Automatic bid invitation file auditing method and system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based bid invitation file automatic auditing method and system, and the method comprises the steps: setting an online database and a local database in the system, carrying out the format auditing, compliance auditing and other abnormal auditing of a bid invitation file through a bid invitation file auditing module, generating auditing suggestions, and carrying out the automatic auditing of the bid invitation file. After the auditing opinions are manually confirmed by a system using unit, a final bid invitation file manuscript is generated, and the system comprises the following modules: a database module, a file uploading module, a bid invitation file auditing module, a suggestion modification module, a bid invitation file generation module, a user interaction interface and a learning module. The method is convenient to use, intelligent auditing of the bid invitation file is achieved through the artificial intelligence technology, and the overall quality and efficiency of the bid invitation file are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of office automation, and in particular relates to an automatic review method and system for bidding documents based on artificial intelligence. Background Art

[0002] At present, bidding procurement has become a special procurement method and has been widely used in various projects in various industries and fields. Under this transaction method, the purchaser of the project procurement (including the purchase of goods, the contracting of projects and the procurement of services) is usually the tenderer, who issues the information of bidding procurement by issuing bidding documents, etc., and proposes the nature of the required procurement project and its quantity, quality, technical requirements, delivery date, completion date or service provision time. After the tenderer reviews and compares the quotations and other conditions of each bidder, it selects the best bidder and signs a procurement contract with it. Specifically, the purchaser generally provides the bidding requirements to the third-party bidding agency, and the third-party bidding agency sends professionals to manually compile the bidding documents according to the bidding requirements provided by the purchaser. After the preparation is completed, the bidding documents are published and announced to notify potential bidders to bid, so as to select the best bidder and sign a procurement contract with it.

[0003] However, after the country abolished the qualification of bidding agents in 2017, a large number of bidding agencies of varying quality have sprung up like mushrooms after a rain. Many bidding agencies have weak strength, insufficient experience, weak manpower, and inadequate understanding of laws and regulations. Their bidding documents are of low quality and poor level, and the manual writing method of bidding documents is inefficient. How to improve the writing quality and efficiency of bidding documents has become an urgent problem to be solved. Therefore, it is of great significance to invent an automatic review method and system for bidding documents based on artificial intelligence, which is simple to use, efficient, high-quality and can continuously learn independently. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an automatic review method and system for bidding documents based on artificial intelligence, so as to improve the quality and efficiency of bidding document writing through autonomous learning of artificial intelligence.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] An automatic review method for bidding documents based on artificial intelligence includes the following steps:

[0007] Step 1: Set up an online database in the system, automatically search and download bidding-related regulations and bidding document templates from the Internet from time to time, and mark them with keywords. The online database is shared by the system users;

[0008] Step 2: Set up a local database in the system, and the system user will manually enter the bidding regulations and project bidding document templates of its specific industry and mark them with keywords. The local database can be shared with all system users after the system user agrees;

[0009] Step 3: The system user uploads the draft of the bidding document that needs to be reviewed. The bidding document review module conducts format review, compliance review and other abnormal reviews on it, and then generates review opinions; the review opinions include the issues found in the review and the specific modification opinions and reasons for each issue;

[0010] Step 4: Based on the audit opinions described in step 3, the system user unit manually confirms the audit opinions and generates the final tender document. The manual confirmation can be to accept the audit opinions and make modifications, to reject the audit opinions and not make modifications, or to make adjustments or partially adopt the audit opinions.

[0011] Furthermore, step 5 may also be included: automatically entering the final draft of the bidding document generated in step 4 into the local database, and recording the result of manual confirmation of the review opinion by the system user in step 4, by continuously recording the result of manual confirmation by the user unit, training the machine to learn the style of the user unit in compiling the bidding document, and optimizing the review opinions proposed in step 3 through the learning model.

[0012] Furthermore, the bidding-related provisions described in step 1 include relevant provisions such as national laws, regulations, rules, normative documents, national standards, industry standards, etc. in the bidding field; the bidding document template is a standard bidding document or bidding document demonstration text issued by the state or local area, and a bidding document for projects of various industries, regions, and types; and the marking keywords described in step 1 include the title of the bidding-related provisions, effective date, expiration date, issuing unit, issuing date, document category, document version, nature of the bidding document, applicable area, regional exclusivity, applicable industry, application scenario, project type, bidding stage, scope of application, contract type, legal hierarchy, legal priority hierarchy, applicable priority, compliance requirements, key terms, penalty measures, applicable subjects, document applicable conditions, additional terms, implementation details, and other keywords.

[0013] Furthermore, the bidding-related provisions for specific industries in step 2 include relevant provisions such as national laws, regulations, rules, normative documents, national standards, industry standards, etc. in the bidding field, and the project bidding document template includes standard bidding documents or bidding document sample texts issued by the state or local areas, and project bidding documents of various industries, regions, and types; and the marking keywords described in step 2 include the title of bidding-related provisions, effective date, expiration date, issuing unit, publication date, document category, document version, nature of bidding documents, applicable area, regional exclusivity, applicable industry, application scenario, project type, bidding stage, scope of application, contract type, legal hierarchy, legal priority hierarchy, applicable priority, compliance requirements, key terms, penalty measures, applicable subjects, document applicable conditions, additional terms, implementation details, and other keywords.

[0014] Furthermore, the compliance audit and other abnormal audits described in step 3 are reviewed by searching, comparing, and retrieving data from online databases and local databases.

[0015] Among them, the review opinions on the bidding documents include: ① whether there are any violations of relevant bidding regulations; ② at the same time, horizontally compare the bidding documents of previous projects in the local database and prompt abnormal bidding documents; ③ whether there are any contradictions, loopholes in contract terms, defects in management models, etc.;

[0016] Among them, abnormal bidding documents specifically refer to:

[0017] 1. The content is incomplete or missing, mainly due to missing clauses (the bidding documents lack key clauses that must be included in laws, regulations, industry standards or practices, such as confidentiality clauses, contract performance clauses, acceptance criteria, etc.), incomplete information (the documents do not contain basic project information, such as project name, budget amount and source of funds, or the bidding content is not described in detail as required) and missing attachments (the attachments required by the bidding documents, such as drawings, technical parameters and bidding instructions, are not attached, or the appendix documents are incomplete).

[0018] 2. The terms or conditions are set unreasonably, mainly because the threshold conditions are too high or too low (the requirements for the bidder's qualifications, performance, etc. are set too high or too low, which is obviously inconsistent with industry practices or regulations, and may be biased or exclusive), the technical parameters are too specific (the technical specifications or product requirements are too detailed, which may restrict potential bidders, and are suspected of being tailor-made for specific manufacturers, violating the principles of openness, fairness and impartiality) and the scoring criteria are unreasonable (the scoring criteria lack objectivity, which may be unfavorable to some bidders, or the scoring items are unbalanced, such as the price weight is too high or too low).

[0019] 3. Abnormal time nodes, mainly due to insufficient announcement time (failure to reserve sufficient announcement period or bid preparation period in accordance with regulatory requirements, such as the 20-day announcement period required by law) and conflicts in bidding schedules (opening time, deadline, clarification time, etc. do not conform to reasonable procedures or conflict with time nodes in other bidding documents).

[0020] 4. The bidding amount and budget are abnormal, mainly due to unreasonable budget (the project budget is too high or too low compared with the market price, especially in equipment procurement, engineering cost, etc., there may be abnormal quotations) and unclear sources of funds (the source of funds is not clear or the source of funds does not meet the requirements, which may affect the project's financial security and contract performance security).

[0021] 5. Legal and compliance issues, mainly including the citation of outdated regulations (the legal clauses or standards cited in the document have expired or become invalid, which may make subsequent contract clauses unenforceable), non-compliance with new regulations (failure to update to the current effective regulatory requirements in a timely manner, or failure to include the latest compliance clauses) and priority conflicts (there are priority conflicts between the cited legal clauses, such as conflicts between local regulations and laws).

[0022] 6. Historical data anomalies are mainly due to inconsistency with past project data (document contents are significantly different in similar projects, such as large differences in the settings of bills of quantities, technical parameters, bidder qualification requirements, etc., which may imply biased or unfair settings) and significant differences in contract terms (contract terms vary greatly in similar projects, such as performance bonds, breach of contract liability, etc., and it is necessary to judge whether there is bias or special tendency).

[0023] 7. Price and cost anomalies, mainly abnormally low or high quotations (the system compares the quotation levels of similar projects. If the price deviates significantly from the market level, there may be suspicion of vicious competition or profit transfer) and abnormal setting of the upper or lower limit of the bid (the setting of the maximum or minimum price is unreasonable, which may lead to reduced competitiveness or benefit specific bidders).

[0024] 8. Violations of the bidding procedures mainly include non-compliant bidding methods (failure to select the appropriate bidding method in accordance with regulations, illegal selection of single-source procurement, etc.), improper prequalification (the prequalification process is not transparent, or there are situations where qualified bidders are excluded), and irregularities in the publication of bidding documents (failure to publish bidding documents through the prescribed announcement platform may affect openness and transparency).

[0025] 9. Related party or conflict of interest issues, mainly potential conflicts of interest (stakeholders may be involved in the documents, which may lead to unfair competition, such as the participation of the bidder’s related parties or stakeholders) and undisclosed related relationships (if the bidder has an undisclosed related relationship with the tenderer or the bid evaluation party, it may violate the fairness requirement).

[0026] 10. Recurrence of problems in historical projects, mainly the recurrence of historical anomalies (the system finds duplicate abnormal content in historical projects, such as setting the same unreasonable clauses, suggesting that there may be systemic problems) and recurrence of quality and performance issues (quality problems or performance issues that occurred in past projects were not mentioned or avoided in the bidding documents, and the contract terms were not optimized for historical problems).

[0027] The above anomalies help the system to identify potential problems in a targeted manner during the document review process and improve the standardization and transparency of the bidding process.

[0028] An automatic tender document review system based on artificial intelligence, including the following modules:

[0029] Module 1: Database module, which stores bidding-related regulations and bidding documents, and sets keywords for them to facilitate matching and comparison by the bidding document review module.

[0030] Module 2: File upload module, which uploads the draft of the bidding document prepared by the system user unit to the bidding document review module after authorization; and uploads the finalized bidding document confirmed by the system user unit to the local database.

[0031] Module 3: Tender document review module, which conducts format review, compliance review and other abnormal review on the draft tender documents uploaded by the system users.

[0032] Module 4: Modification suggestion module, which generates review opinions on the bidding documents reviewed by the bidding document review module. The review opinions include the issues found in the review and the specific modification opinions and reasons for each issue.

[0033] Module 5: The bidding document generation module automatically generates the final version of the bidding document based on the system user's processing results of the specific modification suggestions (such as acceptance, rejection, adjustment or partial adoption).

[0034] Module 6: User interaction interface, which facilitates system users to operate the system.

[0035] Module 7: Learning module, which is also the user style recording module, automatically compares the review opinions proposed by the system with the final version of the bidding documents, and records the opinions proposed by the system but not accepted by the system user. When the number of records of the same situation exceeds the threshold set by the system user, the system will automatically block such review opinions for the system user to avoid repeated reminders and reduce unnecessary workload; record the results of the system user's manual confirmation of the review opinions, and use these feedback data to train the machine learning model to gradually improve the accuracy and efficiency of the output of the bidding document review module.

[0036] Furthermore, the module 1 database module includes an online database and a local database. The online database automatically searches and downloads bidding-related regulations and bidding document templates from time to time through the Internet; the local database is manually entered by the system user unit into the bidding-related regulations and project bidding document templates of its specific industry.

[0037] The beneficial effects of the present invention are as follows: compared with the prior art, the present invention establishes an online database and a local database based on artificial intelligence, stores tender-related regulations and various templates of tender documents in advance, and generates review opinions on the tender documents after format review, compliance review and other abnormal reviews through the tender document review module. The final draft of the tender document is automatically generated after manual confirmation by the system user unit. At the same time, the present invention is also provided with a user style recording module to record the modification habits of the system user unit, automatically filter and shield, and is easy to use. The intelligent review of the tender documents is realized through artificial intelligence technology, which improves the overall quality and efficiency of the tender documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is the workflow and schematic diagram of the present invention. DETAILED DESCRIPTION

[0039] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0040] refer to Figure 1 , this embodiment adopts an automatic review method of bidding documents based on artificial intelligence, including the following steps:

[0041] Step 1: Set up an online database in the system, automatically search and download bidding-related regulations and bidding document templates from the Internet from time to time, and mark them with keywords such as the title of the bidding-related regulations, effective date, expiration date, issuing unit, issuing date, document category, document version, nature of the bidding document, applicable area, regional exclusivity, applicable industry, application scenario, project type, bidding stage, scope of application, contract type, legal level, legal priority level, applicable priority, compliance requirements, key terms, penalty measures, applicable subjects, document applicable conditions, additional terms, implementation details, etc. For example, the applicable industries can be divided into highway industry, railway industry, housing construction industry, municipal industry, energy, etc., the applicable areas can be divided into Guizhou Province, Yunnan Province, Chongqing City, etc., and the project types can be divided into construction, survey, design, supervision, equipment and materials, testing, construction approval, etc. The online database is shared and used by the system users.

[0042] The bidding-related regulations in step 1 include relevant regulations in the bidding field, such as national laws, regulations, rules, normative documents, national standards, industry standards, etc. The bidding document template is a standard bidding document or a bidding document demonstration text issued by the state or local government, and a bidding document for projects of various industries, regions, and types.

[0043] The online database can adopt a distributed database system, which can accommodate a large amount of data and ensure data security. In order to make the database design support efficient query and facilitate fast retrieval, the cleaned legal and regulatory data will be imported into Elasticsearch to improve search efficiency and accuracy; in addition, PostgreSQL is used to store structured data of laws and regulations, including metadata and keywords, so that the corresponding laws and regulations can be accurately obtained when marking and generating corresponding review opinions in erroneous bidding documents.

[0044] The Internet automatically searches and downloads data irregularly. In order to use web crawler technology, the latest bidding-related regulations and bidding document templates are automatically obtained from official government websites and other legal channels. The obtained data is cleaned, deduplicated and standardized, and key information such as the effective and expiration dates of the law, industry, region, type, etc. are extracted through natural language processing (NLP) technology, and this information is marked. The online database is automatically updated through a set period, which can be freely set, preferably once a week or once a month, to ensure the timeliness of the data.

[0045] Step 2: Set up a local database in the system, and have the system users manually enter the bidding-related regulations and project bidding document templates, and mark them with keywords such as the bidding-related regulations title, effective date, expiration date, issuing unit, issuing date, document category, document version, nature of the bidding document, applicable area, regional exclusivity, applicable industry, application scenario, project type, bidding stage, scope of application, contract type, legal level, legal priority level, applicable priority, compliance requirements, key terms, penalty measures, applicable subjects, document applicable conditions, additional terms, implementation details, etc. For example, the applicable industries can be divided into highway industry, railway industry, housing construction industry, municipal industry, energy, etc., the applicable areas can be divided into Guizhou Province, Yunnan Province, Chongqing City, etc., and the project types can be divided into construction, survey, design, supervision, equipment and materials, testing, and construction approval, etc. The local database can be shared with all system users with the consent of the user.

[0046] Among them, the bidding-related regulations for the specific industry in step 2 include national laws, regulations, rules, normative documents, national standards, industry standards and other relevant regulations in the bidding field, and the project bidding document template includes standard bidding documents or bidding document sample texts issued by the state or local area, and project bidding documents of various industries, regions and types.

[0047] The local database uses the easy-to-manage relational database PostgreSQL, which supports system users to upload data according to their own needs. Specifically, system users can upload bidding regulations and project bidding documents for specific industries through a graphical user interface (GUI), and manually add or confirm labels automatically marked by the system. At the same time, it supports classification, sorting and searching of data uploaded by system users to ensure data accuracy and ease of use.

[0048] Step 3: After the system user uploads the draft of the bidding document that needs to be reviewed, the bidding document review module will conduct format review, compliance review and other abnormal reviews on it, and generate a review opinion, which includes the issues found in the review and specific modification opinions and reasons for each issue.

[0049] Among them, the system user uploads the draft of the bidding documents that need to be reviewed through the graphical user interface (GUI) on the system.

[0050] Among them, format review is to use Apache POI to read the format of the uploaded file and format it, and extract style information from the file, including font, font size, color, alignment, etc. The format review process is to check whether the overall structure and layout of the file meet the requirements, define a set of rules to standardize the style and format of the file, and use algorithms to match the style and format of the file with predefined rules. The specific format review is as follows:

[0051] File format parsing: Use Apache POI to read the file type and version (such as DOC, DOCX, XLSX) to ensure that the file format matches the audit standard. If it does not meet the requirements (such as unsupported file type), an error will be prompted.

[0052] Page structure check: Analyze the document's layout structure such as the number of pages, paragraphs, columns, etc. Confirm whether the basic page structure of the file meets the requirements, such as whether the page columns or paragraphs are set correctly.

[0053] Tables, pictures, and other objects: Check whether the file contains tables, pictures, charts, and other objects, and record their location and number to ensure that there are no more or fewer than the expected number.

[0054] Read font and font size information: Traverse each paragraph or text portion in each paragraph in the document, and use Apache POI to read the text's font type, font size, bold, italic and other style information.

[0055] Main title and subtitle: According to regulations, verify whether the font and size of each level of title in the document meet the requirements. For example, the first-level title should be bold Songti 16, the second-level title should be bold Songti 14, etc.

[0056] Text font and size: Check the font type and size of the text to ensure that the text follows the standard (such as Songti size 12; non-standard styles such as color or bold are not allowed).

[0057] Color specification check: Read and parse information such as text color, background color, table line color, etc. to ensure that the colors meet the regulations. For example, all text colors should be black or other specific colors, and the background should be white.

[0058] Text alignment: Check the alignment of each paragraph in the document (such as left, center, right, or justified) to ensure that specific content (such as titles, table titles) meets the alignment requirements. For example, all body paragraphs should be justified and titles should be centered.

[0059] Paragraph spacing check: Check the upper and lower spacing of paragraphs to ensure that the paragraph spacing meets the specifications. For example, the body paragraph requires a blank line at the top and bottom, and there should be appropriate spacing between the title and the body.

[0060] Line spacing check: Verify whether the line spacing of each paragraph meets the requirements. For example, the main text requires single line spacing, and the title can be set to 1.5 times line spacing.

[0061] Page Margin Check: Check the margin settings around the page to make sure the margins are within the specified range. For example, the left and right margins are set to 3 cm, and the top and bottom margins are set to 2 cm.

[0062] Paper size check: Make sure the document's paper size meets the standard (such as A4 size 21x29.7 cm) and check whether the page orientation (portrait or landscape) meets the requirements.

[0063] Table border and style check: traverse the tables in the document, read table borders, cell colors, font sizes and other information, and ensure that the table style is standardized. For example, all tables should have a border line of 0.5 pounds and a font of Songti 10.

[0064] Image resolution and position check: Read and check the image's resolution, size, position, and wrapping mode to ensure that the image meets the specifications (such as resolution not less than 300 DPI and center alignment).

[0065] Title level check: Read and parse the title level relationship to ensure that the title uses the correct format. For example, the title level should be in descending order according to the specification, and the title level number (such as 1.1, 1.2, 2.1, etc.) meets the structural requirements.

[0066] Automatic numbering check: Ensure that the numbers in the file are generated by automatic numbering rather than manually entered, to avoid subsequent numbering confusion.

[0067] Identify and record errors: Record formatting items in the document that do not conform to the rules, including the specific location, problem description, and recommended formatting requirements.

[0068] Automatically generate audit reports: Generate detailed format check reports based on the test results, list all format items that do not comply with regulations, provide correction suggestions, and attach hyperlinks to help users quickly locate specific problem locations.

[0069] Format exception capture: If the document contains unsupported formats or styles (such as complex styles, footnotes, comments, etc.), the exception will be recorded in the report, or an unrecognizable format item will be prompted.

[0070] Fault-tolerance mechanism: Allow format deviations within a certain range and set format tolerances (such as the font size deviation is allowed to be ±1 point) to ensure the rationality of the audit results.

[0071] The above steps can help the system gradually realize a comprehensive check of the bidding document format and ensure that the file format meets the predefined specification requirements.

[0072] The compliance review and other abnormal reviews described in step 3 are reviewed by searching, comparing, and retrieving data from online databases and local databases. The specific steps are:

[0073] 1. Data source connection and data synchronization

[0074] Connect with online databases: Establish real-time or regular synchronization mechanisms with online databases to ensure real-time data. Data connection can be achieved using API interfaces or data exchange protocols (such as RESTful API, SOAP, GraphQL).

[0075] Synchronize with local database: Regularly update local data from online database and set data update mark to ensure that local database can run effectively even when offline or network is unstable.

[0076] Data mirroring backup: Create a mirror or backup for critical data so that it can still be read and called when the database is updated or fails.

[0077] 2. Multi-keyword search function

[0078] Keyword indexing: Establish full-text indexing and keyword indexing (using inverted indexing or hash indexing) for online and local database content to facilitate quick query, such as based on keywords such as effective time, expiration time, applicable region, industry, priority level, etc.

[0079] Fuzzy matching and word segmentation retrieval: Use word segmentation technology (such as Chinese-based jieba word segmentation or NLP-based word segmentation tools) and fuzzy matching to cope with a variety of expressions and non-standardized fields (such as "applicable to construction projects" can match "construction projects").

[0080] Combined search: allows combined queries based on multiple keywords, such as filtering effective time, region, industry and other complex conditions at the same time to ensure the accuracy of the search results.

[0081] 3. Data comparison and difference identification

[0082] Text comparison algorithm: Use the natural language processing (NLP) algorithm to perform semantic comparison between the text content of the bidding documents and the text content in the regulatory standards library to identify whether key terms comply with currently valid legal terms.

[0083] Multi-version comparison: For standard documents or regulations with multiple versions, establish a historical version database, use version control technology or multi-version comparison tools to identify the differences between different versions and avoid the old version content affecting the audit results.

[0084] Compliance rule base: Establish a compliance rule base, extract the key clauses and requirements of laws and standards into rules, compare the compliance of documents through rule matching technology, and detect whether they meet the legal and compliance requirements.

[0085] 4. Data retrieval and review mechanism

[0086] Real-time data call: Automatically call the latest regulations and standards data during the review process, compare and analyze the terms in the bidding documents, and ensure the latest regulatory content.

[0087] Abnormal marking and report generation: Automatically mark non-compliant or questionable clauses and generate detailed abnormal reports, including the specific content of the non-compliant clauses, legal basis and corresponding clause content.

[0088] Historical data call: When the applicability of regulations involves historical clauses or expired clauses, the system can call historical regulatory data for comparison and indicate whether the clauses referenced in the document have expired.

[0089] 5. Application of intelligent review model

[0090] Machine learning and rule engine: Build an intelligent review model that combines machine learning and rule engine. Based on the content in the bidding documents, key terms and fields are automatically extracted and then compared with the standard content in the regulatory database.

[0091] Natural Language Understanding (NLU): Use natural language understanding technology to parse the language of the bidding documents, identify the semantics behind the terms, and match them with the semantics of the regulatory terms.

[0092] Knowledge graph-assisted review: Build a knowledge graph of bidding regulations to visualize the relevance and hierarchical structure of regulations, help identify the applicability and priority relationships between different regulatory provisions, and improve review efficiency.

[0093] 6. User feedback and self-learning

[0094] Manual review and feedback mechanism: After the automatic review results are generated, a manual review and feedback mechanism can be added to verify the accuracy of the model review results.

[0095] Model self-learning: Feedback information is added to the model’s self-learning module to continuously optimize retrieval, comparison, and calling rules to improve the accuracy and audit efficiency of the system.

[0096] Among them, compliance review is to automatically extract key information from the first draft of the bidding documents through NLP technology (including project name, project number, project budget, funding source, qualification requirements, past performance, financial status, personnel requirements, technical specifications, quality standards, acceptance standards, environmental standards, contract period, performance bond, payment terms, breach of contract terms, risk sharing, bidding method, announcement period, bid deadline, bid opening process, bid evaluation process, scoring criteria, price scoring weight, technical scoring details, bonus items, statutory compliance clauses, policy compliance, safety regulations, labor protection, effective date, expiration date, update requirements, intellectual property ownership, confidentiality obligations, risk response measures, force majeure clauses, environmental compliance, production safety requirements, technical certification and other key information), and compare with the data in the online database and local database to check whether there are any clauses that violate the relevant bidding regulations. The specific steps are as follows:

[0097] 1. Data normalization and preprocessing

[0098] Text normalization: Standardize the extracted key information and reference data in the database to eliminate format and expression differences, including unified fonts, uppercase and lowercase letters, and number formats.

[0099] Synonym processing: Create a synonym dictionary for different expressions of the same content (such as "performance bond" and "performance deposit") to facilitate subsequent accurate comparison.

[0100] 2. Logical design of item-by-item comparison

[0101] Comparison of basic project information: project name, number, budget, source of funds and other information are directly and accurately matched with the filing information of the corresponding project in the database. Ensure that the basic information of the project is consistent with the approved or filed information.

[0102] Comparison of qualifications and requirements: Compare the qualification requirements, past performance, financial status and other contents with the compliance requirements in laws and regulations and industry standards to verify whether there are unreasonable threshold conditions or exclusive clauses. For restrictive clauses such as qualifications, the rationality can be verified by referring to the qualification requirements of similar projects in the past.

[0103] Comparison of technical and quality requirements: Compare technical specifications, quality standards, acceptance criteria and other information with the industry standards in the database to ensure that no unreasonable technical requirements are set. If the industry standard in the database stipulates that the standard for a specific device is "XX parameter", check whether the document uses clauses that meet the parameter.

[0104] 3. Comparison of contract and performance conditions

[0105] Comparison of term and amount: Compare the contract term, performance bond and other conditions with the data of similar projects or the prescribed range in the regulations. For example, the performance bond amount should be controlled within a reasonable percentage of the total project amount, and it will be considered abnormal if it exceeds this range.

[0106] Comparison of payment terms and breach of contract terms: Compare payment terms and breach of contract terms to see if they comply with contract law and industry regulations, and ensure that payment arrangements and breach of contract handling comply with the general standards of relevant laws.

[0107] Comparison of risk allocation and force majeure clauses: Check the risk clauses and force majeure clauses in the documents to ensure compliance with the legal standards in the database. Compare similar clauses in past projects to see if there are any deviations.

[0108] 4. Comparison of processes and deadlines

[0109] Comparison of bidding methods and announcement periods: Compare the bidding methods, announcement periods and other contents with the regulations to ensure that the announcement period is not less than the minimum period required by law. For certain project types (such as open bidding), the announcement period should meet the statutory standards.

[0110] Deadline and process comparison: Check whether the bid deadline and bid opening process are in compliance with process specifications. Ensure that the bid opening time meets the requirements of fairness and openness, such as not allowing early bid opening or setting a non-public bid evaluation process.

[0111] 5. Comparison of scoring criteria and bid evaluation rules

[0112] Comparison of scoring weights: Compare the price weights and technical scoring details in the scoring criteria with the standard database to ensure that the weights are set within a reasonable range. For example, the price weight is usually between 30% and 60%, and an abnormality is indicated if it exceeds this range.

[0113] Reasonableness comparison of bonus items: Compare the bonus items in the scoring with the laws and policies to ensure that the bonus items do not favor specific bidders and meet the requirements of fair competition.

[0114] 6. Comparison of compliance and policy requirements

[0115] Comparison of statutory compliance clauses: Compare the statutory compliance clauses (such as fair competition statements and confidentiality requirements) in the documents one by one to see if they are complete, and refer to the standard contract clauses in the database to ensure that all necessary clauses are complete.

[0116] Policy compliance comparison: Check whether the document complies with the latest policies, such as local procurement preferences, environmental protection requirements, etc., and compare it item by item with the content in the policy database to ensure compliance with the latest policy guidelines.

[0117] 7. Comparison of effective and expiration dates

[0118] Compare the effective and expiration dates: Compare the effective and expiration dates in the document with the project approval time and industry specifications to ensure that the time range meets regulatory requirements. For example, if the regulations require that the project must be completed within six months after the contract is signed, if the document stipulates that it takes more than six months, an abnormality will be prompted.

[0119] 8. Comparison of intellectual property, confidentiality obligations and risk management clauses

[0120] Comparison of intellectual property ownership clauses: Check whether the intellectual property ownership is clear in the document and compare it with the standard clauses in the database to ensure that the intellectual property ownership is clear.

[0121] Confidentiality obligation comparison: Compare whether the confidentiality obligation clauses meet the standards to ensure that sensitive bidding information is not leaked.

[0122] Risk clause comparison: Compare risk response measures and force majeure clauses with standard clauses to ensure compliance with industry standards. If there are missing items or unclear clause descriptions, revision suggestions will be provided.

[0123] 9. Comparison of industry-specific requirements

[0124] Comparison of environmental protection and safety requirements: Compare special requirements such as environmental protection and safe production to ensure that the environmental protection, energy conservation and other clauses described in the document comply with national and local environmental protection and safety policies.

[0125] Technical certification comparison: The technical certification requirements for specific products are compared with the certification requirements of similar projects in the database to ensure compliance with industry certification requirements.

[0126] 10. Abnormal records and prompts

[0127] Abnormal clause marking: During the comparison process, the abnormal results of each comparison are recorded and abnormal clauses are highlighted in the final comparison report.

[0128] Generate detailed comparison report: Generate an audit report of the comparison results between the file and the database, list abnormal information and correction suggestions, and help the tender document compilers make modifications.

[0129] Among them, other abnormal audits use machine learning algorithms to further conduct in-depth analysis of the bidding documents, including comparing other project bidding documents in online and local databases to check whether there are abnormalities, inconsistencies, contract clause loopholes, management model defects, etc. The specific steps are as follows:

[0130] 1. Text analysis and comparison

[0131] Natural Language Processing (NLP): Use NLP technology to perform structured analysis on bidding documents, extract key contract terms, conditions, terms, and identify key information. Common methods include:

[0132] Keyword extraction: Extract key contract terms, amounts, deadlines, and other information.

[0133] Semantic analysis: Use word segmentation, part-of-speech tagging, and syntactic analysis to identify the semantics of each clause and determine the logical relationship between the clauses.

[0134] Inconsistency detection: By automatically comparing the content and expressions of different clauses, detect whether there are any logical or expression inconsistencies. For example, whether the time and amount defined in a clause are consistent with other clauses.

[0135] Text similarity analysis: Use methods such as cosine similarity or sentence vectors to determine whether different parts of the content are similar and check whether there is any content redundancy or repeated definitions.

[0136] 2. Logical rules and conditional checks

[0137] Rule-based logic check: Write a rule engine to define the logical relationship and conditional rules of the bidding documents. For example, "payment terms" should correspond to "delivery terms" and there should be no contradictory descriptions.

[0138] Automated condition verification: Automatically verify the consistency of the document through condition rules. For example, whether the relationship between the payment ratio in the contract and the project progress is reasonable and whether it conforms to the general rules of engineering construction.

[0139] 3. Anomaly and vulnerability detection

[0140] Vulnerability identification model: Based on common problems and vulnerabilities in historical contracts or bidding documents, machine learning models are trained to detect anomalies in the documents. Samples of past contract issues can be used to build classification or anomaly detection models to identify potential vulnerabilities.

[0141] Rule base and compliance verification: Establish a rule base based on industry standards or regulations to automatically detect whether the document complies with these standards or regulations. For example, conduct compliance verification on procurement processes, confidentiality clauses in contract terms, intellectual property clauses, etc.

[0142] 4. Analysis of management model defects

[0143] Management model data mining: Identify potential defects by analyzing the management model in project management methods and contracts. For example, whether the division of labor in project implementation is reasonable, whether there is a clear division of responsibilities, and whether the management process is clear.

[0144] Process analysis and simulation: Use process mining and process simulation technology to simulate the management model involved in the contract terms and analyze possible management bottlenecks, unreasonable resource allocation and other issues.

[0145] Expert rule base and knowledge graph: Establish a knowledge graph of the management model and combine it with the rule base of industry experts to detect whether the management model in the contract is reasonable. For example, whether there are necessary risk control measures and emergency plans.

[0146] 5. Intelligent inspection tools and report generation

[0147] Intelligent inspection system: Integrate the above analysis modules and develop automated inspection tools to conduct multi-dimensional intelligent inspections of bidding documents on a regular basis.

[0148] Report generation: The inspection results are automatically generated into a report, marking potential inconsistencies, contract loopholes and management model defects, and providing improvement suggestions to help assessors quickly identify risks.

[0149] 6. Manual review

[0150] Finally, combined with the system analysis results, manual review and secondary evaluation are carried out to ensure that abnormal issues not captured by the model are fully identified. This step is particularly suitable for bidding documents for complex or sensitive projects.

[0151] Through the above technical implementation, problems in the bidding documents can be analyzed and checked efficiently and comprehensively to ensure the rationality and compliance of the contract terms and management model.

[0152] Among them, the review opinions on the bidding documents include: ① whether there are any violations of relevant bidding regulations; ② at the same time, horizontally compare the bidding documents of previous projects in the local database and prompt abnormal bidding documents; ③ whether there are any contradictions, loopholes in contract terms, defects in management models, etc.;

[0153] Among them, abnormal bidding documents specifically refer to:

[0154] 1. The content is incomplete or missing, mainly due to missing clauses (the bidding documents lack key clauses that must be included in laws, regulations, industry standards or practices, such as confidentiality clauses, contract performance clauses, acceptance criteria, etc.), incomplete information (the documents do not contain basic project information, such as project name, budget amount and source of funds, or the bidding content is not described in detail as required) and missing attachments (the attachments required by the bidding documents, such as drawings, technical parameters and bidding instructions, are not attached, or the appendix documents are incomplete).

[0155] 2. The terms or conditions are set unreasonably, mainly because the threshold conditions are too high or too low (the requirements for the bidder's qualifications, performance, etc. are set too high or too low, which is obviously inconsistent with industry practices or regulations, and may be biased or exclusive), the technical parameters are too specific (the technical specifications or product requirements are too detailed, which may restrict potential bidders, and are suspected of being tailor-made for specific manufacturers, violating the principles of openness, fairness and impartiality) and the scoring criteria are unreasonable (the scoring criteria lack objectivity, which may be unfavorable to some bidders, or the scoring items are unbalanced, such as the price weight is too high or too low).

[0156] 3. Abnormal time nodes, mainly due to insufficient announcement time (failure to reserve sufficient announcement period or bid preparation period in accordance with regulatory requirements, such as the 20-day announcement period required by law) and conflicts in bidding schedules (opening time, deadline, clarification time, etc. do not conform to reasonable procedures or conflict with time nodes in other bidding documents).

[0157] 4. The bidding amount and budget are abnormal, mainly due to unreasonable budget (the project budget is too high or too low compared with the market price, especially in equipment procurement, engineering cost, etc., there may be abnormal quotations) and unclear sources of funds (the source of funds is not clear or the source of funds does not meet the requirements, which may affect the project's financial security and contract performance security).

[0158] 5. Legal and compliance issues, mainly including the citation of outdated regulations (the legal clauses or standards cited in the document have expired or become invalid, which may make subsequent contract clauses unenforceable), non-compliance with new regulations (failure to update to the current effective regulatory requirements in a timely manner, or failure to include the latest compliance clauses) and priority conflicts (there are priority conflicts between the cited legal clauses, such as conflicts between local regulations and laws).

[0159] 6. Historical data anomalies are mainly due to inconsistency with past project data (document contents are significantly different in similar projects, such as large differences in the settings of bills of quantities, technical parameters, bidder qualification requirements, etc., which may imply biased or unfair settings) and significant differences in contract terms (contract terms vary greatly in similar projects, such as performance bonds, breach of contract liability, etc., and it is necessary to judge whether there is bias or special tendency).

[0160] 7. Price and cost anomalies, mainly abnormally low or high quotations (the system compares the quotation levels of similar projects. If the price deviates significantly from the market level, there may be suspicion of vicious competition or profit transfer) and abnormal setting of the upper or lower limit of the bid (the setting of the maximum or minimum price is unreasonable, which may lead to reduced competitiveness or benefit specific bidders).

[0161] 8. Violations of the bidding procedures mainly include non-compliant bidding methods (failure to select the appropriate bidding method in accordance with regulations, illegal selection of single-source procurement, etc.), improper prequalification (the prequalification process is not transparent, or there are situations where qualified bidders are excluded), and irregularities in the publication of bidding documents (failure to publish bidding documents through the prescribed announcement platform may affect openness and transparency).

[0162] 9. Related party or conflict of interest issues, mainly potential conflicts of interest (stakeholders may be involved in the documents, which may lead to unfair competition, such as the participation of the bidder’s related parties or stakeholders) and undisclosed related relationships (if the bidder has an undisclosed related relationship with the tenderer or the bid evaluation party, it may violate the fairness requirement).

[0163] 10. Recurrence of problems in historical projects, mainly the recurrence of historical anomalies (the system finds duplicate abnormal content in historical projects, such as setting the same unreasonable clauses, suggesting that there may be systemic problems) and recurrence of quality and performance issues (quality problems or performance issues that occurred in past projects were not mentioned or avoided in the bidding documents, and the contract terms were not optimized for historical problems).

[0164] The above anomalies help the system to identify potential problems in a targeted manner during the document review process and improve the standardization and transparency of the bidding process.

[0165] Step 4: Based on the audit opinions described in step 3, the system user unit manually confirms the audit opinions one by one and generates the final bidding document. The manual confirmation can be to accept the audit opinions and make modifications, to reject the audit opinions and not make modifications, or to make adjustments based on the audit opinions.

[0166] Among them, it can also include step 5: the local database in step 2 automatically enters the final version of the bidding document generated in step 4, and records the results of manual confirmation of the review opinion by the system user in step 4, and trains the machine to learn the style of the user unit in compiling the bidding document by continuously recording the results of the manual confirmation of the user unit, and optimizes the review opinions proposed in step 3 through the learning model.

[0167] Among them, optimization can be achieved by setting a threshold for the confirmation results of the system users for the same audit opinion proposed in step 3. When the number of records is greater than the threshold, the system will automatically block such audit opinions for the system users who refuse to adopt the modification opinions, so as to avoid repeated reminders and reduce unnecessary workload.

[0168] The system records the habits and preferences of the user unit when editing the bidding documents to better adapt to future work needs. The specific steps are as follows:

[0169] 1. Data Collection Module

[0170] 1.1 Definition of User Behavior Events

[0171] Define key user behavior events during the editing process, such as:

[0172] Content operations: such as adding, deleting, and editing clause content.

[0173] Format operations: such as setting font size, paragraph spacing, alignment, etc.

[0174] Template operations: such as selecting commonly used templates and quoting specific content fragments.

[0175] File preservation and version management: record each save, undo, restore and other operations.

[0176] 1.2 Front-end behavior collection

[0177] Embed the behavior collection code in the front-end page of the editor (such as a Web editor based on React, Vue, etc.).

[0178] Use JavaScript or front-end monitoring tools (such as Google Analytics or a custom event tracking system) to capture user behavior in real time and record operational data.

[0179] Every time a user performs a key operation, a log record is generated and sent to the backend storage module through the API interface.

[0180] 1.3 Data upload and filtering

[0181] Upload the captured data to the backend in real time or in batches, filter out irrelevant data, and retain only valid user operation records.

[0182] De-duplicate and verify the data to ensure its integrity and accuracy.

[0183] 2. Data storage module

[0184] 2.1 Database Design

[0185] Use a relational database (such as MySQL, PostgreSQL) or a NoSQL database (such as MongoDB) to store user behavior data. Design the following data tables:

[0186] User table: records basic user information (such as ID, role, department, etc.).

[0187] Operation log table: records detailed information of each user editing operation (such as operation type, specific content, timestamp, etc.).

[0188] Template and clause table: records commonly used templates and clause information in the system to facilitate subsequent correlation analysis.

[0189] 2.2 Data Archiving and Compression

[0190] Since the behavior data may be very large, it is recommended to set up a regular archiving strategy to compress and archive historical data to ensure database performance.

[0191] 3. Data processing and analysis module

[0192] 3.1 Data cleaning and preprocessing

[0193] Clean the collected behavioral data to remove invalid operations and noise data.

[0194] Unify data formats to provide high-quality data sources for subsequent analysis.

[0195] 3.2 Behavior Analysis Model

[0196] Use machine learning or data mining techniques to analyze users' editing behavior and build preference models.

[0197] Frequent pattern mining: Use the Apriori algorithm to mine users' frequently used terms and editing habits and generate frequent operation sequences.

[0198] Cluster analysis: Cluster user operations and group users with similar preferences into one category in order to provide group recommendations.

[0199] Collaborative filtering recommendation: Based on the collaborative filtering algorithm, the behavioral similarities of different users are analyzed to recommend editing modes and content commonly used by others to users.

[0200] 3.3 Preference model training and updating

[0201] Use the collected data to train a preference model (such as editing mode based on time series) and update the model regularly to reflect the user's latest preferences.

[0202] Deep learning models (such as LSTM or Transformer) can be introduced to analyze the time series data of user operations and predict users' future editing needs.

[0203] 4. Recommendation and personalization module

[0204] 4.1 Personalized interface configuration

[0205] Dynamically adjust the editor interface based on user preferences. For example, load the user's most commonly used clause templates and set specific formatting by default.

[0206] Provides a "quick insert" function to set users' frequently used terms as shortcut buttons for quick and easy access.

[0207] 4.2 Real-time recommendations and intelligent prompts

[0208] When users enter content, relevant terms or formats are recommended in real time based on historical behavior. For example, when entering bidding content, matching term templates are automatically recommended.

[0209] Provide intelligent error correction function: If the system detects habitual errors of the user (such as frequent typos, inconsistent formats), it will give real-time prompts.

[0210] 4.3 Personalized reminders

[0211] Provide intelligent reminder function to remind users to supplement key terms or recommend corresponding terms combinations based on their preferences.

[0212] 5. Feedback and optimization module

[0213] 5.1 User Feedback Collection

[0214] Embed feedback functionality in the editor to allow users to rate system recommendations and personalization settings (e.g., satisfactory, irrelevant, try again later).

[0215] 5.2 Model Update and Iteration

[0216] Regularly update the recommendation algorithm based on user feedback data to ensure that the system can more accurately reflect user preferences.

[0217] Optimize model parameters based on feedback and usage data to gradually improve the accuracy and applicability of recommendations.

[0218] 5.3 Data monitoring and continuous optimization

[0219] By monitoring the way users use the system through data, we can optimize recommendation and personalization functions, and enhance the system's responsiveness and user experience.

[0220] Generate preference analysis reports to help management understand users' editing preferences and further optimize workflows.

[0221] 6. Security and Privacy Protection

[0222] Data encryption: Encrypt user data for storage to ensure the security of sensitive information.

[0223] Permission management: Through role and permission control, ensure that different users can only access the data they need.

[0224] Privacy protection and compliance: Design data collection and storage processes in accordance with relevant data privacy regulations (such as GDPR) to ensure compliance.

[0225] Regarding the selection of key technologies for the present invention, the front-end behavior collection can use JavaScript SDK, WebAnalytics tools (such as Google Analytics, Mixpanel), the database can use MySQL, MongoDB or PostgreSQL, data analysis and modeling can use Python (Pandas, NumPy, Scikit-Learn), Spark or TensorFlow / Keras, the recommendation engine can use collaborative filtering algorithm, content recommendation algorithm or time series-based prediction model, and security can use data encryption and user authority management (authority management system based on OAuth or RBAC).

[0226] An automatic tender document review system based on artificial intelligence, including the following modules:

[0227] Module 1: Database module, which stores bidding-related regulations and bidding documents, and sets keywords for them to facilitate matching and comparison by the bidding document review module.

[0228] Module 2: File upload module, which uploads the draft of the bidding document prepared by the system user unit to the bidding document review module after authorization; and uploads the finalized bidding document confirmed by the system user unit to the local database.

[0229] Module 3: Tender document review module, which conducts format review, compliance review and other abnormal review on the draft tender documents uploaded by the system users.

[0230] Module 4: Modification suggestion module, which generates review opinions on the bidding documents reviewed by the bidding document review module. The review opinions include the issues found in the review and the specific modification opinions and reasons for each issue.

[0231] Module 5: The bidding document generation module automatically generates the final version of the bidding document based on the system user's processing results of the specific modification suggestions (such as acceptance, rejection, adjustment or partial adoption).

[0232] Module 6: User interaction interface, which facilitates system users to operate the system.

[0233] Module 7: Learning module, which is also the user style recording module, automatically compares the review opinions proposed by the system with the final version of the bidding documents, and records the opinions proposed by the system but not accepted by the system user. When the number of records of the same situation exceeds the threshold set by the system user, the system will automatically block such review opinions for the system user to avoid repeated reminders and reduce unnecessary workload; record the results of the system user's manual confirmation of the review opinions, and use these feedback data to train the machine learning model to gradually improve the accuracy and efficiency of the output of the bidding document review module.

[0234] Among them, the module 1 database module includes an online database and a local database. The online database automatically searches and downloads bidding-related regulations and bidding document templates through the Internet from time to time; the local database is manually entered by the system user unit into the bidding-related regulations and project bidding document templates of its specific industry.

[0235] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. An automatic review method for bidding documents based on artificial intelligence, characterized in that: The following steps are involved: Step 1: Set up an online database to automatically search and download bidding-related regulations and bidding document templates from the Internet from time to time, and mark them with keywords; Step 2: Set up a local database, and have the system user manually input the bidding regulations and project bidding document templates of their specific industry, and mark them with keywords; Step 3: The system user uploads the draft of the bidding document that needs to be reviewed. The bidding document review module conducts format review, compliance review and other abnormal reviews on it, and then generates a review opinion; Step 4: Based on the review opinions described in Step 3, the system user will manually confirm the review opinions and generate the final bidding documents.

2. According to claim 1, the method for automatic review of bidding documents based on artificial intelligence is characterized in that: It also includes step 5: automatically entering the final draft of the bidding document generated in step 4 into the local database, and recording the results of manual confirmation of the review opinions by the system user in step 4, and training the machine to learn the style of the bidding document prepared by the user by continuously recording the results of the manual confirmation by the user, and optimizing the review opinions proposed in step 3 through the learning model.

3. The method for automatic review of bidding documents based on artificial intelligence according to claim 1 is characterized in that: The bidding-related regulations mentioned in step 1 include national laws, regulations, rules, normative documents, national standards, and industry standards in the bidding field. The bidding document template is a standard bidding document or bidding document sample text issued by the state or local area, and a bidding document for projects of various industries, regions, and types.

4. The method for automatic review of bidding documents based on artificial intelligence according to claim 1 is characterized in that: The bidding-related regulations for specific industries described in step 2 include national laws, regulations, rules, normative documents, national standards, and industry standards in the bidding field. The project bidding document template includes standard bidding documents or bidding document sample texts issued by the state or local areas, and project bidding documents for various industries, regions, and types.

5. The method for automatic review of bidding documents based on artificial intelligence according to claim 1 is characterized in that: The keywords marked in step 1 and step 2 include the title of the bidding-related regulations, effective date, expiration date, issuing unit, publication date, document category, document version, nature of the bidding documents, applicable area, regional exclusivity, applicable industry, application scenario, project type, bidding stage, scope of application, contract type, legal level, legal priority level, applicable priority, compliance requirements, key terms, penalty measures, applicable subjects, document applicable conditions, additional terms, and implementation details.

6. The method for automatic review of bidding documents based on artificial intelligence according to claim 1 is characterized in that: The compliance review and other abnormal review described in step 3 are reviewed by searching, comparing, and retrieving data from online databases and local databases.

7. An automatic tender document review system based on artificial intelligence, characterized in that: Includes the following modules: Module 1: Database module, which stores bidding-related regulations and bidding documents, and sets keywords for them to facilitate matching and comparison by the review module. Module 2: File upload module, which uploads the draft of the bidding document prepared by the system user unit to the bidding document review module after authorization; and uploads the finalized bidding document confirmed by the system user unit to the local database. Module 3: Tender document review module, which conducts format review, compliance review and other abnormal review on the draft tender documents uploaded by the system users. Module 4: Modification suggestion module, which generates review opinions on the bidding documents reviewed by the bidding document review module. The review opinions include the issues found in the review and the specific modification opinions and reasons for each issue. Module 5: The bidding document generation module automatically generates the final version of the bidding document based on the processing results of the system user's specific modification suggestions. Module 6: User interaction interface, which facilitates system users to operate the system. Module 7: Learning module, which is also the user style recording module, automatically compares the review opinions proposed by the system with the final version of the bidding documents, and records the opinions proposed by the system but not accepted by the system user. When the number of records of the same situation exceeds the threshold set by the system user, the system will automatically block such review opinions for the system user to avoid repeated reminders and reduce unnecessary workload; record the results of the system user's manual confirmation of the review opinions, and use these feedback data to train the machine learning model to gradually improve the accuracy and efficiency of the output of the bidding document review module.

8. The automatic tender document review system based on artificial intelligence according to claim 7 is characterized by: The module 1 database module includes an online database and a local database. The online database automatically searches and downloads bidding-related regulations and bidding document templates from time to time through the Internet; the local database is manually entered by the system user unit into the bidding-related regulations and project bidding document templates of its specific industry.

Images