A Bidder Qualification Intelligent Review Method and System Based on RPA
By automating the qualification review of bidders through RPA technology, the problems of low efficiency and poor accuracy of manual operation are solved. It enables batch querying across multiple sites and consistent result generation, adapting to the flexible needs of different business scenarios.
Patent Information
- Application Number
- CN202611295994.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-17
AI Technical Summary
The existing method of reviewing bidder qualifications relies on manual operation, which is inefficient, prone to errors, and fails to meet timeliness requirements. Furthermore, it lacks an automated mechanism for batch querying and result processing across multiple sites, leading to inconsistencies in information and a waste of human resources.
An RPA-based intelligent review method is adopted, which obtains data through standardized Excel templates, automatically queries multiple verification sites, generates a unified report and pushes the results, and introduces an anomaly handling mechanism to achieve batch review of multiple enterprises.
It significantly improves review efficiency and accuracy, reduces human intervention, ensures consistency and traceability of results, adapts to different business needs, and has flexibility and stability.
Smart Images

Figure CN123118019A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of process automation technology, specifically to an intelligent bidder qualification review method and system based on RPA. Background Technology
[0002] In corporate bidding processes, bidder qualification review is a crucial step in ensuring the compliance and fairness of bidding activities. It typically involves verifying bidders' corporate credit, qualification certificates, records of dishonesty, and information on affiliated companies across multiple dimensions. Data sources include various official and commercial platforms such as Credit China, Qichacha, the National Enterprise Credit Information Publicity System, and the National Construction Market Supervision Public Service Platform. Current review methods primarily rely on manual processes. Reviewers must manually compile information on the companies to be queried, log into different query platforms one by one, enter the company name to search, and then screenshot and organize the search results before finally compiling a review report. This process involves multiple platforms and multiple steps, making the overall workflow complex and highly dependent on human experience and operational proficiency.
[0003] However, existing technologies have significant shortcomings: First, manual review is inefficient, requiring individual companies to repeatedly query multiple sites, leading to a linear increase in workload when dealing with multiple bidders, making it difficult to meet timeliness requirements; second, manual operation is prone to errors, such as missing sites, incorrect information input, incomplete screenshots, and result matching discrepancies, affecting the accuracy of the review; third, the overall work is highly repetitive, with personnel engaging in mechanical operations for extended periods, resulting in a waste of human resources; furthermore, review results are mostly manually compiled, leading to inconsistent report formats and disorganized information archiving, hindering subsequent review and traceability. Although some automated tools can perform single-site queries, they cannot automate the entire process from data reception, multi-site batch queries, result compilation to report generation and result delivery, and lack anomaly handling mechanisms and multi-company related query strategies tailored to the characteristics of bidder qualification review, making it difficult to meet actual business needs.
[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0005] The purpose of this invention is to provide a bidder qualification intelligent review method and system based on RPA to solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a bidder qualification intelligent review method based on RPA, comprising the following steps: Obtain a standardized Excel template file sent via email. The standardized Excel template file contains basic information of the enterprise to be queried, query site configuration data, and auxiliary verification data. Monitor the specified trigger email address and initiate the intelligent review process for bidder qualifications upon receiving an email that meets the agreed requirements. The system performs format and data structure validation on the obtained standardized Excel template file. After the validation passes, it reads the enterprise information to be queried, the query site configuration data, and the auxiliary verification data, and completes the local storage of the corresponding data. Based on the read enterprise information and query site configuration data, automated query operations are performed on each enterprise on multiple preset verification sites. This includes accessing the target site through a browser, entering the enterprise name or key information, recognizing and verifying the CAPTCHA, completing account login, executing the query and matching query results that are exactly the same as the enterprise name, and at the same time, standardizing the query results and generating corresponding result records and screenshot files. The obtained query results and screenshots are summarized and processed to generate a standardized query results table. The corresponding screenshots are inserted according to the preset Word template to form a bidder qualification review report. The review report and query results table are packaged and named according to preset rules before being stored. After the packaged storage is completed, the target group chat is automatically retrieved and the packaged file is sent via instant messaging software, thereby automatically pushing the bidder qualification review results.
[0007] To build an automatic startup logic around task identification and trigger control during the email receiving process, the steps are as follows: Receive email messages containing standardized Excel attachments, extract the email subject field, attachment file name, sending timestamp, and sender address information, and perform structural parsing on the email content; The email subject field is matched against the preset trigger rules one by one, and the format of the attachment file name is validated according to the rules. At the same time, the number of attachments is checked to see if it meets the preset conditions. Parse the sender's address and generate a task identifier code, bind the task identifier code to the email metadata, and write it to the task queue cache structure; Write the task identifier code into the trigger log recording area, call the process scheduling interface to enter the subsequent processing link, and maintain the task status marking information at the same time.
[0008] To address the anomaly detection requirements during the data input phase, a structure verification and termination process is introduced, with the following steps: Load email attachments and parse the file header information to identify the file extension, file encoding format, and file integrity status; The number of worksheets, worksheet naming rules, and field column order in the Excel file are compared item by item, and the field types are checked to ensure they meet the preset requirements. Perform a row-by-row scan on key fields, including null value detection, character length detection, and data format validation, and mark the index of abnormal fields. The exception information is written to the exception log table, and the process interruption control signal is triggered at the same time. A copy of the exception file and the corresponding log data are saved.
[0009] Based on the variable requirements of query tasks, the site configuration parsing method is extended, and the steps are as follows: Read the site data area of the Excel template and extract the site identifier field and configuration value information for each row; Perform a tagging and identification operation on each configuration value, filter the records marked as "yes" as valid sites, and record their corresponding index numbers; The filtered results are converted into an ordered list structure, and a mapping table between site identifiers and access addresses is established. The query task sequence is generated based on the mapping relationship and written into the task execution queue structure in sequence for subsequent scheduling.
[0010] To address the issue of operational consistency during multi-platform queries, a unified execution logic is constructed, with the following steps: The browser driver component is invoked to initialize the runtime environment and load the target site URL, while waiting for the page elements to finish loading. Enter the company name or unified social credit code in the input control on the page, and perform input content validation and format correction operations; The system identifies the image area of the CAPTCHA on the page and calls a third-party interface to perform image recognition processing, while filling the recognition result into the CAPTCHA input box. Trigger the query button on the page and listen to the page response status, obtain the returned data content, locate the result list area and perform data extraction.
[0011] To ensure consistency in query result processing, a standardized output method is constructed, following these steps: Perform a line-by-line fetching operation on the result list returned by the query, and parse the company name field and additional information field in each record; Perform exact string matching on the parsed results, filter out records with exactly the same company name and mark them as target data; For cases where no match is found, a preset identifier text is written, and the current page URL is extracted and written into the result data structure. Perform a full-screen screenshot of the page and generate an image file, while recording the file storage path and associating it with the corresponding result record.
[0012] A record consistency verification mechanism is introduced during the query result processing. The target data fields are associated with and verified with the screenshot file path. Consistency comparison is performed on the matching results of enterprise names. At the same time, the correspondence between the page address in the result record and the screenshot source is verified. The data that passes the verification is written into a unified result structure, thereby completing the corresponding storage of result data and image information.
[0013] In line with the requirement for a unified structure in the review result output process, a template-based processing workflow is introduced, with the following steps: The query results from each site are aggregated and processed, integrating text information fields, URL information fields, and screenshot path information; The summary results are written into the corresponding cell range in the Excel result table according to the field mapping rules, and data formatting is uniformly processed at the same time. Load the Word template file and parse the preset placeholders in the template, then insert the screenshot file into the specified position according to the corresponding relationship; The generated file is compressed, and a file name is generated and written to the storage path by combining the time information, task identifier and sender information.
[0014] To address the need for automated execution of the review results distribution process, an information push workflow is constructed, with the following steps: Launch the instant messaging software client and load the user interface, while simultaneously entering the group chat search input area; Enter the target group chat name and perform the matching operation to locate the corresponding group chat window and open it. Read the local compressed file path and perform attachment loading operation, while verifying the existence and integrity of the file; Trigger the send operation and complete the file upload process, while recording the send time and send status information and writing it to the log system.
[0015] An RPA-based intelligent bidder qualification review system includes a triggering module, a data processing module, a multi-site query module, a report generation module, a result push module, an exception handling module, and a configuration management module. The trigger module monitors the specified trigger email address in real time, identifies emails that meet the agreed requirements and retrieves the standardized Excel attachments within them, and initiates the bidder qualification review process after detecting emails that meet the conditions. The data processing module performs format and data structure validation on the standardized Excel attachments obtained by the triggering module. After the validation passes, it reads the data to be queried, configuration data, and auxiliary verification data, parses them, and stores them locally. The multi-site query module, based on the query data parsed by the data processing module, calls the built-in automated operation scripts of multiple verification sites, performs batch query operations on each enterprise through browser driver, CAPTCHA recognition interface and platform login component, and completes query result matching, screenshot saving and text result recording; The report generation module summarizes and processes the query results obtained from the multi-site query module, fills the text information and screenshot information into the preset Word review report template and Excel result table template, generates the bidder qualification review report and corresponding result table, and completes file packaging and standardized naming; The results push module automatically sends the packaged file generated by the report generation module by searching the specified group chat through instant messaging software and pushing the compressed file. The exception handling module classifies and identifies system and business exceptions generated during each processing step, handles them accordingly, and records operation logs and saves exception screenshots. The configuration management module provides a configuration table to store account information, trigger email information, group information, working directory and site access address parameters, and supports parameter maintenance and updates.
[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention constructs a batch automated query mechanism for multiple sites and multiple enterprises, transforming the cumbersome process of manually logging into various platforms, manually entering information, and verifying each item into an automated execution process. Under a unified data input and triggering mechanism, it enables parallel or sequential automatic queries of multiple bidding companies at multiple verification sites, significantly reducing review time. Work that originally required hours or even days can be completed in tens of minutes, significantly improving the overall efficiency of bidder qualification review and effectively meeting the strict timeliness requirements of bidding and tendering business.
[0017] This invention effectively avoids problems such as input errors, missing queries, result matching deviations, and incomplete screenshots in traditional manual operations by using standardized data verification mechanisms, precise enterprise name matching rules, and automated execution processes. During the query process, a structured data processing method and an automated result collection method are uniformly adopted to ensure the consistency and integrity of various data sources, thereby comprehensively improving the accuracy and reliability of the review results and ensuring the authenticity and objectivity of the bidder qualification review conclusions.
[0018] This invention automates the repetitive and mechanized tasks of retrieval, screenshotting, and organization, eliminating the need for reviewers to participate in the specific query process. This significantly reduces human intervention, lowers reliance on professionals, and reduces the company's human resource costs. It also frees up relevant personnel from low-value repetitive labor, allowing them to focus on high-value tasks such as bidding and tendering decision analysis, thereby further improving the overall operational efficiency of the company.
[0019] This invention constructs standardized input and output templates to achieve unified and standardized input of data to be queried and unified and standardized output of review results. During the report generation process, it automatically completes data filling, screenshot embedding, and formatting, ensuring that the generated bidder qualification review reports maintain consistency and standardization in terms of structure, content, and format. This avoids information confusion and inconsistent formatting issues that occur during manual processing, thereby improving the readability, traceability, and subsequent management efficiency of the review results.
[0020] This invention establishes a comprehensive exception handling mechanism to classify, identify, and address various anomalies that may occur during process execution, including attachment format anomalies, site access anomalies, login anomalies, and query result anomalies. When an anomaly occurs, it can perform operations such as retrying, logging, and saving screenshots of the anomaly, effectively preventing single points of failure from causing overall process interruption, significantly improving the stability and fault tolerance of the review process, and ensuring the continuous and reliable operation of the system in complex environments.
[0021] This invention uses a configuration management mechanism to uniformly maintain various parameters required for system operation, enabling flexible configuration of query sites, account information, and operating parameters. It supports the selection of different verification site combinations according to different business scenarios, adapting to the bidder qualification review needs of different industries and enterprises. System adaptation can be completed without changing the overall process structure, possessing strong flexibility and scalability, which is conducive to the promotion and application of the system in multiple scenarios.
[0022] This invention utilizes an automated result push mechanism to send the results directly to a designated group chat via instant messaging tools after the review report is generated and packaged. Report distribution can be completed without manual intervention, achieving closed-loop management of the entire process from data acquisition and review execution to result output and sharing. This improves information transmission efficiency and team collaboration efficiency, making the bidding process more efficient and smooth. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0024] Figure 1 This is a flowchart of a method for intelligent review of bidder qualifications based on RPA according to the present invention.
[0025] Figure 2 This is a schematic diagram of a module of an intelligent bidder qualification review system based on RPA according to the present invention. Detailed Implementation
[0026] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.
[0027] This invention provides, for example Figure 1 The following are the specific steps of an RPA-based intelligent bidder qualification review method: During the data preparation and process triggering phase, users fill in the basic information of the company to be queried, configure the sites to be queried (marking the target sites as yes), and supplement auxiliary verification data such as the list of controlled suppliers whose cooperation is temporarily suspended and the list of Category I / II related companies, according to a pre-set standardized template for intelligent review of bidder qualifications based on RPA. The template file is then sent as an attachment to the designated triggering email address. In this process, the standardized template serves as a unified data input carrier, standardizing the format of the information to be queried, the field content, and the query configuration method. This ensures that different users follow consistent structural requirements when entering data, thereby avoiding anomalies in subsequent processing due to inconsistent data formats. Meanwhile, by setting a query site tag field in the template, users can flexibly select the target sites to be verified according to specific review needs, thus enabling configurable query scope. Supplementing the template with auxiliary verification data such as the list of suppliers whose cooperation with the controlling shareholder is temporarily suspended and the list of Category I / II related enterprises expands the review scope beyond a single enterprise to include related relationships and historical cooperation restrictions, enhancing the comprehensiveness and rigor of the review from a data perspective. Sending the completed template file as an email attachment to the designated trigger email address creates a unified task submission portal, simplifying the user operation process and providing standardized trigger conditions for review requests, which is conducive to the stable execution of subsequent automated processes.
[0028] After data preparation, an RPA-based intelligent review robot for bidder qualifications is configured with email triggers to monitor a designated trigger email address in real time. Upon receiving an email that meets the agreed-upon requirements, the review process is automatically initiated. This email triggering mechanism continuously monitors the email address to automatically detect external business requests. When an email containing standardized attachments (like an Excel file) that meets agreed-upon rules is detected, it is considered a valid review task input, triggering the execution of subsequent processes. In this process, the email serves not only as a data transmission carrier but also as a process initiation signal, giving the entire review process event-driven characteristics and allowing it to run automatically without manual intervention. Preliminary identification of email content and attachments ensures that the data entering the process meets preset requirements, reducing the impact of abnormal data on subsequent processing. This triggering method transforms the traditional model of manually initiating processes into an automatic response mechanism, creating a tight connection between data input and process execution. This improves overall processing efficiency and ensures the timeliness and accuracy of process initiation, providing a stable and reliable starting point for subsequent data verification, batch queries, and report generation.
[0029] During the data verification and retrieval phase, the RPA robot retrieves the latest email from the triggering mailbox, verifying the format (xlsx only) and data structure of the email attachments to ensure they conform to the preset template requirements. If they do not conform, a business exception is triggered, and the process ends directly. In this process, the mailbox serves as the unified data receiving entry point. The robot first filters and identifies the latest emails that meet the triggering conditions, ensuring that the processed data is valid for the current review task. For email attachments, file format verification is prioritized, allowing only standardized Excel files with the .xlsx extension to enter the subsequent processing flow, thus preventing data parsing failures due to file format incompatibility from the outset. After format verification, the internal data structure of the Excel file is further verified, including whether the worksheet names conform to preset requirements, whether each field column is complete, whether the field order is consistent, and whether there are any missing or abnormal key fields. This structural consistency verification ensures the data is parsable and standardized. When an attachment format does not meet the requirements or the data structure deviates from the preset template, the business exception handling logic is directly triggered, suspending the current process execution, thereby preventing erroneous data from entering subsequent query stages and causing a chain reaction. This verification mechanism completes the screening of data at the initial stage of data entering the system, effectively improving the stability of the overall process and the reliability of data processing.
[0030] After verifying the attachments and confirming that the data meets the preset requirements, the data reading and parsing phase begins. The robot reads the enterprise information to be queried from Sheet1 of the Excel template, the query configuration information of Sheet sites, the list of suppliers whose cooperation with Sheet Holdings is temporarily suspended, and the auxiliary verification data of the list of related enterprises of Sheet I / II, completing the local storage of the data. In the specific execution process, the enterprise information to be queried in Sheet1 is first read line by line to extract basic field information such as enterprise name and unified social credit code, and organized according to the established data structure so as to perform multi-site query operations one by one in the future. Then, the query configuration information in Sheet sites is read. By identifying the corresponding marker fields of each verification site, the set of target sites that need to participate in this review is determined, thereby realizing dynamic control of the query scope. Next, the auxiliary verification data in the list of suppliers whose cooperation with Sheet Holdings is temporarily suspended and the list of related enterprises of Sheet I / II are read. This part of the data is used to supplement the verification of enterprises or identify the relationship in the subsequent query process, further enhancing the completeness of the review dimensions. After reading all the aforementioned data types, the parsed structured data is uniformly stored in a local temporary directory in a standardized data format. This allows subsequent query processes to directly access local data without re-accessing the original files, thereby improving execution efficiency and reducing the performance overhead of repeated file reads. Simultaneously, local storage provides a reliable data foundation for data traceability, anomaly detection, and log recording during process execution, ensuring good controllability and traceability of the entire review process at the data level.
[0031] In the multi-site automated batch query phase, the robot, based on the retrieved enterprise information, performs automated queries on 14 preset verification sites for each enterprise. These sites include lists of suppliers whose cooperation with the company is temporarily suspended, lists of Category I / II related enterprises, Credit China, Qichacha, China Government Procurement Network, National Enterprise Credit Information Publicity System, Supreme People's Court China Enforcement Information Disclosure Website, Ministry of Emergency Management's blacklist of untrustworthy entities in work safety, National Construction Market Supervision Public Service Platform, Work Safety License Inquiry Platform, National Energy Administration Qualification and Credit Platform, Surveying and Mapping Qualification Inquiry Platform, Geological Hazard Risk Assessment Qualification Inquiry Platform, and Fire Protection Qualification Inquiry Platform. During execution, the system constructs a query task queue based on the parsed enterprise data, sequentially traversing all target sites configured to participate in the query for each enterprise. A unified scheduling mechanism ensures that query tasks at each site are executed stably in a predetermined order. This dual-loop structure of enterprise and site dimensions ensures both comprehensive verification of multi-dimensional information for a single enterprise and batch processing capabilities for multiple enterprises, giving the query process scalability and standardization. Meanwhile, the data sources from different sites cover multiple dimensions such as credit information, judicial information, qualification information and industry regulatory information, making the overall verification results more comprehensive and effectively supporting subsequent review and judgment.
[0032] For different query sites, the robot executes customized automated operation logic. Core operations include: opening the target site in Chrome browser, automatically inputting the company name / key information, calling a third-party CAPTCHA recognition platform (Tujian) to automatically recognize and verify the CAPTCHA, automatically logging into platforms requiring accounts (such as Qichacha and the National Enterprise Credit Information Publicity System, with account and password read from a configuration table), clicking the query / filter button, and matching query results that match the company name exactly. During execution, each site is pre-configured with corresponding automated operation processes to adapt to differences in website page structure, interaction methods, and login mechanisms. Browser-driven control of page opening and element positioning enables page-level automation. An automatic input mechanism fills in the company name or unified social credit code into the query box, ensuring accurate and consistent query conditions. For websites with CAPTCHA verification, a third-party CAPTCHA recognition platform is introduced to automatically recognize and submit the CAPTCHA, reducing manual intervention. For websites requiring login, the login process is automatically completed by reading account information from the configuration table, ensuring the availability of query permissions. In the query results retrieval stage, by matching result items that are exactly the same as the company name, the risk of misjudgment caused by similar or identical companies is avoided, thereby ensuring the accuracy and consistency of the data source.
[0033] After completing the queries for each site, the query results for each site are standardized: if multiple query results exist, the result with an exact match to the company name is used; if no query results are found, record "no results found" in the results table; if the queried site is a webpage, record the URL of the result page in the corresponding field of the results table; simultaneously, take a full-screen screenshot of all query result pages and save it to a local temporary folder. During the standardization process, the query results are first filtered and standardized, and the data returned by different sites is uniformly converted into a structured format for storage to ensure consistent data format when generating subsequent reports; for cases with multiple records, a unique target result is selected through strict name matching rules to avoid misselection; for cases where no information is found, "no results found" is explicitly marked to ensure data integrity and interpretability; for webpage-type results, the access URL is recorded to make the query process traceable; simultaneously, full-screen screenshots of the query pages are taken to save the original evidence in image form, providing intuitive evidence for subsequent review. Screenshot files are uniformly stored in a local temporary directory and categorized by company and site for easy retrieval and archiving later.
[0034] For specialized sites like Qichacha, personalized operations are implemented: if an SMS verification code is displayed during login, the process is paused and the user is notified to enter it manually; after the query is completed, the user returns to the homepage to perform a relationship search operation. For multiple companies, the data is entered and searched in batches; for more than 10 companies, the search is split and iterated, and the relationship search results are exported as images; clicking the "more" button for the company's phone number or email address allows for separate screenshots to record detailed information. During this process, for login scenarios involving secondary verification, the process is paused and a manual verification code is prompted to ensure successful account login without affecting the overall security of the process. After completing the basic information query, the relationship search function is used to further obtain information about the connections between companies. When there are many companies, batch input and batch processing strategies are used to ensure stable execution of the query operation within the platform's limitations, and the generated relationship graph results are exported and saved as images. Simultaneously, for detailed information such as company contact information, more options are expanded and separate screenshots are taken to ensure more complete information collection. Through these personalized processing mechanisms, the automated process can adapt to the special interaction logic of different sites, ensuring automation while maintaining execution stability and information integrity in complex scenarios.
[0035] During the query result aggregation and report generation phase, the robot generates a standardized query result table based on the user-provided query data table, entering the text results and URL information from each site into the corresponding fields. In the specific execution process, the robot uses the enterprise basic information and query configuration data read and stored locally in the early stages as the basic data source to uniformly aggregate and process various results generated during the multi-site query phase. For each enterprise's query results at each verification site, the text information is filled into the corresponding positions in the result table according to predetermined field rules. Simultaneously, the URLs of the result pages returned by each site are recorded, ensuring the traceability of each query result. During data processing, a unified data structure is used to standardize query results from different sources and in different formats, ensuring the completeness and consistency of field content in the result table and avoiding information confusion caused by differences in data sources. For cases where no results are found, the result table is clearly marked as "no results found," thus ensuring the completeness and logical consistency of the result data. Through the above processing, a standardized query result table with a clear structure and unified fields is formed, providing a reliable data foundation for subsequent report generation and facilitating subsequent statistical analysis, audit review, and historical archiving of the query results.
[0036] After generating the query results table, the system automatically inserts screenshots of the query results from each site into the corresponding positions in a pre-defined Word template, generating a complete bidder qualification review report. During this process, the pre-defined Word template standardizes the overall report structure, chapter content, and the display position of information from each site. The robot automatically inserts the screenshot files saved in previous steps into the designated positions according to the company and site correspondences, ensuring a one-to-one correspondence between the report content and the query results. During insertion, the screenshot order is kept consistent with the query order to ensure clear report logic and hierarchy. The text information in the results table is used to supplement the screenshot content, providing both visual evidence and structured description. In multi-company scenarios, the system generates corresponding report content sequentially according to the company dimension, achieving an orderly presentation of information from multiple companies within the same report. This template-based generation method maintains consistency in report format, content structure, and information presentation, avoiding layout chaos, information omissions, or sequence errors that may occur during manual compilation, thus significantly improving report generation efficiency and standardization.
[0037] After the report generation is completed, the robot packages the review report (Word) and the query result sheet (Excel) into a ZIP archive, names it according to the rule of "Bidder Qualification Review Report-yyyy-MM-dd-hh-mm-ss-Sender Name", and saves it to the Output folder in the robot's working directory. During the packaging process, the generated Word review report and Excel query result sheet are uniformly archived as core output files and encapsulated by compression, which not only reduces the file size, but also facilitates subsequent transmission and storage. In the naming rule, the introduction of timestamp information ensures the uniqueness of each generated file and avoids the problem of file overwriting. At the same time, adding the name of the email sender realizes the identification of the task source, which facilitates subsequent query and management. In terms of the saving path, all files are uniformly stored in the preset Output folder, so that all output results are centrally managed, which is convenient for subsequent process invocation and historical data archiving. After this stage is completed, the query results are transformed from the original scattered state into a combined form of structured files and standardized reports, which not only realizes the transformation from data to achievements, but also provides a standardized output carrier for subsequent automatic push and business use.
[0038] In the automatic push stage of review results, the robot automatically opens instant messaging software (Feishu, DingTalk, WeCom, etc.), retrieves the specified group chat, sends the named archive to the group chat to complete the push of review results, and the process ends. During the execution process, the robot first starts the corresponding instant messaging software client based on the preset automatic operation logic, enters the contact or group chat retrieval interface through interface recognition or a preset path, and automatically completes the positioning and opening of the target group chat according to the pre-set group chat name or identification information in the configuration. After the group chat is positioned, the robot reads the archive file that has been generated and named according to the rules in the previous stage from the local working directory, and performs the file sending action by simulating manual operations, including a series of operation processes such as clicking the attachment button, selecting the local file path, and confirming sending, so as to realize the automatic delivery of the file.
[0039] During the file sending process, the unified naming rule ensures that the recipient can directly identify the report generation time and the corresponding task source from the file name, for example, the specific generation time of the report is clarified through the timestamp information, and the task initiator is identified through the email sender name, thereby improving the efficiency of information identification and the convenience of subsequent management. At the same time, the archive contains two core files, the review report and the query result sheet, so that the recipient can obtain the complete review results in the group chat without additional retrieval or supplementary information, thereby improving the completeness and timeliness of information transmission. In the scenario of multiple task executions, the automatic push mechanism can realize batch sending of continuous task results, avoiding the efficiency reduction problem caused by manual operation one by one.
[0040] Once the push notification is complete, the entire review process ends, forming a closed loop from data input, automated querying, result organization to outcome distribution. This closed-loop process ensures that review results are delivered to relevant personnel immediately after generation, reducing intermediate waiting steps and improving overall business collaboration efficiency. Simultaneously, using instant messaging tools as the result distribution medium facilitates information sharing; relevant personnel can directly view, download, and discuss the results in group chats, thereby improving team collaboration efficiency. In this stage, automated push notifications not only automate the output of results but also extend the process's endpoint, ensuring that the entire review process not only completes data processing tasks but also information transmission tasks, guaranteeing that review results can be utilized promptly.
[0041] To enable those skilled in the art to fully understand and implement this technical solution, the following detailed embodiments of the RPA-based intelligent bidder qualification review method and system are provided, taking a practical application scenario as an example. This embodiment uses a company's bidder qualification review business as an application scenario, with the robot's working directory set to D:\RPA\Business Applications\Automatic Review Tools, and the core application engine being RPA.
[0042] Example 1: In practical application, the process begins with data preparation and process triggering. Bidding personnel fill in basic information such as the names and unified social credit codes of three companies to be queried using a pre-set Excel template. They also mark 10 websites, including Credit China, Qichacha, and the National Enterprise Credit Information Publicity System, as "yes" in the Sheet database. Additionally, they supplement the list with five companies (five in total) of suppliers whose cooperation with them is temporarily suspended, and three companies (three in total) of related companies (three in total). This Excel file is then sent as an attachment to the trigger email address. Upon receiving this email, the RPA robot's email trigger automatically initiates the review process. During this process, the standardized Excel template provides uniform constraints on data entry, ensuring a consistent data structure for company information, query configurations, and auxiliary verification data. Marking query websites allows for flexible control of the query scope. Supplementing the supplier and related company lists expands the review dimensions, making the review results more comprehensive. Email serves as the carrier for data transmission and process triggering, enabling business requests to be submitted in a unified manner. An automatic monitoring mechanism automatically initiates the process, reducing manual intervention and improving overall execution efficiency.
[0043] The process then proceeds to the data verification and retrieval stage. The robot retrieves the aforementioned email, verifies that the attachment is in .xlsx format and that the data structure matches the preset template, thus passing the verification. It then reads the information of three companies to be queried, the configurations of ten sites to be queried, and the list of suppliers and related companies for auxiliary verification from the Excel template, storing the data locally in the D:\RPA\Business Applications\Automatic Review Tools\Temp folder. During this process, dual verification of file format and data structure ensures that the input data conforms to preset specifications, preventing abnormal data from entering subsequent processes. In the data retrieval stage, the basic information of enterprises, site configuration data, and auxiliary verification data are parsed and converted into structured data for storage. Local storage allows subsequent query processes to directly access local data, improving execution efficiency, while also providing data support for log recording and anomaly troubleshooting, thereby ensuring the stability of the process.
[0044] Next, automated batch queries were performed across multiple sites. The robot performed automated queries on 10 sites for each of the 3 companies: It accessed the Credit China website, entered the company name, called the image database interface to recognize the verification code and complete the verification, clicked "Search," matched results with identical company names, clicked the "Serious Dishonesty" button and took a full-screen screenshot, and recorded the URL of the result page in the result table; it accessed the Qichacha website, read the account and password from Config.xlsx to complete automatic login, entered the company name and clicked "Search," matched the query results and took a screenshot of the company information page, clicked the "More" button for phone and email and took screenshots separately, then returned to the homepage to perform the "Find Relationships" operation, batch-entered queries for the 3 companies and exported a screenshot of the relationship diagram; it accessed the National Enterprise Credit Information Publicity System, completed automatic login and verification code recognition, entered the company name to search, clicked the "Included in the Seriously Dishonest List (Blacklist)" information button and took a screenshot; the remaining sites performed automated queries according to a customized script, and all results were matched, screenshots were taken, and text information was recorded. Sites with no query results were recorded as "No Query Results" in the result table. In this phase, browser automation controls page access and data input, CAPTCHA recognition technology addresses access restrictions, automatic account login ensures query permissions, and company name matching rules guarantee the accuracy of query results. Customized scripts are used to adapt to the differences in page structure and interaction methods across different websites, ensuring stable execution of queries across multiple platforms. Complete query results are generated by combining screenshots and text records.
[0045] After the query is completed, the process moves to the result aggregation and report generation stage. Based on the data table to be queried, the robot generates an Excel query result table, inputting text results and URL information from 10 websites of 3 companies. Following a preset Word template, screenshots from each website are automatically inserted into their corresponding positions, generating a "Bidder Qualification Review Report." The Word report and Excel result table are packaged into a ZIP archive, named "Bidder Qualification Review Report 2024-03-10 10:20:30 - Zhang San.zip" (Zhang San being the email sender), and saved to the D:\RPA\Business Applications\Automatic Review Tools\Output folder. In this process, standardized result tables achieve unified aggregation of data from multiple websites; a templated report generation mechanism enables automatic combination of screenshots and text information, ensuring consistency in the structure and format of the output report; and file packaging and naming rules ensure unique identification and traceability of the output results, while centralized storage facilitates subsequent management and use.
[0046] Finally, the review results are automatically pushed out. The robot automatically opens instant messaging software, retrieves the pre-configured target business group chat, and sends the named compressed file to the group chat, completing the process. At this stage, the results are distributed automatically, delivering the review results to the designated recipients without human intervention, achieving instant information sharing. Pushing the results via group chat allows relevant personnel to quickly obtain the report and carry out follow-up work, thus forming a complete closed-loop process from data input, automatic query, result generation to result push, significantly improving overall business processing efficiency and collaboration capabilities.
[0047] Example 2: In another implementation scenario, flexible application can be achieved by configuring query sites to meet different query needs. This method and system allow users to configure query sites as needed. If bidding personnel only need to verify the bidder's credit status and qualification certificates, they can mark only "Credit China," "Supreme People's Court China Enforcement Information Disclosure Website," "National Construction Market Supervision Public Service Platform," and "Safety Production License Inquiry Platform" as "Yes" in the site list of the Excel template. After the robot reads the configuration information, it will only perform automated queries on these four sites, without operating on the remaining sites. The generated review report will only contain the query results from these four sites, achieving flexible adaptation to query needs. In this implementation, by selectively marking the site configuration table, dynamic control of the query scope is achieved, allowing the review process to be tailored according to actual business needs. When the business only focuses on credit records and qualification status, it is not necessary to execute queries on all sites, thereby reducing invalid query operations and improving overall execution efficiency. Simultaneously, since the query sites are driven by configuration data, query strategy adjustments can be completed without modifying the program logic, enhancing the system's adaptability to different business scenarios. In this way, different users can choose the appropriate data sources according to project type, industry requirements, or review focus, making the automated process highly flexible and scalable, and achieving the rational use of resources while ensuring the accuracy of the results.
[0048] In practical applications, comparing automated review with traditional manual review methods clearly demonstrates the advantages of automated solutions. Taking the review of the qualifications of 5 bidders (involving 10 verification sites) in a real-world business scenario as an example, the effectiveness of manual review and RPA intelligent review is compared, with the following results: In terms of comparative metrics, manual review takes approximately 8 hours per person, while RPA intelligent review can be completed in about 20 minutes; regarding the review error rate, the manual method has an error rate of approximately 15% (mainly due to omissions, incomplete screenshots, etc.), while the automated execution process achieves standardized operations with a 0% error rate; in terms of human resource input, manual review requires one professional to operate the entire process, while automated execution requires no human intervention, with the robot automatically completing the entire process; regarding report format, manual compilation results in inconsistencies and disorganized information archiving, while automatically generated reports use standardized templates, have a clear structure, and are easy to trace; regarding anomaly handling, manual methods rely on personnel to discover and handle anomalies, a time-consuming and non-standardized process, while the automated process can identify and handle anomalies according to rules, while simultaneously recording logs and screenshots, improving processing efficiency and standardization.
[0049] The above comparison demonstrates that automated review methods outperform manual review methods in several key indicators. In terms of efficiency, automated batch execution significantly reduces processing time; in terms of accuracy, standardized processes prevent human error; in terms of manpower, automation reduces human involvement; in terms of consistency, templated output ensures consistent results; and in terms of anomaly handling, rule-based mechanisms improve process stability. These advantages collectively highlight the significant value of this technical solution in practical applications, enabling it to play a crucial role in bidding and tendering review processes and demonstrating promising prospects for widespread adoption.
[0050] This invention constructs a batch automated query mechanism for multiple sites and multiple enterprises, transforming the cumbersome process of manually logging into various platforms, manually entering information, and verifying each item into an automated execution process. Under a unified data input and triggering mechanism, it enables parallel or sequential automatic queries of multiple bidding companies at multiple verification sites, significantly reducing review time. Work that originally required hours or even days can be completed in tens of minutes, significantly improving the overall efficiency of bidder qualification review and effectively meeting the strict timeliness requirements of bidding and tendering business.
[0051] This invention effectively avoids problems such as input errors, missing queries, result matching deviations, and incomplete screenshots in traditional manual operations by using standardized data verification mechanisms, precise enterprise name matching rules, and automated execution processes. During the query process, a structured data processing method and an automated result collection method are uniformly adopted to ensure the consistency and integrity of various data sources, thereby comprehensively improving the accuracy and reliability of the review results and ensuring the authenticity and objectivity of the bidder qualification review conclusions.
[0052] This invention automates the repetitive and mechanized tasks of retrieval, screenshotting, and organization, eliminating the need for reviewers to participate in the specific query process. This significantly reduces human intervention, lowers reliance on professionals, and reduces the company's human resource costs. It also frees up relevant personnel from low-value repetitive labor, allowing them to focus on high-value tasks such as bidding and tendering decision analysis, thereby further improving the overall operational efficiency of the company.
[0053] This invention constructs standardized input and output templates to achieve unified and standardized input of data to be queried and unified and standardized output of review results. During the report generation process, it automatically completes data filling, screenshot embedding, and formatting, ensuring that the generated bidder qualification review reports maintain consistency and standardization in terms of structure, content, and format. This avoids information confusion and inconsistent formatting issues that occur during manual processing, thereby improving the readability, traceability, and subsequent management efficiency of the review results.
[0054] This invention establishes a comprehensive exception handling mechanism to classify, identify, and address various anomalies that may occur during process execution, including attachment format anomalies, site access anomalies, login anomalies, and query result anomalies. When an anomaly occurs, it can perform operations such as retrying, logging, and saving screenshots of the anomaly, effectively preventing single points of failure from causing overall process interruption, significantly improving the stability and fault tolerance of the review process, and ensuring the continuous and reliable operation of the system in complex environments.
[0055] This invention uses a configuration management mechanism to uniformly maintain various parameters required for system operation, enabling flexible configuration of query sites, account information, and operating parameters. It supports the selection of different verification site combinations according to different business scenarios, adapting to the bidder qualification review needs of different industries and enterprises. System adaptation can be completed without changing the overall process structure, possessing strong flexibility and scalability, which is conducive to the promotion and application of the system in multiple scenarios.
[0056] This invention utilizes an automated result push mechanism to send the results directly to a designated group chat via instant messaging tools after the review report is generated and packaged. Report distribution can be completed without manual intervention, achieving closed-loop management of the entire process from data acquisition and review execution to result output and sharing. This improves information transmission efficiency and team collaboration efficiency, making the bidding process more efficient and smooth.
[0057] This invention provides, for example Figure 2 The illustrated RPA-based intelligent bidder qualification review system includes a triggering module, a data processing module, a multi-site query module, a report generation module, a result push module, an exception handling module, and a configuration management module. The trigger module monitors the specified trigger email address in real time, identifies emails that meet the agreed requirements and retrieves the standardized Excel attachments within them, and initiates the bidder qualification review process after detecting emails that meet the conditions. The data processing module performs format and data structure validation on the standardized Excel attachments obtained by the triggering module. After the validation passes, it reads the data to be queried, configuration data, and auxiliary verification data, parses them, and stores them locally. The multi-site query module, based on the query data parsed by the data processing module, calls the built-in automated operation scripts of multiple verification sites, performs batch query operations on each enterprise through browser driver, CAPTCHA recognition interface and platform login component, and completes query result matching, screenshot saving and text result recording; The report generation module summarizes and processes the query results obtained from the multi-site query module, fills the text information and screenshot information into the preset Word review report template and Excel result table template, generates the bidder qualification review report and corresponding result table, and completes file packaging and standardized naming; The results push module automatically sends the packaged file generated by the report generation module by searching the specified group chat through instant messaging software and pushing the compressed file. The exception handling module classifies and identifies system and business exceptions generated during each processing step, handles them accordingly, and records operation logs and saves exception screenshots. The configuration management module provides a configuration table to store account information, trigger email information, group information, working directory and site access address parameters, and supports parameter maintenance and updates.
[0058] This invention provides an intelligent bidder qualification review method based on RPA, which is implemented through the aforementioned intelligent bidder qualification review system based on RPA. For details of the specific method and process of the intelligent bidder qualification review system based on RPA, please refer to the aforementioned embodiment of the intelligent bidder qualification review method based on RPA, which will not be repeated here.
[0059] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A method for intelligent qualification review of bidders based on RPA, characterized in that, Includes the following steps: Obtain a standardized Excel template file sent via email. The standardized Excel template file contains basic information of the enterprise to be queried, query site configuration data, and auxiliary verification data. Monitor the specified trigger email address and initiate the intelligent review process for bidder qualifications upon receiving an email that meets the agreed requirements. The system performs format and data structure validation on the obtained standardized Excel template file. After the validation passes, it reads the enterprise information to be queried, the query site configuration data, and the auxiliary verification data, and completes the local storage of the corresponding data. Based on the read enterprise information and query site configuration data, automated query operations are performed on each enterprise on multiple preset verification sites. This includes accessing the target site through a browser, entering the enterprise name or key information, recognizing and verifying the CAPTCHA, completing account login, executing the query and matching query results that are exactly the same as the enterprise name, and at the same time, standardizing the query results and generating corresponding result records and screenshot files. The obtained query results and screenshots are summarized and processed to generate a standardized query results table. The corresponding screenshots are inserted according to the preset Word template to form a bidder qualification review report. The review report and query results table are packaged and named according to preset rules before being stored. After the packaged storage is completed, the target group chat is automatically retrieved and the packaged file is sent via instant messaging software, thereby automatically pushing the bidder qualification review results.
2. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, To build an automatic startup logic around task identification and trigger control during the email receiving process, the steps are as follows: Receive email messages containing standardized Excel attachments, extract the email subject field, attachment file name, sending timestamp, and sender address information, and perform structural parsing on the email content; The email subject field is matched against the preset trigger rules one by one, and the format of the attachment file name is validated according to the rules. At the same time, the number of attachments is checked to see if it meets the preset conditions. Parse the sender's address and generate a task identifier code, bind the task identifier code to the email metadata, and write it to the task queue cache structure; Write the task identifier code into the trigger log recording area, call the process scheduling interface to enter the subsequent processing link, and maintain the task status marking information at the same time.
3. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, To address the anomaly detection requirements during the data input phase, a structure verification and termination process is introduced, with the following steps: Load email attachments and parse the file header information to identify the file extension, file encoding format, and file integrity status; The number of worksheets, worksheet naming rules, and field column order in the Excel file are compared item by item, and the field types are checked to ensure they meet the preset requirements. Perform a row-by-row scan on key fields, including null value detection, character length detection, and data format validation, and mark the index of abnormal fields. The exception information is written to the exception log table, and the process interruption control signal is triggered at the same time. A copy of the exception file and the corresponding log data are saved.
4. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, Based on the variable requirements of query tasks, the site configuration parsing method is extended, and the steps are as follows: Read the site data area of the Excel template and extract the site identifier field and configuration value information for each row; Perform a tagging and identification operation on each configuration value, filter the records marked as "yes" as valid sites, and record their corresponding index numbers; The filtered results are converted into an ordered list structure, and a mapping table between site identifiers and access addresses is established. The query task sequence is generated based on the mapping relationship and written into the task execution queue structure in sequence for subsequent scheduling.
5. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, To address the issue of operational consistency during multi-platform queries, a unified execution logic is constructed, with the following steps: The browser driver component is invoked to initialize the runtime environment and load the target site URL, while waiting for the page elements to finish loading. Enter the company name or unified social credit code in the input control on the page, and perform input content validation and format correction operations; The system identifies the image area of the CAPTCHA on the page and calls a third-party interface to perform image recognition processing, while filling the recognition result into the CAPTCHA input box. Trigger the query button on the page and listen to the page response status, obtain the returned data content, locate the result list area and perform data extraction.
6. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, To ensure consistency in query result processing, a standardized output method is constructed, following these steps: Perform a line-by-line fetching operation on the result list returned by the query, and parse the company name field and additional information field in each record; Perform exact string matching on the parsed results, filter out records with exactly the same company name and mark them as target data; For cases where no match is found, a preset identifier text is written, and the current page URL is extracted and written into the result data structure. Perform a full-screen screenshot of the page and generate an image file, while recording the file storage path and associating it with the corresponding result record.
7. The RPA-based intelligent bidder qualification review method according to claim 6, characterized in that, A record consistency verification mechanism is introduced during the query result processing. The target data fields are associated with and verified with the screenshot file path. Consistency comparison is performed on the matching results of enterprise names. At the same time, the correspondence between the page address in the result record and the screenshot source is verified. The data that passes the verification is written into a unified result structure, thereby completing the corresponding storage of result data and image information.
8. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, In line with the requirement for a unified structure in the review result output process, a template-based processing workflow is introduced, with the following steps: The query results from each site are aggregated and processed, integrating text information fields, URL information fields, and screenshot path information; The summary results are written into the corresponding cell range in the Excel result table according to the field mapping rules, and data formatting is uniformly processed at the same time. Load the Word template file and parse the preset placeholders in the template, then insert the screenshot file into the specified position according to the corresponding relationship; The generated file is compressed, and a file name is generated and written to the storage path by combining the time information, task identifier and sender information.
9. The RPA-based intelligent bidder qualification review method according to claim 1, characterized in that, To address the need for automated execution of the review results distribution process, an information push workflow is constructed, with the following steps: Launch the instant messaging software client and load the user interface, while simultaneously entering the group chat search input area; Enter the target group chat name and perform the matching operation to locate the corresponding group chat window and open it. Read the local compressed file path and perform attachment loading operation, while verifying the existence and integrity of the file; Trigger the send operation and complete the file upload process, while recording the send time and send status information and writing it to the log system.
10. An RPA-based intelligent bidder qualification review system, used to implement the RPA-based intelligent bidder qualification review method described in any one of claims 1-9, characterized in that, It includes a triggering module, a data processing module, a multi-site query module, a report generation module, a result push module, an exception handling module, and a configuration management module. The trigger module monitors the specified trigger email address in real time, identifies emails that meet the agreed requirements and retrieves the standardized Excel attachments within them, and initiates the bidder qualification review process after detecting emails that meet the conditions. The data processing module performs format and data structure validation on the standardized Excel attachments obtained by the triggering module. After the validation passes, it reads the data to be queried, configuration data, and auxiliary verification data, parses them, and stores them locally. The multi-site query module, based on the query data parsed by the data processing module, calls the built-in automated operation scripts of multiple verification sites, performs batch query operations on each enterprise through browser driver, CAPTCHA recognition interface and platform login component, and completes query result matching, screenshot saving and text result recording; The report generation module summarizes and processes the query results obtained from the multi-site query module, fills the text information and screenshot information into the preset Word review report template and Excel result table template, generates the bidder qualification review report and corresponding result table, and completes file packaging and standardized naming; The results push module automatically sends the packaged file generated by the report generation module by searching the specified group chat through instant messaging software and pushing the compressed file. The exception handling module classifies and identifies system and business exceptions generated during each process, handles them accordingly, and records operation logs and saves exception screenshots. The configuration management module provides a configuration table to store account information, trigger email information, group information, working directory and site access address parameters, and supports parameter maintenance and updates.
