Building engineering bidding document automatic generation method and system

Abstract

The invention provides a construction engineering bidding document automatic generation method and system, and belongs to the technical field of engineering construction. The method comprises the following steps of: preprocessing enterprise data, extracting information used in a bidding process from an unstructured enterprise data document by utilizing character recognition and natural language processing technologies, converting the information into structured data, and constructing a dynamic knowledge graph; generating a business bidding document based on a big language model according to the bidding document; constructing a technical scheme knowledge base, generating a response book and a technical scheme based on the technical scheme knowledge base and the bidding document, and integrating the response book and the technical scheme to generate a technical bidding document; and constructing an evaluation matrix, and evaluating the business bidding document and the technical bidding document. According to the method, the compiling efficiency and quality can be improved, and manual errors and time cost are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering construction, in particular to a building engineering tender automatic generation method and system. BACKGROUND

[0002] In the field of engineering construction, the preparation of a tender is a complex and time-consuming task. The traditional method of tender preparation requires manual collection and processing of a large amount of enterprise data, such as project contracts, personnel resumes, audit reports and financial reports, etc., and then manually writing the commercial tender and technical proposal according to the bidding requirements. This process is not only inefficient and prone to human error, but also difficult to ensure the quality and timely delivery of the tender when facing multiple projects or urgent bidding. With the emergence of large models, some solutions have also appeared for automatically writing tenders based on prompt words. However, this method of continuously optimizing the tender through multiple rounds of dialogue by constantly modifying the prompt words is not only difficult for users to understand, but also takes a long time and does not significantly improve efficiency. SUMMARY

[0003] The purpose of the present application is to provide a tender automatic generation method and system in the field of engineering construction, which can automatically process enterprise data and intelligently generate commercial tenders; can automatically create a technical proposal knowledge base and generate a response letter, technical proposal and progress plan based on the knowledge base, and finally generate a technical tender; can automatically evaluate and optimize the generated tender, thereby improving the efficiency and quality of preparation and reducing human error and time cost.

[0004] The present application provides a building engineering tender automatic generation method, comprising the following steps:

[0005] Step S1, enterprise data preprocessing, using text recognition and natural language processing technology to extract information used in the bidding process from unstructured enterprise data documents, convert it into structured data, and build a dynamic knowledge graph;

[0006] Step S2, generating a commercial tender based on a large language model according to the bidding document;

[0007] Step S3, building a technical proposal knowledge base, generating a response letter and technical proposal based on the technical proposal knowledge base and the bidding document, and integrating the response letter and technical proposal to generate a technical tender;

[0008] Step S4, building an evaluation matrix to evaluate the commercial tender and technical tender.

[0009] Preferably, step S1 comprises the following steps:

[0010] Step S101, introducing an OCR-NLP fusion engine to deeply analyze unstructured data;

[0011] Step S102, constructing a dynamic knowledge graph, establishing the correlation network between company qualifications, performance proof, financial documents, and personnel archives.

[0012] Preferably, step S101 specifically comprises:

[0013] Image preprocessing is performed on the scanned unstructured data;

[0014] Text recognition is performed on the preprocessed image of unstructured data using OCR;

[0015] Semantic analysis is performed using an NLP model, and context reasoning and error correction are performed.

[0016] Preferably, step S102 specifically comprises:

[0017] Entity recognition and relationship extraction are performed according to company qualifications, performance proof, financial documents, and personnel archives;

[0018] A multi-dimensional correlation network is constructed;

[0019] The knowledge graph is updated in real time, and the correlation reasoning is performed to update the correlation network.

[0020] Preferably, step S102 specifically comprises:

[0021] For the processing of company qualification files: according to the access person and the cited project, the company qualification files are multi-level encrypted, and the files with watermarks are displayed in the bidding process draft, and according to the user role, the predefined sensitive fields are automatically identified and desensitized;

[0022] For the processing of performance proof files: the contents of historical project contracts and bid-winning notices are automatically parsed by a large language model, and then the first page, the page with seals, and the clause page of the historical project contract are intercepted and merged into contract key pages, and the new merged contract key pages are classified and stored as structured data according to the project information;

[0023] For the processing of financial documents: for audit reports and financial reports, a large language model is used to parse multi-modal documents, and through natural language processing technology and reconciliation calculation, the total operating income, total cost, and cash flow are obtained and stored as structured data;

[0024] For the processing of personnel resumes: a large language model is used to automatically parse the resume files of company personnel, and based on the parsed content, predefined key information is automatically extracted, and the extracted personnel basic information, professional skills, and related project experience are stored as structured data.

[0025] Preferably, step S2 specifically comprises:

[0026] Step S201, perform tender document analysis and extract tender requirements;

[0027] Step S202, call the structured data obtained in step S1, perform dynamic knowledge graph query, and perform associated data matching;

[0028] Step S203, using a large language model and matching associated data, performing adaptive filling of the business tender template through context perception, performing compliance verification of the business tender, and generating the business tender.

[0029] Preferably, the construction of the technical solution knowledge base in step S3 specifically includes:

[0030] Constructing a three-dimensional knowledge graph;

[0031] Based on the three-dimensional knowledge graph, a technical solution knowledge base is constructed.

[0032] Preferably, the knowledge in the field of engineering and construction, such as common technical solutions, process flow, and material selection, is collected and organized, and a three-dimensional knowledge graph is constructed from the dimensions of technical parameters, engineering scenarios, and standard specifications.

[0033] Preferably, an active learning mechanism is introduced in the three-dimensional knowledge graph, and the technical solution knowledge base is continuously expanded through tender document analysis.

[0034] Preferably, a retrieval enhancement strategy is introduced, which specifically includes:

[0035] Calculate semantic similarity by combining cosine similarity and Jaccard coefficient to construct a semantic similarity calculation model;

[0036] Automatically supplement the synonyms of professional terms in the field of engineering and construction for query expansion.

[0037] Preferably, in step S3, generating a response based on the technical solution knowledge base and the tender document specifically includes: analyzing and extracting the content in the tender document that needs to be technically responded to, using natural language retrieval enhancement generation technology, automatically generating a response based on the existing technical solutions in the technical solution knowledge base, and the response includes intuitive conclusions such as positive deviation, negative deviation, and specific response technical items.

[0038] Preferably, in step S3, generating a technical solution based on the technical solution knowledge base and the tender document specifically includes: using natural language retrieval enhancement generation technology, analyzing the construction requirements in the tender document and the content in the technical solution knowledge base, and automatically generating a technical solution, which includes construction plan, construction process description, quality assurance measures, and safety assurance measures. In the process of generating the technical solution, the preset key technical parameters and points are automatically audited and verified through prompt words.

[0039] Preferably, in step S3, a progress plan is further generated, specifically comprising: based on specific response technical items, technical solutions, using natural language processing technology, analyzing and disassembling the construction process and work content in the technical solution, combining the overall project duration requirement, determining the work tasks, start and end time and duration of each stage, and generating a project progress plan table according to the determined work tasks and corresponding time.

[0040] Preferably, the contents of the response, technical solution and project progress plan table are integrated, and the technical bid is automatically typeset according to the format specified in the bidding document.

[0041] Preferably, step S4 comprises the following steps:

[0042] Step S401, a three-dimensional evaluation matrix containing technical score, business score and price score is constructed;

[0043] Step S402, based on reinforcement learning, dynamic weight adjustment is performed between the technical score, business score and price score according to the requirements of different bidding documents, and automatic evaluation and scoring are performed on the automatically generated business bid, technical solution, response and project duration table.

[0044] Preferably, after step S402, step S403 is further included, which automatically identifies index items with scores lower than a preset reference value and gives a prompt warning.

[0045] Preferably, after step S403, step S404 is further included, which performs grammar and spelling checking on the technical bid, gives an evaluation result, proposes improvement suggestions according to the evaluation result, and gives the chapters and page numbers that need to be rectified.

[0046] The application also provides a building engineering bid automatic generation system, comprising a processor, the processor can execute a computer program, the computer program can realize the steps of the building engineering bid automatic generation method.

[0047] Compared with the prior art, the application has the following advantages:

[0048] (1) The application automatically processes enterprise data, intelligently generates business bid, automatically creates technical solution knowledge base, and automatically generates response, technical solution and progress plan based on the knowledge base, and finally generates technical bid, which greatly improves the efficiency of bid preparation, saves a lot of manpower and time cost, and the bid preparation work that originally takes several days or even weeks can automatically generate high-quality results in a short time.

[0049] (2) The application automatically processes enterprise data, reduces the probability of human error, intelligently generates a business bid, automatically creates a technical solution knowledge base, and automatically generates a reply, a technical solution, and a progress plan based on the knowledge base, and finally generates a technical bid, improving the accuracy and reliability of the bid, and enabling customized bid generation based on historical data and project characteristics, highlighting the strengths and characteristics of the enterprise, and automatically evaluating and optimizing the generated bid, which also improves the accuracy and reliability of the bid and improves the probability of winning the bid. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on the structures shown in the drawings.

[0051] Figure 1 The enterprise data preprocessing flowchart of an embodiment of the present application includes two independent branches.

[0052] Figure 2 The schematic diagram of classifying and processing enterprise data according to an embodiment of the present application;

[0053] Figure 3 The business bid generation flowchart of an embodiment of the present application;

[0054] Figure 4 The technical bid generation flowchart of an embodiment of the present application;

[0055] Figure 5 The building engineering bid automatic generation method flowchart of an embodiment of the present application;

[0056] Figure 6 The building engineering bid automatic generation method flowchart of an embodiment of the present application; DETAILED DESCRIPTION

[0057] The specific embodiments of the present application will be described in detail below.

[0058] As shown in Figures 1-6 The present application provides a building engineering bid automatic generation method, comprising the following steps:

[0059] Step S1, enterprise data preprocessing, using character recognition and natural language processing technology to extract information used in the bidding process from unstructured enterprise data documents, convert it into structured data, and construct a dynamic knowledge graph;

[0060] Step S2, according to the bidding document, generating a commercial bid based on a large language model;

[0061] Step S3, constructing a technical scheme knowledge base, generating a response and a technical scheme based on the technical scheme knowledge base and the bidding document, integrating the response and the technical scheme to generate a technical bid;

[0062] Step S4, constructing an evaluation matrix, evaluating the commercial bid and the technical bid.

[0063] According to one embodiment of the present application, step S1 comprises the following steps:

[0064] Step S101, introducing an OCR-NLP fusion engine to deeply analyze unstructured data;

[0065] Step S102, constructing a dynamic knowledge graph to establish a correlation network among company qualifications, performance certificates, financial documents and personnel archives. Step S101 and step S102 can be two independent branches and do not have a specific order.

[0066] According to one embodiment of the present application, step S101 specifically comprises:

[0067] Image preprocessing is performed on the scanned copy of unstructured data;

[0068] OCR is used to recognize the text of the preprocessed unstructured data image;

[0069] NLP model is used for semantic analysis, context reasoning and error correction.

[0070] According to one embodiment of the present application, step S102 specifically comprises:

[0071] Entity recognition and relationship extraction are performed according to company qualifications, performance certificates, financial documents and personnel archives;

[0072] A multi-dimensional correlation network is constructed;

[0073] The knowledge graph is updated in real time, and the correlation relationship is reasoned to update the correlation network.

[0074] According to one embodiment of the present application, step S102 specifically comprises:

[0075] For company qualification file processing: according to the access person and the cited project, the company qualification file is multi-level encrypted, and the file with watermark is displayed in the bidding process draft, and according to the user role, the predefined sensitive field is automatically identified and desensitized;

[0076] For the processing of performance proof type files: the content of historical project contracts and bid-winning notices is automatically parsed by a large language model, and then the first page, the seal page and the clause page of the historical project contract are intercepted page by page to merge into contract key pages, and the newly merged contract key pages are classified and stored as structured data according to project information;

[0077] For the processing of financial files: for audit reports and financial reports, a large language model is used to parse multi-modal documents, and through natural language processing technology and reconciliation calculation, total operating income, total cost and cash flow are obtained and stored as structured data;

[0078] For the processing of personnel resumes: a large language model is used to automatically parse the resume files of company personnel, and based on the parsed content, predefined automatic extraction of key information is performed, and the extracted personnel basic information, professional skills and relevant project experience are stored as structured data.

[0079] According to one specific embodiment of the present application, step S2 specifically comprises:

[0080] Step S201, performing tender document parsing to extract tender requirements;

[0081] Step S202, calling the structured data obtained in step S1 to perform dynamic knowledge graph query and associated data matching;

[0082] Step S203, using a large language model and the matched associated data to perform adaptive filling of the business tender template through context perception, performing compliance verification of the business tender, and generating the business tender.

[0083] According to one specific embodiment of the present application, the construction of the technical scheme knowledge base in step S3 specifically comprises:

[0084] Constructing a three-dimensional knowledge graph;

[0085] Based on the three-dimensional knowledge graph, a technical scheme knowledge base is constructed.

[0086] According to one specific embodiment of the present application, common technical schemes, process flows and material selection knowledge in the field of engineering and construction are collected and organized, and a three-dimensional knowledge graph is constructed from the dimensions of technical parameters, engineering scenarios and specification standards.

[0087] According to one specific embodiment of the present application, an active learning mechanism is introduced in the three-dimensional knowledge graph, and the technical scheme knowledge base is continuously expanded through tender document parsing.

[0088] According to one specific embodiment of the present application, a retrieval enhancement strategy is introduced, specifically comprising:

[0089] The semantic similarity is calculated by combining cosine similarity and Jaccard coefficient, and a semantic similarity calculation model is constructed.

[0090] The professional term synonyms in the field of engineering construction are automatically supplemented to perform query expansion.

[0091] According to one specific embodiment of the present application, in step S3, generating the response letter based on the technical scheme knowledge base and the bidding document specifically includes: analyzing and extracting the content in the bidding document that needs to be technically responded, using natural language retrieval enhancement generation technology, and automatically generating the response letter according to the existing technical scheme in the technical scheme knowledge base, wherein the response letter includes intuitive conclusions such as positive deviation and negative deviation, as well as specific response technical items.

[0092] According to one specific embodiment of the present application, in step S3, generating the technical scheme based on the technical scheme knowledge base and the bidding document specifically includes: using natural language retrieval enhancement generation technology, analyzing the construction requirements in the bidding document and the content in the technical scheme knowledge base, and automatically generating the technical scheme, wherein the technical scheme includes a construction scheme, a construction process description, quality assurance measures, and safety guarantee measures, and in the process of generating the technical scheme, the preset key technical parameters and points are automatically audited and verified through prompt words.

[0093] According to one specific embodiment of the present application, in step S3, generating the progress plan specifically includes: based on the specific response technical items and the technical scheme, using natural language processing technology, analyzing and disassembling the construction process and work content in the technical scheme, combining with the overall project duration requirement to determine the work tasks, start and end times and durations of each stage, and generating a project progress plan table according to the determined work tasks and corresponding times.

[0094] According to one specific embodiment of the present application, the contents of the response letter, the technical scheme and the project progress plan table are integrated, and a technical bid is automatically generated in the format specified in the bidding document.

[0095] According to one specific embodiment of the present application, step S4 includes the following steps:

[0096] Step S401: constructing a three-dimensional evaluation matrix including technical scores, business scores and price scores;

[0097] Step S402: based on the requirements of different bidding documents, performing dynamic weight adjustment among the technical scores, the business scores and the price scores by reinforcement learning, and automatically evaluating and scoring the automatically generated business bid, technical scheme, response letter and project duration table.

[0098] According to one specific embodiment of the present application, after step S402, step S403 is further included, which automatically identifies the index items with scores lower than a preset reference value and gives a prompt warning.

[0099] According to one specific embodiment of the present application, after step S403, step S404 is further included, that is, performing grammar and spelling checking on the technical bid, giving an evaluation result, and giving improvement suggestions according to the evaluation result, and giving chapters and page numbers that need to be rectified.

[0100] The present application also provides a construction engineering bid automatic generation system, comprising a processor, wherein the processor can execute a computer program, and the computer program can realize the steps of the construction engineering bid automatic generation method.

[0101] Embodiment 1

[0102] As shown in the Figure 6 The present application provides a construction engineering bid automatic generation method, comprising the following steps:

[0103] Step S1, enterprise data preprocessing, using character recognition and natural language processing technology to extract information used in the bidding process from unstructured enterprise data documents, convert them into structured data, and build a dynamic knowledge graph;

[0104] Step S2, generating a commercial bid based on a large language model according to the bidding document;

[0105] Step S3, constructing a technical scheme knowledge base, generating a reply and a technical scheme based on the technical scheme knowledge base and the bidding document, and integrating the reply and the technical scheme to generate a technical bid;

[0106] Step S4, constructing an evaluation matrix to evaluate the commercial bid and the technical bid.

[0107] Embodiment 2

[0108] As shown in the Figures 1-6 The present application provides a construction engineering bid automatic generation method, comprising the following steps:

[0109] Step S1, enterprise data preprocessing, using character recognition and natural language processing technology to extract information used in the bidding process from unstructured enterprise data documents, convert them into structured data, and build a dynamic knowledge graph;

[0110] Step S1 specifically comprises the following steps:

[0111] Step S101, introducing an OCR-NLP fusion engine to deeply analyze unstructured data;

[0112] Step S102, constructing a dynamic knowledge graph, establishing the correlation network between company qualifications, performance proof, financial documents, personnel archives. Step S101 and step S102 can be two independent branches, and the order is not important.

[0113] The step S101 specifically comprises:

[0114] Image preprocessing is performed on the scanned unstructured data;

[0115] Text recognition is performed on the preprocessed image of unstructured data using OCR;

[0116] Further semantic analysis is performed using NLP model, and context reasoning and error correction are performed.

[0117] The step S102 specifically comprises:

[0118] Entity recognition and relationship extraction are performed according to company qualifications, performance proof, financial documents, personnel archives;

[0119] A multi-dimensional correlation network is constructed;

[0120] The knowledge graph is updated in real time, and the correlation relationship is inferred, and the correlation network is updated.

[0121] Further, the step S102 specifically comprises:

[0122] For the processing of company qualification files: according to the access person and the reference project, the company qualification files are multi-level encrypted, the files with watermarks are displayed in the process manuscript, and the pre-defined sensitive fields are automatically identified according to the user role, and the hierarchical desensitization is performed; this part generally refers to the company's business license and industry license, etc., because multiple people may be involved in the generation of the tender document, in order to ensure the safety of the company qualification files, the qualification files can be multi-level encrypted according to the access person and the reference project. In the process manuscript, all the files displayed are with watermarks, and a dynamic desensitization algorithm is used to automatically identify sensitive fields for hierarchical desensitization according to the user role.

[0123] For the processing of performance proof type files: automatically parse the contents of historical project contracts and bid-winning notices through a large language model, then intercept the front page, seal page, and clause page of the historical project contract by page, merge into contract key pages, and store the newly merged contract key pages as structured data according to project information. For this part, the general commercial bid part will be attached with similar historical related project information such as project contracts or bid-winning notices. The content of the contract and the bid-winning notice can be automatically understood through a large language model, then the front page, seal page, and clause page of the contract are intercepted and merged, and the newly merged contract key pages are automatically classified and stored according to project information such as name, contract amount, and field.

[0124] For the processing of financial files: for audit reports and financial reports, use a large language model to parse multi-modal documents, obtain total operating income, total cost, and cash flow through natural language processing techniques and reconciliation calculations, and store them as structured data.

[0125] For the processing of personnel resumes: use a large language model to automatically parse the resume files of company personnel, and based on the parsed content, automatically extract predefined key information, store the extracted personnel basic information such as name, education, title, work experience, professional skills, and relevant project experience as structured data.

[0126] Step S2, generate a commercial bid based on a large language model according to the bidding document; step S2 specifically includes:

[0127] Step S201, perform bidding document parsing and extract bidding requirements;

[0128] Step S202, call the structured data obtained in step S1, perform dynamic knowledge graph query, and perform associated data matching;

[0129] Step S203, use a large language model and matched associated data to perform adaptive filling of a commercial bid template through context perception, perform compliance verification of the commercial bid, and generate a commercial bid.

[0130] Step S3, build a technical scheme knowledge base, generate a reply and a technical scheme based on the technical scheme knowledge base and the bidding document, and integrate the reply and the technical scheme to generate a technical bid; the construction of the technical scheme knowledge base in step S3 specifically includes:

[0131] Build a three-dimensional knowledge graph; specifically, collect and organize knowledge of common technical schemes, process flows, and material selection in the field of engineering and construction, and build a three-dimensional knowledge graph from the dimensions of technical parameters, engineering scenarios, and specification standards.

[0132] Based on the three-dimensional knowledge graph, a technical solution knowledge base is constructed, and further, an active learning mechanism is introduced into the three-dimensional knowledge graph, and the technical solution knowledge base is continuously expanded through the analysis of the bidding documents.

[0133] Further, a retrieval enhancement strategy is introduced, which specifically includes:

[0134] The semantic similarity is calculated by combining the cosine similarity and the Jaccard coefficient, and a semantic similarity calculation model is constructed.

[0135] The professional term synonyms in the field of engineering and construction are automatically supplemented for query expansion.

[0136] Further, in step S3, the response letter is generated based on the technical solution knowledge base and the bidding document, which specifically includes: analyzing and extracting the content in the bidding document that needs to be technically responded, using natural language retrieval enhancement generation technology, automatically generating a response letter based on the existing technical solutions in the technical solution knowledge base, the response letter including intuitive conclusions such as positive deviation, negative deviation, and specific response technical items;

[0137] Further, in step S3, the technical solution is generated based on the technical solution knowledge base and the bidding document, which specifically includes: using natural language retrieval enhancement generation technology, analyzing the construction requirements in the bidding document and the content in the technical solution knowledge base, automatically generating a technical solution, the technical solution including construction scheme, construction process description, quality assurance measures, and safety guarantee measures, and in the process of generating the technical solution, the preset key technical parameters and points are automatically audited and verified through prompt words to ensure the scientificity and feasibility of the technical solution.

[0138] Further, in step S3, it also includes generating a progress plan, which specifically includes: based on the specific response technical items and the technical solution, using natural language processing technology, analyzing and disassembling the construction process and work content in the technical solution, combining with the overall project duration requirement to determine the work tasks, start and end time and duration of each stage, and according to the determined work tasks and corresponding time, generating a project progress plan table.

[0139] Further, the contents of the response letter, the technical solution, and the project progress plan table are integrated, and the technical bid is automatically typeset according to the format specified in the bidding document.

[0140] Step S4, an evaluation matrix is constructed to evaluate the commercial bid and the technical bid, and step S4 includes the following steps:

[0141] Step S401, a three-dimensional evaluation matrix including technical score, commercial score, and price score is constructed;

[0142] In step S402, the automatically generated business bid, technical scheme, reply and project duration table are automatically evaluated and scored based on the requirements of different bidding documents, dynamic weight adjustment between technical score, business score and price score is performed based on reinforcement learning.

[0143] In an embodiment, the evaluation result is used to propose improvement suggestions, and the chapters and page numbers that need to be improved are given to facilitate user positioning and optimization of the bid.

[0144] Further, step S402 further includes step S403 of automatically identifying index items with scores lower than a preset reference value and giving a prompt warning.

[0145] Further, step S403 further includes step S404 of performing grammar and spelling checking on the technical bid, giving an evaluation result, proposing improvement suggestions according to the evaluation result, and giving chapters and page numbers that need to be improved.

[0146] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application is included in the protection scope of the present application.

Claims

1. A method for automatically generating construction project bid documents, characterized in that: The steps include: Step S1: Preprocessing of enterprise data: Using text recognition and natural language processing technologies, the information used in the bidding process is extracted from unstructured enterprise data documents, converted into structured data, and a dynamic knowledge graph is constructed. Step S2: Generate a business bid based on the bidding document and the large language model; Step S3: construct a technical solution knowledge base, generate a response document and a technical solution based on the technical solution knowledge base and the bidding documents, and integrate the response document and the technical solution to generate a technical bid; Step S4: construct an evaluation matrix to evaluate the commercial bid and the technical bid.

2. The method for automatically generating construction project bid documents according to claim 1, characterized in that: Step S1 includes the following steps: Step S101: Introduce the OCR-NLP fusion engine to perform in-depth analysis on unstructured data; Step S102: construct a dynamic knowledge graph and establish an association network between company qualifications, performance certificates, financial documents, and personnel files.

3. The method for automatically generating construction project bid documents according to claim 2, characterized in that: Step S101 specifically includes: performing image preprocessing on the scanned copy of the unstructured data; Use OCR to perform text recognition on the pre-processed unstructured data images; Then use the NLP model to perform semantic analysis, contextual reasoning and error correction.

4. The method for automatically generating construction project bid documents according to claim 2, characterized in that: Step S102 specifically includes: performing entity identification and relationship extraction based on company qualifications, performance certificates, financial documents, and personnel files; Construct a multi-dimensional association network; Update the knowledge graph in real time, perform relationship reasoning, and update the association network.

5. The method for automatically generating construction project bid documents according to claim 4, characterized in that: Step S102 specifically includes: Regarding the processing of company qualification documents: We implement multi-level encryption based on the accessor and referenced project, displaying watermarked documents in the tender process draft. Predefined sensitive fields are automatically identified based on user roles and graded desensitization is performed. For performance certification documents, the system automatically parses the content of historical project contracts and notices of winning bids using a large language model. The system then extracts the front page, stamp page, and clause page of the historical project contracts, merging them into key contract pages. These newly merged key contract pages are then classified and stored as structured data based on project information. Financial document processing: For audit reports and financial statements, we use large language models to parse multimodal documents. Through natural language processing technology and cross-reference calculations, we obtain total operating revenue, total costs, and cash flow, and store them as structured data. For processing personnel resumes: Use a large language model to automatically parse the resume files of company personnel, and automatically extract predefined key information based on the parsed content, and store the extracted basic information, professional skills and relevant project experience as structured data.

6. The method for automatically generating construction project bid documents according to claim 1, characterized in that: Step S2 specifically includes: Step S201: parsing the bidding documents and extracting the bidding requirements; Step S202: Call the structured data obtained in step S1 to perform dynamic knowledge graph query and match related data; Step S203 , using the large language model and the matched associated data, adaptively filling in the business proposal template through context awareness, performing compliance verification on the business proposal, and generating the business proposal.

7. The method for automatically generating construction project bid documents according to claim 1, characterized in that: Constructing the technical solution knowledge base in step S3 specifically includes: Build a three-dimensional knowledge graph; Build a technical solution knowledge base based on the three-dimensional knowledge graph.

8. The method for automatically generating construction project bid documents according to claim 7, characterized in that: Collect and organize knowledge on common technical solutions, process flows, and material selection in the field of engineering construction, and construct a three-dimensional knowledge graph from the dimensions of technical parameters, engineering scenarios, and specifications and standards.

9. The method for automatically generating construction project bid documents according to claim 7, characterized in that: Introduce active learning mechanism into the three-dimensional knowledge graph, and continuously expand the technical solution knowledge base through bidding document analysis.

10. The method for automatically generating construction project bid documents according to claim 7, characterized in that: Introducing search enhancement strategies, including: Combine cosine similarity and Jaccard coefficient to calculate semantic similarity and build a semantic similarity calculation model; Automatically supplement synonyms of professional terms in the field of engineering and construction to expand queries.

11. The method for automatically generating construction project bid documents according to claim 1, characterized in that: In step S3, generating a response based on the technical solution knowledge base and the bidding documents specifically includes: analyzing and extracting the content that requires technical response in the bidding documents, using natural language retrieval enhancement generation technology, and automatically generating a response based on the existing technical solutions in the technical solution knowledge base. The response includes intuitive conclusions such as positive deviation, negative deviation, and specific response technical items.

12. The method for automatically generating construction project bid documents according to claim 1, characterized in that: In step S3, generating a technical solution based on the technical solution knowledge base and the bidding documents specifically includes: using natural language retrieval to enhance generation technology, analyzing the construction requirements in the bidding documents and the content in the technical solution knowledge base, and automatically generating a technical solution. The technical solution includes a construction plan, construction process instructions, quality assurance measures, and safety measures. In the process of generating the technical solution, the preset key technical parameters and key points are automatically reviewed and verified through prompt words.

13. The method for automatically generating construction project bid documents according to claim 1, characterized in that: Step S3 also includes generating a schedule, which specifically includes: based on specific response technical items and technical solutions, using natural language processing technology to analyze and disassemble the construction process and work content in the technical solution, combined with the overall project schedule requirements, determine the work tasks, start and end times and duration of each stage, and generate a project schedule based on the determined work tasks and corresponding time.

14. The method for automatically generating construction project bid documents according to claim 13, characterized in that: Integrate the contents of the response, technical plan, and project schedule, and automatically generate the technical bid in the format specified in the bidding documents.

15. The method for automatically generating construction project bid documents according to claim 1, characterized in that: Step S4 includes the following steps: Step S401, constructing a three-dimensional evaluation matrix including technical score, business score, and price score; Step S402: Based on the requirements of different bidding documents, dynamic weight adjustment is performed among technical scores, business scores, and price scores based on reinforcement learning, and the automatically generated business bids, technical proposals, response documents, and project schedules are automatically evaluated and scored.

16. The method for automatically generating construction project bid documents according to claim 15, characterized in that: Step S402 also includes step S403, which automatically identifies indicator items with scores lower than a preset benchmark value and issues a prompt warning.

17. The method for automatically generating construction project bid documents according to claim 16, characterized in that: Step S403 is followed by step S404, in which the technical specification is checked for grammar and spelling, an evaluation result is given, improvement suggestions are made based on the evaluation result, and the chapters and pages that need to be corrected are given.

18. A construction project bidding document automatic generation system, characterized by: It includes a processor, which can execute a computer program, and the computer program can implement the steps of the method for automatically generating construction project bids as described in any one of claims 1 to 17.

Images