initial", "Adopted", "Rejected", "Needs reconsideration", etc., to facilitate process control and status tracking.

[0096] 8. Timestamp field: Records the creation time, last modification time, and archiving time of this rejection opinion, supporting the whole process of audit with traceability.

[0097] 9. Associated task fields: Indicate the bid task ID, review approval level ID, grouping ID, etc. associated with this rejection opinion, ensuring that the system can automatically file this opinion into the specified review report document structure.

[0098] When the data is stored in the database, the system will automatically verify and uniformly format the field content to ensure that expressions under different expert writing styles can also be unified into a standard data format, facilitating archiving, analysis, and reuse. This structured field can support being reorganized into natural language paragraphs as needed during front-end display, and can also support the report system to directly generate a rejection list with numbers and cited articles.

[0099] Through the structured field storage method of this embodiment, the system has significantly improved the standardized expression level of rejection opinions during the bid evaluation process, solved problems such as semantic ambiguity, inconvenient retrieval, and difficulty in statistics caused by traditional handwritten or free-form input of review opinions, and provided basic data guarantee for realizing full-process electronic and intelligent bid evaluation.

[0100] Furthermore, the intelligent auxiliary bid evaluation method for the power bidding profession disclosed in this embodiment includes a consistency verification step, including but not limited to checking for missing scoring items, judging the reasonableness of rejection items, and verifying the validity of rule references.

[0101] In a specific embodiment, to ensure that various types of review data generated during the intelligent auxiliary bid evaluation process are legal, compliant, and have internal logical consistency, the system executes a set of predefined consistency verification mechanisms before the expert completes the review task and submits the rejection opinion, which is used to automatically detect possible data anomalies, rule conflicts, or semantic contradictions, ensuring that the final review result can be directly filed for the bid evaluation report.

[0102] This consistency verification step mainly includes the following three core sub-modules: 1. Checking for missing scoring items. The system verifies each scoring item required in the current review task item by item to determine whether there is a situation where the score is not filled or missing. Especially for procurement projects with mandatory scoring items (such as "quotation score", "technical solution score"), the system will compare the scoring item configuration table with the actual scoring record table, automatically mark the missing items and prompt the expert to supplement the record with highlights or red frames to prevent the total score from being abnormal or the review result from being invalid due to missed scoring.

[0103] 2. Rationality Judgment of Rejection Items. This module cross - verifies the rejection opinions submitted by experts with the requirements of the procurement documents through semantic comparison and context consistency analysis techniques. For example, if the rejection opinion is marked as "technical agreement not provided", but the system finds that the technical agreement document is already included when parsing the uploaded materials, or this item is a non - mandatory clause, the system will automatically trigger a prompt: "This rejection situation does not match the actual content of the materials or procurement requirements. Please review." In addition, this module also checks whether the reasons for rejection match the attributes of the bidders, the category of the package segment, and the project type, to prevent misusing clauses or incorrect scope of application.

[0104] 3. Validity Verification of Rule References. The system performs real - time legality checks on the rejection clause entries cited by experts, mainly including three types of verifications: First, whether the cited clause exists in the current procurement document or the regulatory database; Second, whether the clause number corresponds to the clause content, to avoid number deviations caused by copy - paste or oral citation; Third, whether the cited clause is still within the valid period, to prevent using old version content that has been repealed, changed, or conflicts with new documents. During the verification process, the system also associates with the version control library to compare whether the version of the procurement document used in the review approval is consistent with the version of the rule - cited article, to ensure version context consistency.

[0105] After all consistency checks pass, the system allows experts to submit the current review task and automatically generates a "review result status" mark for the current bidding unit, including "passed", "rejected", "to be corrected", etc. At the same time, the results of all consistency checks will be recorded in the log table for subsequent auditing, review dispute explanation, and full - process compliance assessment.

[0106] Through the consistency check steps of this embodiment, while maintaining the subjective judgment right of experts, the system realizes quality control and intelligent error correction of the review process through a dual - verification mechanism of structured data and logical rules, greatly improving the accuracy, fairness, and efficiency of the review work, and laying a solid foundation for realizing the standardization and intelligentization of the power bidding evaluation work.

[0107] Figure 2 The intelligent auxiliary bidding evaluation system for the power bidding profession provided by this embodiment of the present disclosure includes: A review task list generation module, which is used to obtain the uploaded procurement information and create a procurement batch based on the configured basic parameters and procurement review rules, so as to generate a review task list with a unique batch number.

[0108] A feature vector construction module, which is used to extract the context information of each review task based on the review task list and the configured basic parameters, so as to construct the context feature vector of the review task.

[0109] A veto rule matching module is used to match the context feature vector with a preset veto rule to obtain a list of veto suggestions corresponding to each review task in the review task list.

[0110] A binding and pushing module is used to bind the list of veto suggestions to the corresponding bidding tasks and push the bidding tasks with the bound list of negative suggestions to the user terminal.

[0111] A veto opinion generation module is used to receive the operations of the user on each veto suggestion through the user terminal to generate a final veto opinion, and store and archive the final veto opinion.

[0112] In a specific embodiment, to achieve the automatic initialization and batch management of various review tasks in the power bidding profession, the system provides an integrated review task list generation module. This module has functions such as information entry, structured parsing, review rule matching, and automatic batch creation. As the starting step of the entire intelligent auxiliary bid evaluation process, it provides an accurate and complete parameter basis for subsequent context extraction, veto opinion generation, etc.

[0113] First, in the project initiation stage, the system user (such as a bidding agent or a system administrator) uploads the section information table of the procurement project of this batch through the front-end interface. This table includes fields such as project number, section name, section quantity, package number allocation, project type (materials, engineering, services), etc. The system automatically parses the original data in Excel or CSV format, identifies and verifies the format integrity and semantic normativity of the content of each field. For example, it performs a uniqueness check on the "section number" field and a hierarchical matching and parsing on the "package number" field.

[0114] Subsequently, the user continues to upload the corresponding list of requirements for this batch. This table contains information such as technical requirements, qualification conditions, scoring weights, and quotation methods under each section. After receiving this table, the system automatically constructs a requirements-section mapping relationship table and extracts the review parameters (such as weighted allocation ratio, maximum price limit, minimum qualification requirements, etc.) as one of the input parameters for subsequent context construction and veto rule matching.

[0115] Next, the system provides an upload entry for "jury information", supporting the batch import of the list of review experts, including expert names, ID numbers, unit information, professional categories, role identifiers (team leader, deputy team leader, team member), etc. The system automatically generates expert account information and initial login passwords, and performs grouping and role assignment according to the import rules to ensure permission isolation and clear review paths in subsequent batch management.

[0116] After the above information is imported, the system starts the review rule matching engine. Based on parameters such as project category (e.g., engineering), review type (e.g., prequalification, preliminary review), and the rule version of the affiliated platform, it automatically calls the preset rule template library to generate the corresponding "batch review rule parameter set". This includes core rule items such as whether to allow the bid price to exceed the limit, whether the technical bid is a mandatory requirement, and whether the scoring method is a point system or a grading system.

[0117] Finally, the system constructs a complete review "batch object" based on the above information and writes it into the database "veto_bid (veto batch table)". This batch object is assigned unique batch ID, timestamp, creator information and other identifiers, serving as the full-process primary key index for the current bid evaluation task and being used for data reference in all subsequent review tasks, veto records, expert behaviors, report outputs, etc.

[0118] Through the review task list generation module of this embodiment, the import efficiency and accuracy of various basic data in a large number of bid evaluation projects can be significantly improved, realizing a one-key conversion from the original table to system parameters, providing a high-quality data basis for subsequent structured processing, context analysis and automated review decision-making, and comprehensively supporting the construction of an intelligent and batch modern bid evaluation system.

[0119] The veto rule matching module is used to match the context feature vector C with the veto rules to obtain a list of veto suggestions that match the current review task.

[0120] In a specific embodiment, to achieve the intelligent matching and push of veto clauses, the system designs a veto rule matching module. The core task of this module is: based on the review context feature vector C generated in the previous steps, automatically screen out the veto rules highly relevant to the current review task from the veto rule database and generate a list of veto suggestions for experts' reference.

[0121] The system first obtains the context feature vector C of the current review task. This vector consists of multiple dimensions, covering information such as project category, scoring criteria, bid response situation, historical rule trigger tags, review stage, etc. After standardization processing, each feature dimension is encoded as a dense vector to form a complete semantic vector representation.

[0122] Next, the system traverses the preset veto rule entries in the veto rule database. Each veto rule is bound with a set of structured semantic tags when entered, such as "qualification not stamped", "no power of attorney from the legal representative provided", "bid price exceeding the limit", etc., and is converted into a vector representation through a language model. The system uses a semantic matching algorithm to compare the context vector C with the label vector of each rule and calculate the similarity score.

[0123] When the semantic similarity of a certain veto rule to the current context is higher than the system - preset threshold (e.g., 0.85), this rule will be determined as "suitable for the current review task" and enter the candidate rule set. For candidate rules, the system further performs a legality check to ensure that the current version of this clause is legal and effective, applicable to the current project type and scoring items, and there is no content conflicting with the review document.

[0124] Finally, the system restructures the selected veto rules into a "veto recommendation list", and attaches context - prompt content and default parameter filling items. Each entry in this list consists of three parts: one is the clause basis, that is, the clause reference information from the procurement document or the regulation library; the second is the veto recommendation statement, that is, the standardized language expression that experts can directly quote; the third is the parameter filling section, which has editable fields preset for experts to further adjust, such as the quoted amount, qualification years, etc.

[0125] Through this module, the system can automatically complete the complex rule - matching process, which not only reduces the time cost for experts to retrieve clauses, but also ensures that the sources of veto opinions are clear and the expressions are standardized, avoiding veto mistakes caused by inconsistent human judgments, thus significantly improving the compliance and work efficiency of the bid - evaluation process.

[0126] The veto opinion generation module is used to accept the veto recommendation list, and adopt, expand or semantically modify the veto recommendations to form the final veto opinion.

[0127] In a specific embodiment, this module, as the core functional unit of human - machine interaction, undertakes the task of transforming "recommended veto clauses" into "final veto opinions", supporting experts to complete operations such as adopting, modifying, supplementing and confirming veto statements without leaving the system environment.

[0128] After the veto rule matching module generates a veto recommendation list that matches the current bidding task, the veto opinion generation module will automatically push this list to the expert terminal interface. The interface is presented in the form of task cards, and each veto recommendation includes information such as clause basis, veto statement template, system - pre - filled parameters, applicable bid sections, etc. The bid - evaluation experts can perform the following processing behaviors on the operation interface: 1. One - click adoption: If the expert confirms that the recommended clause does not match the bidder's response and the description is accurate, the expert can click the "Adopt" button, and the system will automatically write this veto statement and its cited clause into the final opinion list.

[0129] 2. Semantic modification: If the expert believes that the recommended expression does not match the project language style or additional explanatory content is needed, the expert can click the "Edit" button to directly perform semantic adjustment on the original template, keeping the clause basis field unchanged and only modifying the veto statement field or parameter description content.

[0130] 3. Content Expansion: For complex or multi-factor coexisting veto situations, experts can use the "Add New Paragraph" function to append custom content to the recommendation. The system supports experts in supplementing factual basis and clause evidence by citing historical cases, uploading attachment descriptions, inserting screenshots, or legal texts, etc.

[0131] 4. Ignore or Replace: If the system's recommendation does not apply to the current bid response situation, the expert can click the "Ignore" button to remove it from the list, or manually search for other clauses in the veto library and replace the current recommendation.

[0132] During the entire processing process, every operation performed by the expert is recorded by the system, including the operator, operation time, content version before and after modification, ensuring the traceability ability and review consistency requirements in the subsequent process.

[0133] Finally, the veto content confirmed by the expert will be automatically structured and stored to form a "Final Veto Opinion" object, which includes fields such as original clause reference, processing type (adoption / modification / expansion), expert confirmation content, optional additional explanations, attachments, etc. This object is then bound to the specific review task and is ready to enter the report archiving process of the next stage.

[0134] Through the veto opinion generation module of this embodiment, a closed-loop processing mechanism for veto clauses from recommendation to confirmation is realized. It not only retains the professionalism of manual judgment but also reduces the omission rate and the risk of non-standard expression of experts in high-intensity work with the help of system recommendations, significantly improving the efficiency, compliance, and intelligent level of power bidding review work.

[0135] It also includes a storage and archiving module for collecting and storing the final veto opinions and transmitting the veto opinions to the corresponding preliminary evaluation report documents.

[0136] In a specific embodiment, to achieve the automatic collection and document archiving of the output results in the review stage, the system designs a storage and archiving module. This module is used to receive and store the final veto opinions confirmed by the evaluation experts and ensure that all veto information is accurately classified into the corresponding preliminary evaluation report documents, realizing the structured management and automated arrangement of review conclusions.

[0137] When the veto opinion generation module outputs the confirmed final veto opinions, the system persists each veto opinion as a structured data object. This object contains the following key fields: reference basis for the veto clause (from the procurement document or regulatory source); text of the veto statement (original text of the expert or system recommendation statement); batch ID, report ID, evaluation expert ID, bidder ID; processing type (adoption / modification / expansion) and whether there is an attachment description; confirmation timestamp and version identifier (supporting historical traceability); category of the evaluation (such as qualification / formality / technicality, etc.) and section / package number information.

[0138] This module will automatically sort and classify the above structured information, write it into the corresponding entries in the "Rejection Report Details Table" and the "Rejection Report Details Opinion Table", and insert it into the specified location of the report document according to the chapter structure of the bid evaluation report. For example: rejections in the technical part are written into the technical chapter, and rejections in the qualification part are filed in the qualification chapter to ensure the standardization of the document structure and the integrity of the content.

[0139] In addition, the storage and archiving module also incorporates a document synchronization engine. When the system detects an update (such as addition, modification, or deletion) of the rejection opinions confirmed by the evaluation experts, it will automatically trigger an incremental update mechanism for the report document, only modifying the relevant fields or paragraphs, avoiding repeated recompilation of the entire report, and improving the document generation efficiency. To ensure the rigor and traceability of document archiving, the module also provides the following functions: operation record log: the entire process of archiving each rejection opinion is recorded, including the writing time, writer, and writing location; version comparison mechanism: supports viewing the historical versions and modification differences of the rejection opinions; manual review prompt: if the system determines that there are multiple similar rejection items for the same bidder, the system will mark and prompt the bidding agency or team leader for further verification. Through the storage and archiving module of this embodiment, the system realizes the closed-loop processing of rejection opinions from manual confirmation to document archiving, greatly reducing repetitive labor such as manual text sorting and copy-pasting. At the same time, it ensures the unity of the report structure, the accuracy of information, and the standardization of expression, improving the automation level and document quality management ability of the entire power bidding evaluation process.

[0140] Furthermore, in an intelligent auxiliary bid evaluation system for the power bidding profession disclosed in this embodiment, the matching method of the rejection rules is generated based on the semantic matching similarity between the context vector and the rule tags.

[0141] In a specific embodiment, the matching method of the rejection rules is intelligently judged based on the semantic matching similarity between the evaluation context vector and the rejection rule tags to achieve the automatic screening and recommendation of applicable rejection clauses in the current evaluation context.

[0142] Specifically, during the execution of the bid evaluation task, the system will first extract a set of context information based on factors such as the content of the bid documents, project categories, scoring criteria, and the status of response document fields, and convert this context content into a unified format of evaluation context feature vector C through natural language processing techniques (such as word vectors, pre-training model encoding, etc.). This vector is a high-dimensional semantic expression form that can comprehensively reflect the semantic environment and judgment requirements of the current evaluation task.

[0143] Meanwhile, all the veto clauses in the veto rule library are bound to corresponding structured semantic tags, which include not only the rule names and clause contents, but also descriptive information such as applicable scenarios, general terms, and key negative words. The system also encodes the contents of these tags into rule tag vectors through a semantic model.

[0144] The system pairwise compares the context feature vector C with the veto rule tag vectors through a semantic similarity calculation method (such as cosine similarity), and scores the comparison results. When the matching similarity of a certain rule is higher than a set threshold (such as 0.80), the system considers that this rule is highly relevant to the current review context and adds it to the candidate veto recommendation set.

[0145] In the present invention, the "context feature vector C" is designed as the core input for driving veto rule matching, and its content generally covers multi-dimensional information such as tender batch configuration, scoring rule selection, material response status, bidder attributes, expert role types, and scoring stage identifiers. After being integrated by preset logical rules or semantic extraction algorithms, these information form a feature vector representing the context of the review task. However, the context feature vector C has the following flaws: the construction mechanism of the vector C lacks explicit structural constraints and determination order definitions. The currently commonly used solution is "keyword aggregation + field tagging", that is, simply splicing the keywords, status representations, or field values captured in the context to form a vector set, but the triggering dependencies, logical sequence, or stage attribution among the fields are not clearly identified (such as qualification judgment should be earlier than technical veto). Such an unordered splicing vector representation form is likely to cause semantic mismatches or judgment conflicts when matching with veto rules. In addition, the veto rules themselves have explicit logical sequences and applicable preconditions, that is, different rules are attached to different review stages, role types, or field statuses. For example, rules such as "if the qualification review fails, the bidder should be vetoed" should be completed before the technical review, while vetoes such as "the technical scheme is inconsistent with the parameters" should be established after the qualification is passed. If the vector C does not indicate the stage boundaries or execution paths, it is very likely that the system will push subsequent veto clauses before a certain prerequisite step of the review is completed, resulting in logical disorder. In the AI-driven bid evaluation decision-making, once there is a conflict in the logical topological sorting between the vector C and the veto rules, the system will output veto recommendations that do not conform to the real process based on the wrong context judgment path, resulting in a chain of misjudgments of "mis-push - mis-adoption - mis-filing", which is particularly difficult to explain and hold accountable in scenarios of multi-expert collaboration and responsibility tracing.

[0146] It should be understood that the context feature vector C can actually be a set containing multiple feature vectors.

[0147] Since the veto clause itself has a logical relationship, in some embodiments, the first topological sort and the second topological sort of the veto clause can be constructed based on the extraction of keywords in the veto clause.

[0148] Among them, the first topological sort is the necessary condition of the veto clause: that is, when the conditions of all preceding nodes in the sequence are not met in the bidding document, the veto clause should not be invoked; correspondingly, the second topological sort is the sufficient condition of the veto clause. It can be understood that the directed graph formed by the first topological sort is actually a subgraph of the directed graph formed by the second topological sort. It can be understood that the topological sort regards keywords as nodes in the directed graph, and it should at least include: keywords and the logical dependencies of keywords; it can be understood that the logical dependencies are equivalent to the directed edges of the graph.

[0149] If the veto clause includes: "Those who do not provide the product factory certificate shall not be recognized as having valid supply qualifications and shall be regarded as not meeting the requirements of qualification review"; then a directed edge should be generated from the "product certificate" node to "supply qualification". If the veto clause includes: "If the type test report with third-party CNAS certification cannot be provided, the test data is not valid and shall be treated as an invalid response"; then a directed edge should be generated from the "third-party CNAS certification" node to "valid test data".

[0150] The first topological sort, as a mapping of the veto clause, facilitates determining whether to use the veto clause for the keywords in the bidding document; when the first topological sort is satisfied, the second topological sort is used to compare with the third topological sort formed by the bidding document itself. If in the third topological sort, if the weight of a certain directed edge is positive and negative inconsistently in the second topological sort and the third topological sort, the system shall veto; otherwise, if the traffic of a certain node does not meet the traffic interval of the nodes in the second topological sort, it shall be handed over to an expert for judgment.

[0151] In the initial situation, the second topological sort can be the first topological sort. However, as the number of tender documents increases, the expert judgment can also be based on the rich content of the tender documents to expand the second topological sort twice. For example, assume that the second topological sort includes: "If a type test report with third-party CNAS certification cannot be provided, its test data is not valid and shall be treated as an invalid response"; and the third topological sort of a certain tender document is: "A factory self-inspection type report has been provided, accompanied by enterprise quality inspection". After semantic approximation processing, it is not difficult to infer that the test data is similar to the quality inspection, and the third topological sort attempts to replace the type test report with CNAS certification with a factory self-inspection type report. At this time, a node for the factory self-inspection type report should be generated, and a directed edge with a negative weight pointing from the node of the factory self-inspection type report to the node of the test data should be generated. In addition, two attributes, flow, can be set for the node itself. In the second topological sort or the third topological sort, the flow is equal to the sum of the weight values of the directed edges starting from this node. It can be understood that the flow is used to describe the situation of classification discussion. For example, the tender document includes: "For the validity verification of the insulation level of the transformer winding, a type test report certified by the state or an A-level equipment inspection report shall be provided, and either one being satisfied can be regarded as qualified". And assume that the weights of the directed edges from the node of the transformer winding insulation level to the node of the type test report certified by the state and from the node of the transformer winding insulation level to the node of the A-level equipment inspection report are 4 and 7 respectively. Then the corresponding flow interval of this node can be set as [4, M], where M is equivalent to positive infinity.

[0152] It can be understood that in the second topological sort, the setting of the weight value is actually random, but the positive and negative attributes cannot be random. The setting of the value only needs to ensure that the logical relationship between the weight value and the flow is not contradictory. Therefore, in the second topological sort, the flow of the node is actually a flow interval; while the third topological sort is the specific flow value calculated based on the actual weight value of the directed edge. In addition, if the weight value of the directed edge in the third topological sort inherits the weight value of the corresponding directed edge in the second topological sort. In the first topological sort, the directed edge does not set a weight; while in the second topological sort, the directed edge needs to set a weight interval, and the weight value in the weight interval can be 0 or a negative number.

[0153] To further improve the business accuracy of the matching result, on the basis of similarity screening, the system will also combine conditions such as the current review stage, the type of the section to which it belongs, and the project category to filter the candidate results twice, only retaining the veto rules with effective applicable scopes under the current task, and finally generating a personalized list of veto suggestions.

[0154] In this embodiment, by introducing a context-aware semantic matching mechanism, the one-sidedness and manual configuration burden of traditional keyword-based retrieval are avoided, making the rejection suggestions recommended by the system more in line with the actual review situation, improving the accuracy and practicality of the recommendations, and providing more scientific and efficient decision-making support for the judgment of review experts.

[0155] Furthermore, an intelligent auxiliary bid evaluation system for the power bidding profession disclosed in this embodiment has the rejection suggestion list expressed in structured content, including but not limited to a reference clause section, a rejection statement section, and a parameter filling section.

[0156] In a specific embodiment, to ensure that the rejection suggestions generated during the review process have high readability, editability, and standardization, the system designs the structured expression of the content of the rejection suggestion list. In a modular manner, each proposed rejection opinion is split into multiple logical fields to form a unified data object format, which is convenient for bid evaluation experts to quickly understand, directly operate, and achieve automatic archiving.

[0157] Specifically, each rejection suggestion list contains the following three core structural sections: 1. Reference clause section. This section is used to clearly indicate the clause source on which this rejection suggestion is based, which can come from the bidding documents (such as the procurement requirement list, scoring rules, and business terms), applicable laws, regulations, or industry standards. The system configures a clear "reference source" field for each rule in the rejection rule library, and the content includes clause numbers, clause originals, source documents, etc. For example: "According to Article 3.2 of the 'Bidding Documents': 'The quotation document should be stamped with the legal person's seal and signed'."

[0158] 2. Rejection statement section. This section is a standard rejection expression sentence pattern for experts to quickly call, usually expressing the specific reasons why the bidder does not meet the terms in a clear, affirmative, compliant, and standardized tone. The system supports parameter placeholder for the template statement. For example: "The bidder did not stamp the legal representative's signature seal on the quotation document. According to the [reference clause section], it is rejected." This statement can be directly cited by experts or semantically modified.

[0159] 3. Parameter filling section. This section is a specific parameter input area associated with the review situation, supporting experts to fill in information such as the bidder's name, response number, missing material name, amount field, and time field in a fixed statement structure. The system supports some fields to be automatically captured by the program (such as the unit name and section number), and also allows experts to manually modify or supplement unrecognized data. All parameter fields are stored in a structured form, which is convenient for subsequent report compilation and data analysis.

[0160] In addition, while generating structured rejection suggestions, the system also supports advanced functions: automatic numbering and clause classification to ensure clear logic and easy reading when presenting the list; multi-language support to adapt to bid evaluation environments in different languages; and field linkage checks. For example, if the "quoted clause section" is empty, the "rejection statement section" will not be submitted.

[0161] Through the above-mentioned structured expression mechanism, this embodiment effectively avoids common problems in the expression of rejection opinions, such as subjective language confusion, content omission, or inconsistent formatting, improves the efficiency of expert review, the readability of the system, and the standardization degree of the report, and provides a standardized corpus basis for subsequent data backtracking and large model training.

[0162] Furthermore, an intelligent auxiliary bid evaluation system for the power bidding profession disclosed in this embodiment further includes a consistency verification module for checking the absence of scoring items, judging the reasonableness of rejection items, and verifying the validity of rule references.

[0163] In a specific embodiment, to ensure the integrity, compliance, and consistency of the rejection judgment and scoring operations throughout the bid evaluation process, the system integrates a consistency verification module. This module is used to automatically review the integrity and logical reasonableness of the content before the bid evaluation experts submit rejection opinions and scoring conclusions, and serves as one of the key quality assurance mechanisms for the system to generate the preliminary evaluation report.

[0164] This module mainly includes the following three sub-functions: 1. Checking for missing scoring items. The system first traverses each scoring item that the bidder should be scored on, and based on information such as "required scoring items", "weight settings", and "jury scoring permissions" defined in the procurement review rule library, determines whether there are items that have not been scored or have abnormal scores (such as exceeding the reasonable score range).

[0165] If a scoring item is required and not scored, the system will mark it as "missing" and prompt the jury to supplement it.

[0166] If the score value exceeds the reasonable upper and lower limit intervals (for example, the full score is 10 but 12 is filled in), the system will issue a format warning.

[0167] If the scoring data structure does not match (such as missing a description of the scoring basis or an attachment), the system will force a pop-up window to prompt for supplementary instructions.

[0168] This step ensures the integrity and compliance of the review process at the data structure and scoring content levels.

[0169] 2. Judging the reasonableness of rejection items. For each rejection opinion submitted by the jury, the system calls the historical case library, the clause context parsing model, and the review template rules for semantic analysis to judge whether the rejection reason has logical consistency, content adequacy, and clause adaptability.

[0170] If it is found that the clause cited in the rejection opinion does not belong to the current scoring stage (such as citing an eligibility clause in technical scoring), an adaptability error prompt will be issued; If there is a vague description in the rejection statement (such as "insufficient information" without specifying what kind of information), the system will mark it as "semantically incomplete" and suggest completion; If it is found that this rejection reason has been ruled as a "non-rejection matter" by others in this project, the system will prompt "inconsistent with the group results" for the expert to verify again.

[0171] This step is mainly used to control the rationality and consistency of the rejection behavior, preventing wrong rejections, missed rejections or excessive rejections.

[0172] 3. Verification of the validity of rule citation. The system compares the clause number or regulatory document cited in each rejection opinion with the actual version existing in the current procurement document to verify the accuracy and timeliness of its citation source.

[0173] If the cited clause number does not exist, the system will prompt "invalid cited clause"; If the cited legal provision has been repealed or is not applicable to this batch of procurement types, the system will automatically call the regulatory version comparison mechanism to give update suggestions; If the clause number is correct but its content is significantly inconsistent with the rejection reason (determined through semantic conflict detection), the system will prompt "citation does not match content".

[0174] This mechanism guarantees the traceability and validity of clause citation, avoiding the invalidation of the review due to citing incorrect clauses.

[0175] In summary, through the construction of a perfect consistency check module in this embodiment, various potential problems in the scoring, rejection and citation processes can be systematically detected and prompted before the review submission, which not only guarantees the review quality, but also significantly reduces the manual revision burden of the bidding agency in the report review stage, and further improves the intelligent level and decision-making credibility of the power bidding professional evaluation system.

[0176] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, which may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The processor can call the logical instructions in the memory to execute the method for implementing the soft authorization of the configuration software.

[0177] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present disclosure. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0178] On the other hand, the present disclosure also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to implement the implementation method based on the soft license of the configuration software provided by the above-mentioned various methods.

[0179] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0180] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disks, optical discs, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments.

[0181] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than limiting them; although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. An intelligent auxiliary bid evaluation method for the power bidding profession, characterized in that, The method includes: Obtaining the uploaded procurement information, and creating a procurement batch based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number; Extracting the context information of each review task based on the review task list and the configured basic parameters to construct a context feature vector of the review task; Matching the context feature vector with a preset veto rule to obtain a list of veto suggestions corresponding to each review task in the review task list; Binding the list of veto suggestions to the corresponding bidding tasks, and pushing the bidding tasks bound with the list of negative suggestions to the user terminal; Receiving, through the user terminal, the operations of the user on each veto suggestion to generate a final veto opinion, and storing and archiving the final veto opinion.

2. The intelligent auxiliary bid evaluation method for the power bidding specialty according to claim 1, wherein The obtaining the uploaded procurement information, and creating a procurement batch based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number includes: Obtaining a procurement requirement form, a judge information form, and procurement documents, where the procurement information includes the procurement requirement form, the judge information form, and the procurement documents; among them, the procurement requirement form contains multiple section information, the judge information form contains personal information of multiple expert judges, and the procurement documents contain multiple negative clauses; Configuring the basic parameters based on the procurement information; the basic parameters include procurement category, bid evaluation model version, veto format, and judge authority.

3. The intelligent auxiliary bid evaluation method for the power bidding specialty according to claim 2, characterized in that The obtaining the uploaded procurement information, and creating a procurement batch based on the configured basic parameters and procurement review rules to generate a review task list with a unique batch number further includes: Parsing the procurement documents to obtain the procurement review rules, and creating the procurement batch based on the procurement review rules; Determining the allocated batch numbers for each batch based on the procurement batch, and generating a subcontracting structure list, review task items, and an initial veto clause library based on the allocated batch numbers; Writing the subcontracting structure list, review task items, and the initial veto clause library into a context initialization parameter table to generate the review task list.

4. The intelligent auxiliary bid evaluation method for the power bidding specialty according to claim 3, wherein, The extracting the context information of each review task based on the review task list and the configured basic parameters to construct a context feature vector of the review task includes: Extracting structured information from the configured basic parameters; the structured information at least includes batch number, package number in the subcontracting structure list, equipment type, qualification clause number, maximum bid price limit, qualification type of the bidding unit, identity information of the review expert, and bidding unit number; Performing semantic analysis on the unstructured text paragraphs in the bidding documents through natural language parsing and keyword extraction techniques to identify the technical test requirements corresponding to the equipment type and the guiding language from the subjective judgment of the review expert from the bidding documents; Extracting keyword tags from the technical test requirements and the guiding language, and performing standardization processing on the keyword tags to obtain semantic signal fields; Integrating the semantic signal fields, keyword tags, and structured information, and encoding to generate the context feature vector of the review task.

5. The intelligent auxiliary bid evaluation method for the power bidding specialty according to claim 4, characterized in that, Matching the context feature vector with a preset veto rule to obtain a veto recommendation list corresponding to each review task in the review task list, including: Extracting a structured set of veto rules from a veto rule database; each veto rule in the set of veto rules includes a review criterion category, a rule label, a reference clause or document basis, a veto scenario type, and a veto template; Invoking a preset semantic similarity matching algorithm to compare the keyword fields labeled with keyword tags in the context feature vector with the veto rules to obtain veto rule entries with a similarity exceeding a first threshold; Based on the identity information of the review expert, invoking the semantic similarity matching algorithm to filter out veto rule entries with a technical standard similarity exceeding a second threshold corresponding to the identity information of the review expert; Sorting the filtered veto rule entries according to the magnitude of the similarity to generate the veto recommendation list corresponding to the scenarios of each review task; Wherein, the veto recommendation list includes a veto clause reference, a veto reason, a parameter field, and a text recommendation paragraph.

6. The intelligent auxiliary bid evaluation method for the power bidding specialty according to claim 5, characterized in that Binding the veto recommendation list to the corresponding bidding task and pushing the bidding task bound with the negative recommendation list to the user terminal, including: Obtaining the bidding task to be bound and determining a veto recommendation list with a matching degree exceeding a third threshold with the review task based on the bidding task; Binding the veto recommendation list to the bidding task through a task binding engine to obtain a binding result and writing the binding result into the task table in JSON format; Pushing the veto recommendation list to other user terminals of the technical review experts corresponding to the review task through the user terminal.

7. The intelligent auxiliary bid evaluation method for the power bidding specialty according to claim 6, wherein, Receiving, through the user terminal, operations of the user on each veto recommendation to generate a final veto opinion, and storing and archiving the final veto opinion, including: Obtaining, through the user terminal, the interaction information between the review expert and the operation panel; the interaction information includes adopting a veto recommendation, modifying a veto recommendation, and ignoring a veto recommendation; Obtaining a final status record of each veto recommendation based on the interaction information and converting the final status record into a structured field to generate the final veto opinion; the structured field includes the original allocation number of the veto recommendation, interaction information, interaction content, reference basis, parameter filling value, timestamp, and the identity information of the review expert.

8. An intelligent auxiliary bid evaluation system for the power bidding profession, characterized in that, A system for implementing the intelligent assisted bid evaluation method for the power bidding profession according to any one of claims 1 to 7, the system includes: A review task list generation module, configured to obtain uploaded procurement information and create a procurement batch based on configured basic parameters and procurement review rules to generate a review task list with a unique batch number; A feature vector construction module, configured to extract context information of each review task based on the review task list and configured basic parameters to construct a context feature vector of the review task; A veto rule matching module, configured to match the context feature vector with a preset veto rule to obtain a veto recommendation list corresponding to each review task in the review task list; The binding and pushing module is used to bind the list of veto suggestions to the corresponding bidding tasks and push the bidding tasks with the bound negative suggestion list to the user terminal; The veto opinion generation module is used to receive the operations of the user on each veto suggestion through the user terminal to generate the final veto opinion, and store and archive the final veto opinion.

9. An electronic device, characterized in that, It includes: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Computer-executable instructions are stored in the computer-readable storage medium, and when the computer-executable instructions are executed by the processor, they are used to implement the steps of the method according to any one of claims 1-7.

Images