Building construction cost calculation intelligent system and method based on large language model

Abstract

The invention relates to the technical field of engineering cost calculation, in particular to a building construction cost calculation intelligent system and method based on a large language model. Comprising a user-defined submenu module, a data acquisition and analysis module, a form file analysis module, a large language model matching module, an optimization and auditing module, an internet collaboration and sharing module and a one-key quotation generation module. By constructing an end-to-end automation system integrated with a special large language model intelligent matching core, the whole process from automatic acquisition and standardization of multi-source heterogeneous data, intelligent project matching based on semantic comprehension, compliance auditing and optimization of man-machine collaboration and final quotation file generation is realized; therefore, on the premise of keeping the compliance of industry specifications, the processing efficiency and the result accuracy of the cost calculation amount are remarkably improved, and the adaptability of the system to globalized complex projects is enhanced.

Description

Technical Field

[0001] This invention relates to the field of engineering cost estimation technology, specifically an intelligent system and method for calculating construction cost based on a large language model. Background Technology

[0002] In the field of construction engineering, cost estimation is a core component of project investment control, bidding decisions, and construction cost management. Its accuracy directly impacts the project's economic benefits and market competitiveness. Traditional cost estimation work relies heavily on manual operation. Professionals must manually compare the tender bill of quantities with the company's internal project cost guidance documents, matching detailed information such as project name, technical characteristics, and units of measurement one by one. This process is tedious and time-consuming, especially for large projects with thousands of sub-items, where the manual processing cycle often takes weeks, making it difficult to meet the demands of rapid bidding.

[0003] With the accelerating globalization of the construction industry and the increasing number of international engineering projects, cost estimation and quantity calculation work faces new challenges. On the one hand, tender documents may involve multiple languages, and differences in the expression of professional terminology can easily lead to errors in manual recognition and translation. On the other hand, pricing in different currencies requires real-time exchange rate conversion, further increasing the complexity and risk of errors in the operation. Although some cost estimation software exists on the market that can achieve basic data import and calculation functions, it still relies on manual preprocessing when handling unstructured or multi-format documents (such as PDFs and complex Excel spreadsheets). Moreover, at the information matching level, it is mostly based on keyword literal matching, which is difficult to deal with situations where the semantics are the same but the expressions are different (such as "C30 commercial concrete" and "pre-mixed C30 concrete"). Furthermore, it lacks the ability to automatically parse and intelligently convert multilingual text.

[0004] In recent years, artificial intelligence technology, especially large language models, has made significant progress in natural language understanding and cross-modal information processing, providing new technological pathways for automation and intelligent applications in vertical industries. However, how to deeply integrate these cutting-edge technologies with the specific business scenario of construction engineering cost estimation—which is highly specialized, standardized, and involves diverse data sources—to build an intelligent quantity calculation system capable of end-to-end processing of multi-source data, achieving accurate semantic matching, and supporting collaborative operations, remains a pressing technical challenge for the industry. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide an intelligent system and method for calculating construction costs based on a large language model, which improves the efficiency and accuracy of cost calculation and quotation.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] An intelligent system for calculating construction cost based on a large language model includes a custom sub-menu module, a data acquisition and parsing module, a table file analysis module, a large language model matching module, an optimization and review module, an internet collaboration and sharing module, and a one-click quotation generation module.

[0008] The custom submenu module is connected to the data acquisition and parsing module, the table file analysis module, and the large language model matching module, respectively, and is used to pass configuration parameters to the modules.

[0009] The data acquisition and parsing module and the tabular file analysis module are respectively connected to the large language model matching module, and are used to provide the large language model matching module with standardized project cost data and tender list data;

[0010] The large language model matching module is connected to the optimization and review module and is used to output preliminary matching results to the optimization and review module.

[0011] The optimization and review module is connected to the one-click quote generation module and is used to transmit the reviewed and confirmed quote data to the one-click quote generation module.

[0012] The Internet collaboration and sharing module is connected to the data acquisition and parsing module, the table file analysis module, the optimization and review module, and the one-click quotation generation module, respectively, and is used to provide the collaborative operation environment and data storage management services for the modules.

[0013] The custom submenu module includes configuration submenu items with a visual configuration interface and smart input submenu items that provide process triggering functionality;

[0014] The data acquisition and parsing module includes a multi-source data acquisition unit for acquiring project cost management guidance price documents from multiple sources based on user configuration, and a multi-language parsing and structuring unit for performing multi-language parsing and structuring processing on the acquired documents.

[0015] The table file analysis module includes a file format adaptation unit and a key information extraction unit for adapting to various formats of tender list files and extracting key information, as well as a currency conversion unit for performing automatic currency conversion;

[0016] The large language model matching module includes a large language model construction unit fine-tuned based on the construction industry cost calculation corpus and an intelligent matching unit for performing intelligent matching and outputting results;

[0017] The optimization and review module includes an automatic optimization unit with a built-in library of industry standards and enterprise standard rules, and a manual review unit that provides a visual user interface.

[0018] The Internet collaboration and sharing module includes a collaborative work platform that supports multi-role permission management and real-time collaboration, and a cloud storage and version management unit for realizing cloud storage and version management of data.

[0019] The one-click quotation generation module includes a total price calculation unit for calculating the total project quotation and a bill of quantities generation and export unit for generating and exporting a bill of quantities with prices.

[0020] This invention also discloses an intelligent method for calculating construction cost based on a large language model, the specific steps of which are as follows:

[0021] S1: Receives user-configured system parameters and process triggering instructions through a custom submenu module.

[0022] S2: Automatically acquires project cost management guidance price data through the data acquisition and parsing module, and performs multilingual parsing, structured conversion and verification on the data to generate standardized guidance price structured data.

[0023] S3: Read the tender list file through the table file analysis module, extract key information, and perform automatic currency conversion when the pricing currency is found to be inconsistent with the target currency, generating structured list data that is compatible with the guidance price data format.

[0024] S4: The large language model matching module calls the large language model that has been fine-tuned with corpus in the construction field. Based on multi-level matching rules, it intelligently matches the structured data of the list with the structured data of the guidance price, and outputs the matching results and similarity analysis report.

[0025] S5: The optimization and review module automatically verifies the compliance of the matching results, generates optimization suggestions, and submits them for manual review and confirmation through a visual interface.

[0026] S6: Provides a permission-based real-time collaboration environment for multi-role users through the Internet collaboration and sharing module, and encrypts and stores the approved data and operation logs to the cloud server.

[0027] S7: The one-click quotation generation module automatically calculates the total quotation based on the final confirmed unit price and quantity, and generates a bill of quantities file with price that conforms to national standards.

[0028] The present invention, which adopts the above technical solution, has the following prominent features compared with the prior art:

[0029] By constructing an end-to-end automated system that integrates a dedicated large language model intelligent matching core, the entire process is seamlessly connected, from automatic acquisition and standardization of multi-source heterogeneous data, intelligent project matching based on semantic understanding, compliance review and optimization through human-machine collaboration, to the generation of the final quotation document. This significantly improves the processing efficiency and accuracy of cost estimation while maintaining compliance with industry standards, and enhances the system's adaptability to complex global projects.

[0030] As a preferred embodiment, a further technical solution of the present invention is:

[0031] Preferably, the configuration sub-menu item has a parameter configuration interface for setting the data acquisition source, multilingual recognition preference, currency conversion benchmark, and data matching accuracy threshold. Through the integrated parameter configuration interface, the configuration sub-menu item allows users to flexibly set the data source, language, currency, and matching accuracy, enabling the system to be personalized to meet the needs of different projects, thereby improving the system's versatility and the ease of user operation.

[0032] Preferably, the multi-source data acquisition unit is equipped with a first acquisition interface for downloading guidance price documents from the enterprise integrated management system via application programming interface calls, and a second acquisition interface for reading guidance price data from the local database via structured query language. By providing standardized data interfaces for the enterprise management system and the local database, the multi-source data acquisition unit realizes unified and automated acquisition of structured and unstructured cost guidance price data, effectively reducing manual intervention and time costs in the data preparation stage.

[0033] Preferably, the multilingual parsing and structuring unit integrates optical character recognition technology and a multilingual natural language processing model. It is used to extract project name, project characteristics, and unit of measurement fields from unstructured documents and convert them into structured data of a preset data model. The multilingual parsing and structuring unit, combining optical character recognition and multilingual natural language processing technology, can automatically and accurately extract and structure key cost information from files of different formats. This solves the problems of low efficiency and easy omissions in manual processing of unstructured documents, and provides a high-quality data foundation for subsequent intelligent matching.

[0034] Preferably, the large language model building unit is equipped with a model training architecture that uses low-rank adaptation technology to fine-tune the general large language model. This architecture is based on a construction industry cost calculation corpus that includes construction engineering quantity list pricing specifications, enterprise historical matching cases, and multilingual project feature description corpora. The large language model building unit adopts fine-tuning technology based on the construction industry-specific corpus, enabling the general model to have the ability to deeply understand engineering cost professional terminology, specifications, and differences in multilingual expressions, providing industry knowledge support for core matching tasks and improving the reliability and professionalism of semantic matching.

[0035] Preferably, the intelligent matching unit is configured with multi-level matching logic that includes matching weights for project name and unit of measurement, as well as matching weights for project feature details. This unit calculates the total similarity between the tender list projects and the guidance price projects based on the semantic understanding capabilities of the large language model. Through the multi-level matching logic with configurable weights, the intelligent matching unit simulates the comprehensive judgment process of professionals, and can more comprehensively and evenly consider the semantic similarity of project name, unit, and feature details, thereby providing more accurate and reasonable matching suggestions when facing real business scenarios with diverse expressions.

[0036] Preferably, the collaborative work platform is equipped with a permission management component that assigns data operation permissions based on user roles, a real-time collaboration component that uses WebSocket technology to achieve data synchronization, and a message notification component for pushing audit reminders and anomaly prompts. By integrating role permission management, real-time data synchronization and message notification mechanisms, the collaborative work platform builds a working environment that supports online collaboration among multiple users, which helps to break down information silos, promote team collaboration, and ensure the timely progress and transparent management of key operation processes. Attached Figure Description

[0037] Figure 1 This is a system hierarchy architecture diagram according to an embodiment of the present invention;

[0038] Figure 2 This is a system module architecture diagram of an embodiment of the present invention;

[0039] Figure 3 This is a flowchart illustrating the principle of the large language model matching module in an embodiment of the present invention.

[0040] Figure 4 This is a schematic diagram of the optimization and review module of this invention.

[0041] Figure 5 This is a system operation flowchart of an embodiment of the present invention. Detailed Implementation

[0042] The present invention will be further illustrated below with reference to specific embodiments. The purpose of this illustration is solely to provide a better understanding of the invention. Therefore, the examples given do not limit the scope of protection of the present invention.

[0043] like Figures 1 to 5 As shown in the figure, this embodiment presents an intelligent system for calculating construction cost based on a large language model, including a custom sub-menu module, a data acquisition and parsing module, a table file analysis module, a large language model matching module, an optimization and review module, an Internet collaboration and sharing module, and a one-click quotation generation module;

[0044] The custom submenu module is connected to the data acquisition and parsing module, the table file analysis module, and the large language model matching module, respectively, and is used to pass configuration parameters to the modules.

[0045] The data acquisition and parsing module and the tabular file analysis module are respectively connected to the large language model matching module, and are used to provide the large language model matching module with standardized project cost data and tender list data;

[0046] The large language model matching module is connected to the optimization and review module and is used to output preliminary matching results to the optimization and review module.

[0047] The optimization and review module is connected to the one-click quote generation module and is used to transmit the reviewed and confirmed quote data to the one-click quote generation module.

[0048] The Internet collaboration and sharing module is connected to the data acquisition and parsing module, the table file analysis module, the optimization and review module, and the one-click quotation generation module, respectively, and is used to provide the collaborative operation environment and data storage management services for the modules.

[0049] The custom submenu module includes configuration submenu items with a visual configuration interface and smart input submenu items that provide process triggering functionality. The smart input submenu item provides a "one-click start" button to trigger the system to execute the entire process in the order of "data acquisition → parsing → matching → review → quotation generation"; it also provides a "step-by-step start" option, allowing users to trigger a specific module individually.

[0050] The data acquisition and parsing module includes a multi-source data acquisition unit for acquiring project cost management guidance price documents from multiple sources based on user configuration, and a multi-language parsing and structuring unit for performing multi-language parsing and structuring processing on the acquired documents.

[0051] The tabular file analysis module includes a file format adaptation unit and a key information extraction unit for adapting to various formats of tender list files and extracting key information, as well as a currency conversion unit for performing automatic currency conversion. The file format adaptation unit supports reading Excel, CSV, and PDF tabular tender list files, automatically identifies file encoding and table structure, and processes merged cells and nested tables through a table reconstruction algorithm. The key information extraction unit uses a multilingual NLP model to extract project name, project characteristics, unit of measurement, quantity, pricing currency type, and schedule requirements from the tender list. The currency conversion unit automatically calls a third-party exchange rate interface to obtain the real-time exchange rate if the pricing currency in the tender list is inconsistent with the user-configured target currency, calculates the converted amount (formula: converted amount = original amount × real-time exchange rate), and adds the fields original_currency, exchange_rate, and converted_currency to the structured data to record the conversion process.

[0052] The large language model matching module includes a large language model construction unit fine-tuned based on the construction industry cost calculation corpus and an intelligent matching unit for performing intelligent matching and outputting results;

[0053] The optimization and review module includes an automatic optimization unit with a built-in industry standard and enterprise standard rule library, and a manual review unit with a visual operation interface. The automatic optimization unit is implemented by using the built-in rule library of "Construction Engineering Quantity List Pricing Specification" and "Enterprise Cost Control Standard" to automatically check whether the unit price of the matching result is within the range of the guidance price fluctuation and whether the project characteristics fully cover the bidding requirements, and generates optimization suggestions. The manual review unit is implemented by using Vue.js + Element UI to build a visual review interface, displaying the matching results, optimization suggestions and standard clauses, and supporting users to confirm, modify and rematch operations; it automatically records the review operations and generates operation logs.

[0054] The Internet collaboration and sharing module includes a collaborative work platform that supports multi-role permission management and real-time collaboration, and a cloud storage and version management unit for data cloud storage and version management. The cloud storage and version management unit is implemented by using AES-256 encryption technology to store data on a cloud server and supporting multi-region backup; it automatically records file modification versions and supports historical version rollback.

[0055] The one-click quotation generation module includes a total price calculation unit for calculating the total project price and a bill of quantities generation and export unit for generating and exporting a bill of quantities with prices. The total price calculation unit calculates the total price of each item based on the approved guidance price and the quantities in the tender list and summarizes them. It supports automatic calculation of taxes and fees according to the configured tax rate. The bill of quantities generation and export unit has a built-in bill of quantities template that conforms to the national standard GB 50500-2013. It supports exporting the bill of quantities with prices to Excel, PDF, and Word formats, and annotates the original currency type, exchange rate, converted amount, reviewer, and generation time information in the exported file.

[0056] A method for intelligent construction cost estimation based on a large language model is also disclosed, with the following specific steps:

[0057] S1: Receives user-configured system parameters and process triggering instructions through a custom submenu module.

[0058] S2: Automatically acquires project cost management guidance price data through the data acquisition and parsing module, and performs multilingual parsing, structured conversion and verification on the data to generate standardized guidance price structured data.

[0059] S3: Read the tender list file through the table file analysis module, extract key information, and perform automatic currency conversion when the pricing currency is found to be inconsistent with the target currency, generating structured list data that is compatible with the guidance price data format.

[0060] S4: The large language model matching module calls the large language model that has been fine-tuned with corpus in the construction field. Based on multi-level matching rules, it intelligently matches the structured data of the list with the structured data of the guidance price, and outputs the matching results and similarity analysis report.

[0061] S5: The optimization and review module automatically verifies the compliance of the matching results, generates optimization suggestions, and submits them for manual review and confirmation through a visual interface.

[0062] S6: Provides a permission-based real-time collaboration environment for multi-role users through the Internet collaboration and sharing module, and encrypts and stores the approved data and operation logs to the cloud server.

[0063] S7: The one-click quotation generation module automatically calculates the total quotation based on the final confirmed unit price and quantity, and generates a bill of quantities file with price that conforms to national standards.

[0064] Preferably, the configuration sub-menu has a parameter configuration interface for setting data acquisition sources, multilingual recognition preferences, currency conversion benchmarks, and data matching accuracy thresholds. It provides a visual configuration interface divided into four independent sections: "Data Source Configuration," "Multilingual Recognition Configuration," "Currency Conversion Configuration," and "Matching Accuracy Configuration." Each section has a parameter description pop-up window. Data source configuration allows users to select either "Enterprise Integrated Management System" or "Local Database." When selecting "Enterprise Integrated Management System," users enter the system access address, an authenticated account with access permissions to cost data, and the file storage path. When selecting "Local Database," users configure the database IP address, port number, and target data table name. Multilingual recognition configuration supports target language selection and automatically associates with the built-in multilingual terminology database for the construction industry. It also supports users uploading custom terminology lists. Currency conversion configuration supports target currency selection and exchange rate update frequency settings. Matching accuracy configuration allows setting a similarity threshold via a slider, ranging from 0% to 100% (default 85%), used to determine the validity of the matching results from the large language model. The configuration submenu items, through an integrated parameter configuration interface, allow users to flexibly set the data source, language, currency, and matching precision, enabling the system to be personalized to meet the needs of different projects, thus improving the system's versatility and ease of use for users.

[0065] Preferably, the multi-source data acquisition unit is equipped with a first acquisition interface for downloading guidance price files from the enterprise integrated management system via application programming interface (API) calls, and a second acquisition interface for reading guidance price data from a local database via structured query language (SCL). When the data source is the "enterprise integrated management system," a file download request is sent via a RESTful API interface, supporting breakpoint resumption. When the data source is the "local database," structured guidance price data is directly read by generating and executing SQL query statements, supporting filtering by "project type" and "update time." By providing standardized data interfaces for the enterprise management system and the local database, the multi-source data acquisition unit achieves unified and automated acquisition of structured and unstructured cost guidance price data, effectively reducing manual intervention and time costs in the data preparation stage.

[0066] Preferably, the multilingual parsing and structuring unit integrates optical character recognition (OCR) technology and a multilingual natural language processing (NLP) model. This unit extracts project name, project features, and unit of measurement fields from unstructured documents and converts them into structured data based on a preset data model. For PDF-format price guidance documents, it uses OCR technology combined with a multilingual NLP model to extract text and identify key fields such as project name and project features. For Excel-format documents, it automatically identifies table headers and data rows and extracts corresponding field information. The extracted information is then converted into JSON-format structured data according to a preset data model. The preset data model fields include project_id, project_name, project_feature, unit, unit_price, currency_type, and applicable_project_type. A rule engine with built-in construction industry data validation rules is used to validate the structured data and generate an anomaly report. This multilingual parsing and structuring unit, combining OCR and NLP technologies, can automatically and accurately extract and structure key cost information from files of different formats. This solves the problems of low efficiency and easy omissions in manual processing of unstructured documents, providing a high-quality data foundation for subsequent intelligent matching.

[0067] Preferably, the large language model construction unit is equipped with a model training architecture that uses low-rank adaptation technology to fine-tune the general large language model. This architecture is based on a corpus of construction industry cost estimation corpus containing the "Construction Engineering Quantity List Pricing Specification", enterprise historical matching cases, and multilingual project feature description corpus. The implementation method of the large language model construction unit is as follows: collect relevant corpus of construction industry cost estimation to construct a corpus, including the "Construction Engineering Quantity List Pricing Specification", enterprise historical matching cases, and multilingual project feature description corpus, with a total corpus size of not less than 100,000 entries; fine-tune the general large language model using LoRA technology; and deploy the fine-tuned model to a GPU server using TensorRT acceleration technology. The large language model construction unit adopts fine-tuning technology based on a construction industry-specific corpus, enabling the general model to have the ability to deeply understand engineering cost professional terminology, specifications, and differences in multilingual expressions, providing industry knowledge support for the core matching task and improving the reliability and professionalism of semantic matching.

[0068] Preferably, the intelligent matching unit is configured with multi-level matching logic that includes matching weights for project name and unit of measurement, as well as matching weights for project feature details. This unit calculates the total similarity between the tender list items and the guidance price items based on the semantic understanding capability of a large language model. The implementation of the intelligent matching unit is as follows: it has a built-in "multi-level matching logic" that first matches "project name + unit of measurement" (weight 60%), and then matches "project feature details" (weight 40%), supporting user-defined weight allocation; it calculates similarity based on a large language model, for example, the similarity between "C30 commercial concrete pouring" and "pre-mixed C30 concrete pouring" is calculated to be 95%; if the similarity reaches the accuracy threshold configured by the user, it outputs the matching result and the corresponding guidance price unit price; otherwise, it outputs the recommended items with the highest similarity ranking and a similarity analysis report, selects the top 3-5 guidance price items with the highest similarity ranking as recommended items, and generates a "similarity analysis report" (annotating the difference fields, such as "the difference between '20mm rebar diameter' in the project feature and the recommended item '18mm'"). The intelligent matching unit simulates the comprehensive judgment process of professionals through multi-level matching logic with configurable weights. It can more comprehensively and evenly consider the semantic similarity of project names, units and feature details, thus providing more accurate and reasonable matching suggestions when facing real business scenarios with diverse expressions.

[0069] Preferably, the collaborative work platform includes a permission management component that assigns data operation permissions based on user roles, a real-time collaboration component that uses WebSocket technology to achieve data synchronization, and a message notification component for pushing audit reminders and anomaly alerts. The implementation of the collaborative work platform is as follows: it supports setting operation permissions according to the roles of "project manager," "cost engineer," and "auditor"; it uses WebSocket technology to achieve real-time data synchronization among multiple users; and it pushes audit reminders and anomaly alerts through system messages or emails. By integrating role permission management, real-time data synchronization, and message notification mechanisms, the collaborative work platform constructs a working environment that supports online collaboration among multiple users, which helps to break down information silos, promote team collaboration, and ensure the timely progress and transparent management of key operation processes.

[0070] By integrating a large language model and fully automated processing, the system achieves intelligent matching between the bill of quantities and project cost guidance prices, significantly shortening the traditional time-consuming and lengthy manual comparison cycle and greatly improving the overall efficiency of cost estimation. Secondly, based on a finely tuned large language model using a corpus specific to the construction industry, combined with a multi-level matching mechanism for project names, units of measurement, and feature details, it can accurately understand the differences in professional terminology and multilingual expressions, effectively improving the accuracy and reliability of quotations. Simultaneously, the system's built-in multilingual recognition and automatic currency conversion functions can directly process tender documents in multiple languages ​​and currencies, eliminating the errors and delays caused by manual translation and conversion in traditional models, and expanding the system's applicability in overseas engineering projects. Furthermore, by integrating industry-standard verification rules, full-process operation log recording, and cloud version management, the system not only ensures the compliance of quotation data but also achieves full-process traceability, strongly supporting project auditing and risk management. Finally, relying on an internet collaboration platform and role-based access control, the system supports real-time multi-user collaboration and cross-regional data sharing, promoting team collaboration efficiency and adapting to the decentralized and collaborative work scenarios of the construction industry.

[0071] The above description is merely a preferred embodiment of the present invention and does not limit the scope of the present invention. All equivalent changes made based on the description and drawings of the present invention are included within the scope of the present invention.

Claims

1. An intelligent system for calculating construction cost based on a large language model, characterized in that: It includes a custom submenu module, a data acquisition and parsing module, a spreadsheet file analysis module, a large language model matching module, an optimization and review module, an internet collaboration and sharing module, and a one-click quote generation module; The custom submenu module is connected to the data acquisition and parsing module, the table file analysis module, and the large language model matching module, respectively, and is used to pass configuration parameters to the modules. The data acquisition and parsing module and the tabular file analysis module are respectively connected to the large language model matching module, and are used to provide the large language model matching module with standardized project cost data and tender list data; The large language model matching module is connected to the optimization and review module and is used to output preliminary matching results to the optimization and review module. The optimization and review module is connected to the one-click quote generation module and is used to transmit the reviewed and confirmed quote data to the one-click quote generation module. The Internet collaboration and sharing module is connected to the data acquisition and parsing module, the table file analysis module, the optimization and review module, and the one-click quotation generation module, respectively, and is used to provide the collaborative operation environment and data storage management services for the modules. The custom submenu module includes configuration submenu items with a visual configuration interface and smart input submenu items that provide process triggering functionality; The data acquisition and parsing module includes a multi-source data acquisition unit for acquiring project cost management guidance price documents from multiple sources based on user configuration, and a multi-language parsing and structuring unit for performing multi-language parsing and structuring processing on the acquired documents. The table file analysis module includes a file format adaptation unit and a key information extraction unit for adapting to various formats of tender list files and extracting key information, as well as a currency conversion unit for performing automatic currency conversion; The large language model matching module includes a large language model construction unit fine-tuned based on the construction industry cost calculation corpus and an intelligent matching unit for performing intelligent matching and outputting results; The optimization and review module includes an automatic optimization unit with a built-in library of industry standards and enterprise standard rules, and a manual review unit that provides a visual user interface. The Internet collaboration and sharing module includes a collaborative work platform that supports multi-role permission management and real-time collaboration, and a cloud storage and version management unit for realizing cloud storage and version management of data. The one-click quotation generation module includes a total price calculation unit for calculating the total project quotation and a bill of quantities generation and export unit for generating and exporting a bill of quantities with prices.

2. The intelligent construction cost calculation system based on a large language model as described in claim 1, characterized in that: The configuration submenu has a parameter configuration interface for setting data acquisition sources, multilingual recognition preferences, currency conversion benchmarks, and data matching accuracy thresholds.

3. The intelligent construction cost calculation system based on a large language model according to claim 1, characterized in that: The multi-source data acquisition unit is equipped with a first acquisition interface for downloading guidance price documents from the enterprise integrated management system via application programming interface calls, and a second acquisition interface for reading guidance price data from the local database via structured query language.

4. The intelligent construction cost calculation system based on a large language model according to claim 1, characterized in that: The multilingual parsing and structured processing unit integrates optical character recognition technology and a multilingual natural language processing model to extract project name, project features, and unit of measurement fields from unstructured documents and convert them into structured data based on a preset data model.

5. The intelligent construction cost calculation system based on a large language model according to claim 1, characterized in that: The large language model building unit is equipped with a model training architecture that uses low-rank adaptation technology to fine-tune the general large language model. This architecture is based on a construction industry cost calculation corpus that includes construction engineering quantity list pricing specifications, enterprise historical matching cases, and multilingual project feature description corpus.

6. The intelligent construction cost calculation system based on a large language model according to claim 1, characterized in that: The intelligent matching unit is configured with multi-level matching logic that includes matching weights for project name and unit of measurement, as well as matching weights for project feature details. This unit calculates the total similarity between the tender list projects and the guidance price projects based on the semantic understanding capabilities of the large language model.

7. The intelligent construction cost calculation system based on a large language model according to claim 1, characterized in that: The collaborative work platform is equipped with a permission management component that assigns data operation permissions based on user roles, a real-time collaboration component that uses WebSocket technology to achieve data synchronization, and a message notification component for pushing audit reminders and anomaly alerts.

8. An intelligent method for calculating construction cost based on a large language model, characterized in that, The intelligent construction cost calculation system based on a large language model, as described in any one of claims 1 to 7, is used, and the specific steps are as follows: S1: Receives user-configured system parameters and process triggering commands through a custom submenu module; S2: Automatically acquire project cost management guidance price data through the data acquisition and parsing module, and perform multilingual parsing, structured conversion and verification on the data to generate standardized guidance price structured data; S3: Read the tender list file through the table file analysis module, extract key information, and perform automatic currency conversion when the pricing currency is found to be inconsistent with the target currency, generating structured list data that is compatible with the guidance price data format; S4: The large language model matching module calls the large language model that has been fine-tuned with the corpus in the construction field. Based on the multi-level matching rules, it intelligently matches the structured data of the list with the structured data of the guidance price and outputs the matching results and similarity analysis report. S5: The matching results are automatically verified for compliance with standards through the optimization and review module, optimization suggestions are generated, and manual review and confirmation are submitted through a visual interface. S6: Provides a permission-based real-time collaboration environment for multi-role users through the Internet collaboration and sharing module, and encrypts and stores the reviewed and confirmed data and operation logs to the cloud server; S7: The one-click quotation generation module automatically calculates the total quotation based on the final confirmed unit price and quantity, and generates a bill of quantities file with price that conforms to national standards.

9. The intelligent method for calculating construction cost based on a large language model as described in claim 8, characterized in that: In step S2, the multilingual parsing and structuring of the project cost management guidance document specifically includes: for PDF format files, using optical character recognition technology and a multilingual natural language processing model to extract text and identify key fields; for Excel format files, automatically identifying table headers and data rows and extracting field information; and converting the extracted information into JSON format structured data according to a preset data model.

10. The intelligent method for calculating construction cost based on a large language model as described in claim 8, characterized in that: In step S4, intelligent matching based on multi-level matching rules specifically includes: first matching the project name and unit of measurement to obtain the first similarity, then matching the project feature details to obtain the second similarity, and weighting the first similarity and the second similarity according to preset weights to obtain the total similarity; if the total similarity is not lower than the precision threshold configured by the user, the unit price of the successfully matched project is output; otherwise, several recommended items with the highest similarity ranking and a difference analysis report are output.

Images