Bidding document automatic writing method and system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based bidding document automatic writing method and system, and the method comprises the steps: recognizing a project bidding document through a bidding document understanding module, extracting the requirements of the bidding document from the project bidding document, automatically recognizing materials according to the extracted requirements of the bidding document, and continuously updating a material database; according to the method, related materials are automatically obtained from a material database, a first to-be-processed bidding file is automatically generated according to the requirements of the recognized bidding file, a second to-be-processed bidding file is generated after the first to-be-processed bidding file is automatically compared, optimized and modified by a learning module, and a final version of bidding file is generated after manual confirmation by a system using unit. The method automatically identifies the bid invitation file requirements, continuously updates the material database, automatically matches the bid invitation file requirements, automatically generates the bid invitation file, and is convenient to use, short in time consumption, efficient and accurate.

Description

Technical Field

[0001] The present invention belongs to the field of public resource transaction and office automation technology, and specifically relates to an artificial intelligence-based bidding document automatic writing method and system. Background Art

[0002] At present, bidding and procurement has become a common way of concluding contracts and has been widely used in various projects in various industries and fields. In the bidding and procurement process, a large amount of bidding document preparation work needs to be completed. On the one hand, it is limited by the limitation of manual efficiency. If the bidder has a large number of projects to bid in a short period of time, it may not be able to complete the bidding task; on the other hand, statistics show that the error rate of manual operation is much higher than that of computer operation. Even experienced bidders may have their bidding documents rejected due to a small mistake, resulting in the whole team's efforts for many days going to waste. In addition, if the bidding document preparation personnel change and there are problems with the handover of materials, the relevant supporting materials will be lost, and the bidding document supporting materials need to be collected again.

[0003] With the development of Internet technology, computer technology, electronic bidding and other technologies, it has become possible for the same bidder to bid for projects in multiple locations simultaneously through the Internet. This requires a change in the traditional way of preparing bidding documents. It is of great significance to invent an automatic bidding document writing method and system based on artificial intelligence that is simple, efficient and accurate. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for automatically writing bidding documents based on artificial intelligence, so as to improve the efficiency and accuracy of bidding document writing and solve the shortcomings of the prior art.

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

[0006] An automatic bidding document writing method based on artificial intelligence comprises the following steps:

[0007] Step 1: Establish a material database and enter the bidding information of the system user, including basic information of the enterprise and common certification materials for bidding;

[0008] Step 2: Establish a project bidding document understanding module to identify the project bidding documents and extract the requirements of the bidding documents, including the bidding document format requirements, bidding scope and various response clause requirements in the bid evaluation method;

[0009] Step 3: According to the requirements of the extracted bidding documents, automatically identify whether the materials in the material database are complete and valid. For materials that are incomplete or some of which have expired, generate a list to be updated and update the material database according to the requirements of the list;

[0010] Step 4: After the materials in the material database are complete and valid, relevant materials are automatically obtained from the material database, and the first bidding document to be processed is automatically generated according to the requirements of the identified bidding documents;

[0011] Step 5: Based on the final version of the bidding documents of the user's previous bidding projects, automatically compare, optimize and modify the first bidding document to be processed, and generate the second bidding document to be processed in the specified format. If the user uses this system for bidding for the first time and there is no final version of the bidding document of the user's previous bidding project in the material database, this step will not be performed;

[0012] Step 6: The system user manually confirms the second bid document to be processed and generates the final bid document for this project; if the user is using this system to bid for the first time, this step is for the system user to manually confirm the first bid document to be processed and generate the final bid document for this project.

[0013] At the same time, the final version of the bidding document for this project generated in step 6 is automatically uploaded to the material database, and while being stored in the material database, it is marked according to key information such as project industry (such as highway, water transport, railway, building construction, etc.), bidding type (such as construction, supervision, design, auditing, cost consulting, etc.), and bidding requirements, so that data can be quickly retrieved when the bidding documents for other projects are prepared later (that is, in step 5).

[0014] Furthermore, the basic information of the enterprise and the commonly used certification materials for bidding entered by the system user in step 1 include the business license, qualification certificate, award certification materials, financial audit statements, tax payment certificates, past performance contract agreements and their supporting materials, personnel certificates, personnel social security payment certificates, personnel performance contract agreements, commonly used technical plans for various types of projects and other supporting materials.

[0015] Furthermore, in step 3, the material database is updated. The system automatically searches and downloads materials to update the material database based on the list to be updated, and uses crawler technology to download and automatically update the material library, or starts a scheduled task to regularly obtain materials from a specific website.

[0016] Furthermore, in the step 3, the material database is updated, and the system user unit can be reminded to manually enter the updated materials in the material database in time according to the list to be updated through system messages, system administrators receiving emails, text messages, etc.

[0017] Furthermore, the list to be updated generated in step 3 may also be supplemented or deleted manually by the user unit.

[0018] An automatic bidding document writing system based on artificial intelligence includes the following modules:

[0019] Module 1: Database module, used to store materials and other data.

[0020] The materials include basic information of the enterprise, common certification materials for bidding, and various versions of the generated bidding documents. Keywords are set for the materials to facilitate identification, matching, comparison, and retrieval in the later stage.

[0021] Module 2: Tender document understanding module, which identifies tender documents and extracts the requirements of tender documents.

[0022] Module 3: Bidding document preparation module, used to generate bidding documents. Bidding documents are prepared according to the requirements of the identified and extracted bidding documents.

[0023] Module 4: Data upload module, uploads the data of this project and stores it in the material database.

[0024] Module 5: User interaction interface, which facilitates system users to operate the system.

[0025] Module 6: Learning module, which is also the user style recording module, automatically compares, optimizes and modifies the first bid document to be processed with the final version of the bid document of the previous bidding project in the material database, records the processing opinions of the system user unit, and records the results of the system user unit's manual confirmation of the bid document compiled by the system, and uses these feedback data to continuously train the machine learning model to gradually improve the accuracy and efficiency of the output of the bid document preparation module.

[0026] The beneficial effects of the present invention are as follows: compared with the prior art, the present invention establishes a material database based on artificial intelligence to store the commonly used bidding materials of enterprises in advance, establishes a project bidding document understanding module, automatically identifies the requirements of the bidding documents, and continuously updates the material database according to the requirements of the bidding documents, establishes a learning module, records the usage habits of the user units, and further optimizes and modifies the generated bidding documents according to the final version of the bidding documents of the previous bidding projects, so as to generate bidding documents that are easy to use and time-saving, and the more times the user units use them, the higher the efficiency and accuracy of producing bidding documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the present invention;

[0028] Figure 2 It is a module schematic diagram of the present invention. DETAILED DESCRIPTION

[0029] 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.

[0030] Example 1: Reference Figure 1 This embodiment adopts an automatic bidding document writing method based on artificial intelligence, including the following steps:

[0031] Step 1: Enter the basic information of the system user and the commonly used certification materials for bidding into the material library, including but not limited to the business license, qualification certificate, award certification materials, financial audit statements, tax payment certificates, past performance contract agreements and their certification materials, personnel certificates, personnel social security payment certificates, personnel performance contract agreements, commonly used technical solutions for various types of projects and other certification materials and other commonly used bidding elements. The system builds a file server to store business licenses, qualification certificates, award certification materials, financial audit statements, tax payment certificates, past performance contract agreements and their certification materials, personnel certificates, personnel social security payment certificates, personnel performance contract agreements, commonly used technical solutions for various types of projects and other certification materials, etc. Image-type (jpg, png, etc. formats) and file-type (word, pdf, etc. formats) data.

[0032] Among them, the file server can use the open source file service software Minio to independently deploy a local file server. The required business license, qualification certificate, award-winning proof materials, financial audit reports and other pictures and files are first uploaded to the file server, and the unique ID returned by the file server is obtained. This unique ID corresponds to a unique file, and then the unique ID is stored in the relational database Postgresql. A table dedicated to storing file meta-information is established in the Postgresql library. The fields roughly include the above-mentioned ID and file classification: for example, the unit business license, qualification certificate, award-winning proof materials, financial audit reports, tax payment certificates, past performance contract agreements and their supporting materials, personnel certificates, personnel social security payment certificates, personnel performance contract agreements, commonly used technical plans for various types of projects and other supporting materials, as well as the corresponding validity period, owner, upload timestamp, etc. of the file; it has a detailed field description to facilitate the rapid and accurate retrieval when the corresponding file needs to be accurately located and obtained in the future.

[0033] Step 2: Establish a bidding document understanding module to identify project bidding documents and extract the requirements of the bidding documents, including the bidding document format requirements, bidding scope and various response clause requirements in the bid evaluation method.

[0034] First, a model is trained to identify and classify different parts of the bidding documents, such as the bidding document format requirements, the bidding scope, and the various response clause requirements in the bid evaluation method. A large number of samples with correct labels are provided to train the model. Then a set of rules matching specific keywords or phrases are designed to determine whether the response clause requirements are met. The rules can be written based on business knowledge or generated by a machine learning model.

[0035] Secondly, use NLP technology to parse the text and understand the sentence structure and meaning. Named Entity Recognition (NER) can be used to identify and extract specific types of entities such as company names, names, dates, etc. Dependency analysis can help find the connections between entities. Perform in-depth analysis of the text to find hidden relationships and patterns. It can be used to discover potential risk points or opportunities. The extracted information is displayed in charts or other forms to facilitate understanding and decision-making.

[0036] Step 3: According to the requirements of the extracted project bidding documents, automatically identify whether the materials in the material database are complete and valid, understand the requirements and keywords of the bidding documents according to the above artificial intelligence algorithm, and search whether the materials in the data material library are completely matched. For incomplete materials or some materials that have expired, generate a list of materials that need to be updated in the material database. According to the list, on the one hand, conduct an online search, use crawler technology to download and automatically update the material library, or start a scheduled task to obtain materials regularly from a specific website. On the other hand, remind system users to manually update the material library materials in a timely manner through system messages, system administrators receiving emails and text messages, etc.

[0037] Step 4: After the materials in the material database are complete and valid, automatically obtain relevant materials from the material database, and automatically generate the first tender document to be processed according to the requirements of the tender document identified in step 2. The specific implementation method is as follows:

[0038] 1. Identify and generate the style and structure of the first tender document to be processed

[0039] File parsing: Use Apache POI to read the format of the tender documents, including paragraph style, font, font size, alignment, title structure, etc. The parsed structural information is used to match the style of the generated files.

[0040] Title structure recognition: By reading the structure of the title, table of contents, chapter title, etc., the formatting rules of each part, such as the font, size, and alignment of the title, are obtained.

[0041] Paragraph and list styles: Detect body paragraphs, indents, list styles, etc., laying the foundation for maintaining format consistency for subsequent content filling.

[0042] Style rule extraction: Structuring the acquired style information, creating a style rule table (such as the style of the title, the format of the text, etc.) in order to unify the style of the generated files.

[0043] Style rule templateization: Convert the acquired style definitions into reusable templates, and establish template structures according to different chapters to facilitate subsequent content filling.

[0044] 2. Get relevant materials from the material library

[0045] Material keyword extraction: Analyze the content structure of the bidding documents, extract keywords and determine content requirements. Keywords may include project budget, qualification requirements, personnel requirements, technical parameters, etc.

[0046] Database search: Use PostgreSQL to query the material library and find content that matches keywords. You can use project name, chapter name, keywords, etc. as search conditions to filter out the best matching materials from the material library.

[0047] Content priority: If there are multiple matching materials in the database, the latest version or the material that meets the specific requirements of the bidding project will be given priority.

[0048] Material format matching: After retrieving the material in PostgreSQL, ensure that the content and format of the material conform to the format requirements of the target file (such as font, paragraph alignment, chart layout, etc.).

[0049] 3. Content filling and format adjustment

[0050] Content insertion template: Fill the relevant material content obtained from the material library in each chapter and position of the target document to keep it consistent with the format requirements of the bidding document. You can use the following steps:

[0051] Template mapping: Fill the retrieved content into the corresponding sections of the bidding document according to the style rule template, such as basic project information, technical specifications, qualification requirements, etc.

[0052] Paragraph and title formatting: Use Apache POI to set the style of each part (such as font, color, alignment, etc.) to ensure that the style of the filled content is consistent with the original file format.

[0053] Automatic format adjustment: For materials with long paragraphs or large tables, content scaling, paragraph merging, and font size and line spacing can be adjusted to ensure that the file is neat and readable.

[0054] Content adaptation adjustment: If the material content is too long, the font size and paragraph indentation will be automatically adjusted to prevent the text from exceeding the page range and ensure a beautiful page layout.

[0055] Table and graphic formatting: If you fill in a table or graphic material, the cell width, alignment, border line, etc. will be automatically set to ensure that the table and graphic are displayed beautifully in the document.

[0056] 4. Generation and format check of the first pending bidding document

[0057] Content integrity check: After all content is filled in, the first bid document to be processed is automatically checked to ensure that there are no missing items in the document and that the content and format are complete and consistent.

[0058] Format consistency adjustment: Use Apache POI to process the entire document consistently to ensure that the format, such as font, font size, line spacing, etc., conforms to the overall file style, especially to unify the layout of the automatically generated content.

[0059] 5. Export and save the first pending bidding document

[0060] File format export: Export the generated first bidding document to be processed into a specific format (such as PDF, Word, etc.) as required.

[0061] File preservation and archiving: The first generated bidding document to be processed will be stored in the database or local file system for further manual or automatic review, and the relevant file index will be generated to facilitate subsequent calls and version management.

[0062] Through the above steps, the system can automatically obtain matching content from the material library, fill in and generate the first bid document to be processed that meets the specifications according to the predefined format, thereby improving the automation and standardization of document preparation.

[0063] Furthermore, step 4 also includes the generation of a report on content filling and formatting, which records in detail the source of each part of the material, the matching status and the adjustment steps, so as to facilitate manual review and improvement. If the system finds unmatched materials, inconsistent formats or missing information during the filling process, it will automatically generate a prompt to remind the user to supplement the content or adjust the format.

[0064] Step 5: Based on the final version of the bidding documents of the user's previous bidding projects, automatically compare, optimize and modify the first bidding document to be processed, and generate the second bidding document to be processed in the specified format. If the user uses this system for bidding for the first time and there is no final version of the bidding document of the user's previous bidding project in the material database, this step will not be performed;

[0065] The automatic comparison includes the comparison of text content, clauses, paragraphs, format structure, scoring items and technical clauses, specifically:

[0066] 1. Text content comparison: Use natural language processing (NLP) technology to compare the first bid document to be processed with the final version of the bid documents of previous bidding projects stored in multiple material libraries item by item to identify differences. Comparative identification can use the following methods: (1) semantic matching, using word vectors or pre-trained models (such as BERT, GPT) for semantic analysis to identify subtle differences in content (such as content with different wording but the same meaning); (2) keyword matching, extracting important keywords (such as project name, budget, qualification requirements, etc.) and comparing the accuracy of these keywords and whether the format meets the requirements.

[0067] 2. Clause and paragraph comparison: By comparing the clauses one by one, identify missing, repeated or conflicting clauses to ensure that the document content is complete and correct.

[0068] 3. Format structure comparison: Check whether the structure of the file conforms to the standard format, including title, chapter order, font size, paragraph style, etc., to ensure that the generated file is consistent with the reference file in structure.

[0069] 4. Comparison of scoring items and technical terms: Automatically compare scoring rules, technical parameters and other details, identify inconsistent terms and mark them for modification. For example, compare scoring items such as price weight and technical scoring rules to ensure that the scoring criteria are reasonable and fair.

[0070] Among them, automatic optimization includes optimization of language and format, clause optimization, scoring rules, and content compliance, specifically:

[0071] 1. Language and format optimization: Use language generation models (such as GPT-4, etc.) to optimize the language expression of the bidding documents to make the language more concise, professional, and standardized. For example, break down long sentences into short sentences to improve the readability of the document. Simplify long sentences and remove unnecessary wording to ensure that the document is concise and clear. Optimize the terms and expressions in the document to ensure that the wording meets the standards of the bidding documents.

[0072] 2. Clause optimization: Improve important clauses in the document based on best practices. For example, adjust the weight distribution of scoring criteria, optimize the description of payment terms, and avoid content that may affect fairness. Automatically identify unreasonable clauses (such as excessively high qualification thresholds, unreasonable budget requirements, etc.) and recommend compliant alternatives.

[0073] 3. Scoring criteria optimization: Automatically adjust the weight distribution of scoring criteria based on historical data or best cases to make it more in line with market standards.

[0074] 4. Content compliance optimization: Automatically compare the content with relevant laws and regulations and industry standards to ensure that the document terms meet compliance requirements. For example, if the contract period and liquidated damages stipulated in the terms do not meet industry standards, the system will automatically propose compliant modification suggestions.

[0075] Among them, automatic modification includes clauses, rule-driven, templated corrections, scoring rules, and format adjustments, specifically:

[0076] 1. Automatic adjustment of terms: According to the optimization plan and regulatory requirements, the system automatically modifies the terms and contents in the bidding documents of the current bidding project to make their contents comply with the requirements of the bidding documents.

[0077] 2. Rule-driven automatic adjustment: Automatically modify clauses that do not meet the specifications by setting modification rules (such as budget ratio, contract period, etc.). The rules are based on industry regulations or successful cases in past documents.

[0078] 3. Templated amendments: For clauses with high repetitiveness (such as payment clauses, confidentiality agreements, etc.), use predefined templates to make amendments to ensure consistency.

[0079] 4. Intelligent adjustment of scoring details: For the scoring details, automatically adjust the scoring weight, scoring criteria and other terms. For example, if the project technology weight is high, increase the weight of the technical score, while ensuring that the sum of the weights of each item is 100%.

[0080] 5. Format adjustment: According to the required export format (such as PDF, DOCX, XLSX, etc.), automatically adjust the content layout, font format, page layout, etc. to make the exported file meet the specified format requirements.

[0081] Among them, the specific implementation steps of generating and exporting the second pending bid document of this project in the specified format are: exporting the content to a specific format (such as PDF or Word) according to the needs to ensure that the final output format is unified and complies with the regulations; before exporting, perform final optimization of the format, such as adjusting paragraph spacing and page margins to ensure that the file layout is beautiful and clear; storing the generated second pending bid document in a database or local file system for further manual or automatic review, and generating relevant file indexes to facilitate subsequent calls and version management.

[0082] Furthermore, step 5 also includes the generation of optimization reports, intelligent prompts and feedback of review results. After the second pending bid document is generated, the system provides an automatic optimization report that lists the details of the document comparison, optimization, and modification, and explains the basis for optimization and modification for the convenience of user review. If anomalies are found in the document comparison or optimization (such as missing clauses, content that conflicts with regulations), the system generates detailed prompts for users to conduct manual inspection and final confirmation. This step can help realize the automated preparation, review and optimization of bidding documents, ensure that the content of the documents is compliant and the format is standardized, and improve the efficiency and quality of preparation.

[0083] Step 6: The system user manually confirms the second bid document to be processed and generates the final bid document for this project. If this is the first time the user uses this system to bid, this step is for the system user to manually confirm the first bid document to be processed and generate the final bid document for this project.

[0084] At the same time, the final version of the bidding document for this project generated in step 6 is automatically uploaded to the material database, and while being stored in the material database, it is marked according to key information such as project industry (such as highway, water transport, railway, building construction, etc.), bidding type (such as construction, supervision, design, auditing, cost consulting, etc.), and bidding requirements, so that data can be quickly retrieved when the bidding documents for other projects are prepared later (that is, in step 5).

[0085] An automatic bidding document writing system based on artificial intelligence includes the following modules:

[0086] Module 1: Database module, used to store data such as materials. The materials include basic information of the enterprise, common bidding certification materials, and various versions of the generated bidding documents. Keywords are set for the materials to facilitate later identification, matching, comparison, and retrieval.

[0087] Module 2: Tender document understanding module, which identifies tender documents and extracts the requirements of tender documents.

[0088] Module 3: Bidding document preparation module, used to generate bidding documents. Bidding documents are prepared according to the requirements of the identified and extracted bidding documents.

[0089] Module 4: Data upload module, uploads the data of this project and stores it in the material database.

[0090] Module 5: User interaction interface, which facilitates system users to operate the system.

[0091] Module 6: Learning module, which is also the user style recording module, automatically compares, optimizes and modifies the first bid document to be processed with the final version of the bid document of the previous bidding project in the material database, records the processing opinions of the system user unit, and records the results of the system user unit's manual confirmation of the bid document compiled by the system, and uses these feedback data to continuously train the machine learning model to gradually improve the accuracy and efficiency of the output of the bid document preparation module.

[0092] 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 bidding document writing method based on artificial intelligence, characterized in that: The following steps are involved: Step 1: Establish a material database and enter the bidding information of the system user, including basic information of the enterprise and common certification materials for bidding; Step 2: Establish a bidding document understanding module to identify the project bidding documents and extract the requirements of the bidding documents, including the bidding document format requirements, bidding scope and various response clause requirements in the bid evaluation method; Step 3: According to the requirements of the extracted bidding documents, automatically identify whether the materials in the material database are complete and valid. For materials that are incomplete or some of which have expired, generate a list to be updated and update the material database according to the requirements of the list; Step 4: After the materials in the material database are complete and valid, relevant materials are automatically obtained from the material database, and the first bidding document to be processed is automatically generated according to the requirements of the identified bidding documents; Step 5: Based on the final version of the bidding documents of multiple previous bidding projects of the user unit, automatically compare, optimize, and modify the first bidding document to be processed to generate a second bidding document to be processed; If the user unit is bidding for the first time using this system, and there is no final version of the bidding documents of the user unit's previous bidding projects in the material database, this step will not be required; Step 6: The system user unit manually confirms the second pending bid document to generate the final bid document for this project.

2. The method for automatically writing bidding documents based on artificial intelligence according to claim 1 is characterized in that: The basic information of the enterprise and the commonly used certification materials for bidding entered by the system user in the step 1 include business license, qualification certificate, award certification materials, financial audit statements, tax payment certificate, past performance contract agreement and its certification materials, personnel certificates, personnel social security payment certificates, personnel performance contract agreement, commonly used technical plans for various types of projects and other certification materials.

3. The method and system for automatically writing bidding documents based on artificial intelligence according to claim 1, characterized in that: In step 3, the material database is updated. The system automatically searches and downloads materials to update the material database based on the list to be updated.

4. The method for automatically writing bidding documents based on artificial intelligence according to claim 1 is characterized in that: In step 3, the material database is updated, and the system user unit may be reminded to manually enter the updated material database according to the list to be updated.

5. The method and system for automatically writing bidding documents based on artificial intelligence according to claim 1, characterized in that: The list to be updated generated in step 3 may also be supplemented or deleted manually by the user unit.

6. The automatic bidding document writing system based on artificial intelligence according to claim 1 is characterized in that: It includes the following modules: database module, bidding document understanding module, bidding document preparation module, data upload module, user interaction interface and learning module.

Images