A compliance risk control and payment management system and method for the engineering construction industry

Abstract

The application belongs to the field of computer software and engineering information technology, and particularly relates to a compliance risk control and payment management system and method for the engineering construction industry, which comprises: an engineering document intelligent analysis module, which adopts a constructed multi-modal AI model to identify and structure four-tuple information; an engineering project structured modeling module, which constructs a multi-level regional tree model, and associates each leaf node in the tree model with a unit price, a weight and a payment trigger condition agreed in a project contract; a mapping engine, which stores and manages payment rules, and supports parsing natural language clauses into executable logic rules; a dynamic progress payment calculation module, which traverses the multi-level regional tree model, and automatically accumulates corresponding unit prices and engineering quantities; an intelligent early warning and risk control module, which collects multi-dimensional process data of each approval link in real time, and dynamically determines an overdue probability through a pre-constructed risk scoring model. The application can realize high-precision semantic analysis, accurate mapping, dynamic pricing and intelligent risk control.

Description

Technical Field

[0001] This invention relates to the field of computer software and engineering information technology, specifically to a compliance risk control and payment management system and method for the engineering construction industry. Background Technology

[0002] In current payment management practices in the engineering construction field, mainstream information management systems are mainly built on a model based on electronic forms, workflow engines, and manual review. Typical management systems usually include the following technical components:

[0003] The form filling module allows contractors to fill out payment applications online and manually enter structured information such as contract number, application amount, and corresponding project milestones.

[0004] The document attachment upload module supports uploading unstructured files such as project progress reports, site photos, supervisor's signature sheets, and invoices;

[0005] The business process management engine can be configured with multi-level approval processes, such as project department, cost department, finance and owner, to realize online process flow;

[0006] Relational databases are used to store structured data such as contract terms, historical payment records, and approval logs;

[0007] The basic reminder function sends notifications via email or in-site message at preset times.

[0008] Currently, some systems are attempting to introduce general artificial intelligence technologies to improve efficiency; for example, using optical character recognition technology to automatically extract fields such as amount and date from standardized invoices.

[0009] Rule engines are used to validate simple conditions (such as "block payments if cumulative payment ratio exceeds 90%)." These technologies are already maturely applied in finance, e-commerce, and other fields, and are gradually penetrating into engineering project management software. However, their application in the specific scenario of engineering payments remains at a superficial automation stage. That is, although the above system has achieved the digitalization of the payment process, due to its technical architecture not being deeply adapted to the business characteristics of the engineering industry, the following fundamental problems caused by limitations in technical principles and structure still exist:

[0010] 1. Unstructured content with engineering semantics cannot be parsed. Engineering progress reports usually contain mixed text and images (such as "C Zone 3rd Floor Mechanical and Electrical Pipeline Laying Completed" with on-site photos). Existing OCR or NLP models are only applicable to standard documents or fixed templates and lack the ability to jointly identify engineering semantic units such as region, profession, process and completion status. Key information still needs to be extracted manually.

[0011] 2. Lack of a structured modeling mechanism for engineering projects. In actual projects, construction is organized according to dimensions such as contract sections, buildings, floors, and specialties. However, the existing system has not established such a spatial-logical hierarchical model, and cannot accurately map the actual completed content with the payment milestones agreed in the contract (such as "±0.00 30% payment completed").

[0012] 3. Progress payment calculation relies on manual intervention. Since the system cannot automatically link the completed work quantity with the contract unit price list, the progress payment amount still needs to be manually calculated and entered by cost estimators. This makes it difficult to achieve dynamic and fine-grained pricing based on the actual scope of completion, resulting in low efficiency and a high risk of errors.

[0013] 4. The early warning mechanism lacks context awareness. The existing reminder function is only triggered based on a fixed time offset and does not integrate multi-dimensional data such as the statutory response period of the project contract, historical approval time, and current backlog. It cannot dynamically assess the risk of overdue, resulting in delayed early warnings or a high false alarm rate.

[0014] 5. Fragmented information among multiple parties, severe separation of roles and permissions among general contractors, subcontractors, owners, and supervisors, lack of a unified dynamic view of funds, inconsistent understanding of payment status among parties, and easy to cause disputes.

[0015] In summary, although general AI and process automation technologies have been introduced into the field of engineering management, a payment management system that can deeply integrate engineering semantic understanding, regional structured modeling, dynamic pricing, and intelligent risk control has yet to emerge. Existing technological solutions still cannot meet the urgent needs of modern engineering construction for efficient, transparent, and compliant payments in terms of accuracy, automation, and collaborative efficiency. Summary of the Invention

[0016] In view of the technical defects mentioned in the background art, the purpose of this invention is to provide a compliance risk control and payment management system and method for the engineering construction industry, aiming to at least solve one of the technical problems in the related art to a certain extent.

[0017] To achieve the above objectives, in a first aspect, embodiments of the present invention provide a compliance risk control and payment management system for the engineering construction industry, the system comprising:

[0018] The intelligent engineering document parsing module is used to receive unstructured engineering files uploaded by users, and uses a pre-built multimodal AI model to identify and output structured information including a quadruple of region, specialty, process and completion status; wherein, the multimodal AI model includes a fusion of optical character recognition, natural language processing and computer vision.

[0019] The project structured modeling module is used to build a multi-level regional tree model based on the project contract and drawings, and associate each leaf node in the tree model with the unit price, weight and payment triggering conditions stipulated in the project contract.

[0020] The mapping engine is used to store and manage payment rules in project contracts. It supports parsing natural language terms into executable logical rules and binding them to nodes in the region tree model. The model nodes are leaf nodes in the region tree model. Each leaf node corresponds to an independent minimum pricing unit and is associated with the unit price, weight, and payment triggering conditions stipulated in the project contract.

[0021] The dynamic progress payment calculation module is used to traverse the multi-level region tree model based on the list of completed regions output by the intelligent parsing module of the project document, automatically accumulate the corresponding unit price and quantity of work, generate the amount payable for this payment, and generate detailed calculation basis for audit traceability and verification; wherein, the list of completed regions is derived from the completion status information in the quadruple;

[0022] The intelligent early warning and risk control module is used to collect multi-dimensional process data of each approval stage in real time, and dynamically judge the probability of delinquency through a pre-built risk scoring model. When the threshold is exceeded, multi-level early warnings are triggered according to preset strategies.

[0023] As a preferred implementation of this application, the system further includes a multi-party collaborative funds view module, which provides a unified data dashboard to display the payment application status, paid / unpaid amount, approval progress and warning information within the scope of their authority to different roles, so as to ensure that all parties have a consistent understanding of the dynamics of funds.

[0024] As a specific implementation of this application, the multi-level region tree model is constructed by organizing the project, section, building, floor and profession hierarchically;

[0025] The specific steps to associate each leaf node in the tree model with the unit price, weight, and payment trigger conditions stipulated in the project contract are as follows:

[0026] During the system initialization phase, the Bill of Quantities (BQ) and payment terms in the project contract are parsed, and the construction area, professional category, and contract unit price corresponding to each smallest pricing unit are extracted. Based on the construction drawings and the component attributes and spatial topology relationships in the pre-built corresponding BIM model, the physical space division structure is determined, and tree nodes for projects, sections, buildings, floors, and professions are generated level by level. The BIM model provides the geometric location, floor, professional category, and unique identifier at the component level for calibrating the area division.

[0027] During the operation phase, the actual construction completion status is obtained by identifying the collected panoramic images of the site; the regional location and professional type in the identification results are matched with the leaf nodes of the regional tree model; when the physical area corresponding to a certain leaf node is identified by the panoramic image as having completed the specified process and meeting the acceptance conditions agreed in the contract, it is considered to trigger the payment conditions of that node; the BQ unit price and quantity of work bound to that node are then used for subsequent pricing calculations.

[0028] As a specific implementation of this application, the step of parsing natural language terms into executable logical rules and binding them to region tree model nodes specifically includes:

[0029] A semantic parser based on domain dictionary enhancement is used to segment and extract keywords from the contract payment terms, identifying the subject of the condition, unit of measurement, payment ratio or amount, and preconditions.

[0030] The above elements are converted into structured rule expressions, expressed as IF[region plus process plus completion degree] AND [acceptance status] THEN trigger payment [percentage / amount]; wherein, the real-time determination of [region plus process plus completion degree] depends on the AI ​​recognition results of the on-site panoramic image;

[0031] The identified component completion status is mapped to the corresponding node in the region tree and used as the input variable of the rule engine; the rule expression is finally bound to the corresponding pricing unit node in BQ to ensure that the payment logic is consistent with the contract pricing system.

[0032] As a specific implementation of this application, the step of traversing the multi-level region tree model, automatically accumulating the corresponding unit price and quantity of work, and generating the amount payable for this transaction specifically includes:

[0033] Based on the panoramic image recognition results output by the multimodal AI model, the status of completed areas and their professional processes is extracted to form a structured list of quadruplets.

[0034] For each quadruple, locate the corresponding leaf node in the region tree model and verify whether it meets the bound payment triggering condition;

[0035] If satisfied, the contract unit price, unit of measurement, and quantity coefficient are obtained from the BQ entry associated with this node, and the amount payable for this node is calculated in combination with the actual completion ratio.

[0036] Summarize all leaf node payables that meet the conditions, and generate a total payable amount and itemized details table. The details table includes node path, BQ code, unit price, quantity, calculation formula, source panoramic image timestamp, and identification confidence level.

[0037] As a specific implementation of this application, a dynamic risk scoring model is constructed by integrating multi-dimensional data such as the statutory deadline of the project contract, historical approval time, and current process load. The dynamic determination of the delinquency probability through the pre-constructed risk scoring model specifically includes:

[0038] Real-time data collection includes the current approval stage of the payment application, the statutory response deadline stipulated in the contract, the historical average processing time for this stage, and the number of pending work orders.

[0039] Simultaneously, the panoramic image recognition record on which the payment application is based is linked. If the recognition confidence level is lower than the threshold or there is a disputed mark, the risk score weight is automatically increased; the remaining available time window is calculated as: statutory deadline - current date.

[0040] Based on the probability distribution of processing time fitted from historical data, and combined with the current backlog of work orders for load correction, the expected processing time of this application in this stage is predicted.

[0041] If the predicted time exceeds the remaining available time window, a risk of overdue payment is identified. Further, a comprehensive risk score is calculated based on the extent of the overdue payment and the reliability of the panoramic data. When the score exceeds a preset threshold, a warning action of the corresponding level is triggered.

[0042] Secondly, embodiments of the present invention also provide a compliance risk control and payment management method for the engineering construction industry, applied to the compliance risk control and payment management system for the engineering construction industry described in the first aspect, the method comprising the following steps:

[0043] The system receives unstructured engineering files uploaded by users, uses a pre-built multimodal AI model to identify and output structured information including a quadruple of region, specialty, process, and completion status; wherein, the multimodal AI model includes fused optical character recognition, natural language processing, and computer vision.

[0044] Based on the project contract and drawings, construct a multi-level regional tree model, and associate each leaf node in the tree model with the unit price, weight and payment triggering conditions stipulated in the project contract;

[0045] Store and manage payment rules in project contracts, support parsing natural language terms into executable logical rules, and bind them to nodes in the region tree model; wherein, the model nodes are leaf nodes in the region tree model, each leaf node corresponds to an independent minimum pricing unit, and is associated with the unit price, weight and payment triggering conditions stipulated in the project contract;

[0046] Based on the output list of completed areas, the multi-level area tree model is traversed, the corresponding unit price and quantity of work are automatically accumulated, the amount payable for this period is generated, and detailed calculation basis is generated for audit traceability and verification; wherein, the list of completed areas is derived from the completion status information in the quadruple;

[0047] The system collects multi-dimensional process data from each approval stage in real time and dynamically judges the probability of delinquency through a pre-built risk scoring model. When the threshold is exceeded, multi-level warnings are triggered according to preset strategies.

[0048] The beneficial effects of this invention: By deeply integrating multimodal artificial intelligence, structured modeling of engineering projects, and dynamic rule calculation into the payment management process, this invention has the following significant technical effects and advantages compared to existing technologies:

[0049] 1. Achieve high-precision semantic parsing of unstructured engineering documents: By integrating a multimodal AI model that combines optical character recognition, natural language processing and computer vision, the system can accurately identify the four-tuple information of region-profession-process-completion status in complex documents such as progress reports and on-site photos. The accuracy of structured extraction is improved, and manual input errors and omissions are greatly reduced.

[0050] 2. Establish a precise mapping mechanism between the physical space of the project and the contract terms: Based on a multi-level region tree model, the system can automatically associate the actual construction progress with the contract payment nodes, solving the problem of the disconnect between the completed content and the payment conditions caused by the lack of engineering semantic modeling in traditional systems, and ensuring that the basis for payment is objective, verifiable and more accurate.

[0051] 3. Supports fine-grained and dynamic automatic calculation of progress payments: The system can match the contract unit price in real time according to the smallest completed project unit, realizing automated pricing as each unit is completed, eliminating the need for manual calculation by cost estimators, improving calculation efficiency, and ensuring that the results are fully traceable and auditable.

[0052] 4. The early warning mechanism has context awareness and risk prediction capabilities: By integrating multi-dimensional data such as the legal term of the contract, historical approval time and current process load to build a dynamic risk scoring model, the system can predict the risk of overdue payment in advance and effectively avoid compliance default and cash flow crisis.

[0053] 5. Achieve real-time collaboration and transparency of fund information among multiple parties: All participants can view payment status based on the same data platform, eliminating information silos and cognitive biases, reducing disputes caused by poor communication, and improving overall collaboration efficiency.

[0054] This invention not only solves the long-standing technical problems in the field of engineering payment, such as low automation, poor accuracy, and difficulty in collaboration, but also achieves a leap from process digitization to intelligent decision-making automation through technological innovation, providing reliable technical support for the digital transformation of the engineering construction industry. Attached Figure Description

[0055] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below.

[0056] Figure 1 This is a schematic diagram of a compliance risk control and payment management system for the engineering construction industry provided in an embodiment of the present invention;

[0057] Figure 2 This is a schematic model of the processing procedure of a structured modeling module for engineering projects provided in an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram illustrating the process of parsing natural language terms into executable logical rules and binding them to nodes of a region tree model, as provided in an embodiment of the present invention.

[0059] Figure 4 This is a flowchart of a compliance risk control and payment management method for the engineering construction industry provided by an embodiment of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be understood that when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0061] BIM model (Building Information Modeling) is a digital tool used in engineering applications. It can be understood as a 3D model with rich building information.

[0062] BQ: Bill of Quantities.

[0063] It should be noted that, unless otherwise stated, the technical terms used in this embodiment have the common meaning as understood in the relevant technical field.

[0064] Please refer to Figure 1 This invention provides a compliance risk control and payment management system for the engineering construction industry, the system comprising:

[0065] The intelligent engineering document parsing module is used to receive unstructured engineering files uploaded by users, and uses a pre-built multimodal AI model to identify and output structured information including a quadruple of region, specialty, process and completion status; wherein, the multimodal AI model includes a fusion of optical character recognition, natural language processing and computer vision.

[0066] The project structured modeling module is used to build a multi-level regional tree model based on the project contract and drawings, and associate each leaf node in the tree model with the unit price, weight and payment triggering conditions stipulated in the project contract.

[0067] The mapping engine is used to store and manage payment rules in project contracts. It supports parsing natural language terms into executable logical rules and binding them to nodes in the region tree model. The model nodes are leaf nodes in the region tree model. Each leaf node corresponds to an independent minimum pricing unit and is associated with the unit price, weight, and payment triggering conditions stipulated in the project contract.

[0068] The dynamic progress payment calculation module is used to traverse the multi-level region tree model based on the list of completed regions output by the intelligent parsing module of the project document, automatically accumulate the corresponding unit price and quantity of work, generate the amount payable for this payment, and generate detailed calculation basis for audit traceability and verification; wherein, the list of completed regions is derived from the completion status information in the quadruple;

[0069] The intelligent early warning and risk control module is used to collect multi-dimensional process data of each approval stage in real time, and dynamically judge the probability of delinquency through a pre-built risk scoring model. When the threshold is exceeded, multi-level early warnings are triggered according to preset strategies.

[0070] In this embodiment, by integrating a multimodal AI model that combines optical character recognition (OCR), natural language processing (NLP), and computer vision (CV), the system can accurately identify the "region-specialty-process-completion status" quadruple information in complex documents such as progress reports, on-site photos, and on-site panoramic views.

[0071] In application, the necessary modules of the multimodal AI model include an image text extraction module, an image intelligent recognition module, and a multi-source information fusion module;

[0072] The image text extraction module is used to extract visible text information (such as device labels and serial numbers) from panoramic images or accompanying documents to assist in subsequent component matching.

[0073] The image intelligent recognition module is used for:

[0074] Deep learning-based target detection technology, after training, can automatically locate and identify various engineering components in panoramic images;

[0075] Each identified component must be labeled with its type (e.g., "galvanized steel cable tray - 200×100mm"), and this type must strictly correspond to the pricing item in BQ;

[0076] The output is the number of components of a certain type that were successfully identified within the specified area.

[0077] The multi-source information fusion module is used to cross-validate the components identified in the image with the text extraction results, thereby improving the accuracy and uniqueness of the identification.

[0078] Furthermore, based on the above technical solution, the system also includes a multi-party collaborative funds view module. This module provides a unified data dashboard to display the payment application status, paid / unpaid amounts, approval progress, and warning information within the scope of their authority to different roles, ensuring that all parties have a consistent understanding of the dynamics of funds.

[0079] The working principle of this system is as follows:

[0080] After users upload project progress-related documents (such as progress reports, site photos, acceptance forms, etc.), the system first uses a multimodal artificial intelligence model (multimodal AI model) to identify and semantically analyze the unstructured content, extracting four-tuple information of key information such as construction area, professional type, process stage and completion status;

[0081] Subsequently, the system matches the extracted results with a pre-built structured region tree model of the engineering project. This model is organized hierarchically by project → section → building → floor → profession, and is associated with the payment terms and unit price list in the contract.

[0082] Based on the matching results, the system automatically calculates the amount of progress payment due and verifies whether the payment conditions stipulated in the contract are met.

[0083] Meanwhile, the system monitors the processing time of each stage in real time, and dynamically assesses the risk of overdue payments by combining the statutory response period of the contract with historical approval data, triggering tiered early warnings.

[0084] Ultimately, all payment status, approval progress, and fund flow information are synchronized to a unified collaborative view, which can be viewed in real time by various parties such as general contractors, subcontractors, owners, and supervisors, avoiding disputes caused by information fragmentation.

[0085] Reference Figure 2 The multi-level regional tree model is constructed by organizing the data according to the levels of project, section, building, floor, and profession.

[0086] The specific steps for associating each leaf node in the tree model with the unit price, weight, and payment trigger conditions stipulated in the project contract are as follows:

[0087] During the system initialization phase, the Bill of Quantities (BQ) and payment terms in the project contract are parsed, and the construction area, professional category, and contract unit price corresponding to each smallest pricing unit are extracted. Based on the construction drawings and the component attributes and spatial topology relationships in the pre-built corresponding BIM model, the physical space division structure is determined, and tree nodes for projects, sections, buildings, floors, and professions are generated level by level. The BIM model provides the geometric location, floor, professional category, and unique identifier at the component level for calibrating the area division.

[0088] During the operation phase, the actual construction completion status is obtained by identifying the collected panoramic images of the site; the regional location and professional type in the identification results are matched with the leaf nodes of the regional tree model; when the physical area corresponding to a certain leaf node is identified by the panoramic image as having completed the specified process and meeting the acceptance conditions agreed in the contract, it is considered to trigger the payment conditions of that node; the BQ unit price and quantity of work bound to that node are then used for subsequent pricing calculations.

[0089] In this embodiment, the multi-level region tree model serves as a unified index hub for BIM, BQ, and site identification results, used to query the total number of theoretical components corresponding to a certain BQ item in a certain region; the key structure and connections of this model specifically include:

[0090] 1. Leaf node: Represents a combination of (region, profession), such as the 3rd floor of Building A1 - Mechanical and Electrical Engineering;

[0091] 2. Leaf node attributes:

[0092] Link the BQ Item list to list all pricing items involved in this area; such as ITM-2056: Galvanized cable tray installation;

[0093] Link the BIM component list, extract it from the BIM model, and for each BQ Item, list all component instance IDs and quantities that should be installed in this area; for example: ITM-2056 → corresponds to 15 cable tray segments in the BIM; therefore, the theoretical total is 15.

[0094] When the multimodal AI model is fed with {region:"Building A1, 3rd Floor", BQ_item_id:"ITM-2056", detected_count:12}:

[0095] The region tree locates the leaf node A1, 3rd floor - Mechanical and Electrical;

[0096] Query the total number of theoretical components corresponding to ITM-2056 under this node; from BIM=15.

[0097] Calculate the percentage completed: 12 / 15 = 80%;

[0098] Write 80% of the progress status to the BQ Item for subsequent pricing.

[0099] Reference Figure 3 The process of parsing natural language terms into executable logical rules and binding them to region tree model nodes specifically includes:

[0100] A semantic parser based on domain dictionary enhancement is used to segment and extract keywords from the contract payment terms, identifying the subject of the condition, unit of measurement, payment ratio or amount, and preconditions.

[0101] The above elements are converted into structured rule expressions, expressed as IF[region plus process plus completion degree] AND [acceptance status] THEN trigger payment [percentage / amount]; wherein, the real-time determination of [region plus process plus completion degree] depends on the AI ​​recognition results of the on-site panoramic image;

[0102] The identified component completion status is mapped to the corresponding node in the region tree and used as the input variable of the rule engine; the rule expression is finally bound to the corresponding pricing unit node in BQ to ensure that the payment logic is consistent with the contract pricing system.

[0103] Furthermore, the step of traversing the multi-level region tree model, automatically accumulating the corresponding unit price and quantity of work, and generating the amount due for this payment specifically includes:

[0104] Based on the panoramic image recognition results output by the multimodal AI model, the status of completed areas and their professional processes is extracted to form a structured list of quadruplets.

[0105] For each quadruple, locate the corresponding leaf node in the region tree model and verify whether it meets the bound payment triggering condition;

[0106] If satisfied, the contract unit price, unit of measurement, and quantity coefficient are obtained from the BQ entry associated with this node, and the amount payable for this node is calculated in combination with the actual completion ratio.

[0107] Summarize all leaf node payables that meet the conditions, and generate a total payable amount and itemized details table. The details table includes node path, BQ code, unit price, quantity, calculation formula, source panoramic image timestamp, and identification confidence level.

[0108] In this embodiment, a dynamic risk scoring model is constructed by integrating multi-dimensional data such as the statutory deadline of the project contract, historical approval time, and current process load. The dynamic determination of the probability of delinquency using the pre-constructed risk scoring model specifically includes:

[0109] The system collects real-time data on the current approval stage of a payment application, the statutory response deadline stipulated in the contract, the historical average processing time for that stage, and the number of pending work orders. The historical average processing time is calculated based on the average processing time for that stage over the past three months.

[0110] At the same time, the panoramic image recognition record on which the payment application is based will be linked. If the recognition confidence is lower than the threshold or there is a dispute mark, the risk score weight will be automatically increased.

[0111] Calculate the remaining available time window = statutory deadline - current date;

[0112] Based on the probability distribution of processing time fitted from historical data, and combined with the current backlog of work orders for load correction, the expected processing time of this application in this stage is predicted.

[0113] If the predicted time exceeds the remaining available time window, a risk of overdue payment is identified. A comprehensive risk score is then calculated based on the extent of the delay and the reliability of the panoramic data. When the score exceeds a preset threshold, a corresponding level of warning action is triggered. Multiple preset thresholds are included, corresponding to low, medium, and high risks. Preset strategies include triggering corresponding level of warning actions when the score exceeds the preset threshold, including triggering multi-level warnings such as in-site messages, SMS messages, and large-screen notifications.

[0114] During implementation, the reasonableness and reliability of the completion percentage based on the BQItem level in the dynamic risk scoring model, whereby this percentage is derived from the number of identified items / total number of BIMs, specifically includes:

[0115] 1. Reasonableness of completion rate: If the completion rate suddenly increases to 100%, but the adjacent areas are still at 0%, it may be abnormal; if the completion rate is more than 30% higher than the planned progress, it triggers the risk of false reporting.

[0116] 2. Recognition confidence level: The average confidence level of the image intelligent recognition model for the detected components (if <0.7, it is unreliable); the proportion of components whose numbers were not matched by OCR is too high;

[0117] 3. BIM consistency: Does the identified component type exist in the BIM? If the AI ​​identifies the component even though it does not exist in the BIM, it is a high-risk situation.

[0118] To facilitate a better understanding of this solution, a specific embodiment will be used as an example.

[0119] Example Background

[0120] A real estate development company launched the XX Garden residential project, comprising two sections (Section A and Section B). Section A includes two high-rise residential buildings (Building A1 and Building A2), each 30 stories high, with a shear wall structure. The contract stipulates the following payment schedule:

[0121] Upon completion of ±0.00, 10% of the total contract price will be paid;

[0122] 3% payment for every 5 standard structural layers completed;

[0123] Payment will be made up to 70% of the total contract price after the topping-out period.

[0124] The structured modeling module for engineering projects constructs a region tree:

[0125] xx Garden → Section A → Building A1 → [1st floor, 2nd floor, ..., 30th floor] → [Civil Engineering];

[0126] Each floor node is associated with a contract unit price (e.g., single-floor structural cost = 800,000 yuan) and payment rules (3% payment is triggered every 5 floors).

[0127] Specific work process

[0128] Upload progress file

[0129] On October 10, 20xx, the general contractor uploaded a PDF document titled "Progress Report for Building A1 in October". The document contained the text: "As of October 9, the main structure of Building A1 from floors 1 to 15 has been completed." It also included 15 on-site photos showing that the floor slab of the 15th floor had been poured.

[0130] Intelligent parsing and semantic extraction

[0131] The intelligent parsing module for engineering documents receives the file and performs the following operations:

[0132] Use OCR to extract text content;

[0133] The key phrase "The main structure of layers 1-15 is complete" was identified using an NLP model.

[0134] The computer vision model was used to analyze the photos, confirming that the 15-layer structure was completed.

[0135] Regional matching and payment terms verification

[0136] The system inputs the above results into the structured modeling module of the project, traverses the floor nodes of Building A1, and marks floors 1-15 as "completed".

[0137] Contract terms - payment node mapping engine automatically identifies:

[0138] Three complete 5-layer units have been completed (1–5, 6–10, 11–15).

[0139] If the "pay 3% every 5 levels" rule is triggered 3 times, the cumulative additional payment ratio should be 9%.

[0140] Dynamic calculation of progress payments

[0141] The dynamic progress payment calculation module queries the total contract price (e.g., 100 million yuan) and calculates the current payment due:

[0142] 100 million yuan × 9% = 9 million yuan;

[0143] Simultaneously generate detailed data:

[0144] "Building A1, floors 1-5: 800,000 x 5 = 4,000,000";

[0145] "Building A1, floors 6-10: 4 million";

[0146] "Building A1, floors 11-15: 4 million";

[0147] Total: 12 million → The system automatically detected duplicate calculations and corrected it to 9 million according to the rule of "3% payment for every 5 levels".

[0148] Intelligent early warning assessment

[0149] The system records that this application was submitted on October 10, 20xx. According to the contract, "the owner must respond within 30 days of receiving the application," and the statutory deadline is November 9, 20xx.

[0150] Intelligent early warning and risk control module

[0151] Based on historical data (the average approval time for homeowners is 22 days), the current risk is considered low, and no warning will be issued for now; if the issue remains unresolved by November 5th, it will be automatically upgraded to high risk, and a text message reminder will be sent to the homeowner representative.

[0152] Multi-party collaborative view update

[0153] All data is synchronized to the multi-party collaborative funds view module:

[0154] The general contractor can see: "Application submitted, amount 9 million, status: awaiting owner's review";

[0155] The owner can see: "New payment application, A1 Building, floors 1-15, 9 million, deadline: November 9th";

[0156] The supervisor can see: "The corresponding progress has been accepted, and approval is recommended."

[0157] The entire process requires no manual input of quantities or calculation of amounts; the system completes the entire process from file upload to collaborative view update within 5 minutes.

[0158] It should be noted that the relevant training process steps for each model should be understood by those skilled in the art, and will not be elaborated here.

[0159] This invention provides a compliance risk control and payment management system for the engineering construction industry. Its core lies in the deep integration of artificial intelligence, structured modeling and rule engine to achieve automatic parsing of payment applications, dynamic calculation of progress payments and collaborative monitoring of fund status. Moreover, the modules work closely together through data interfaces to form a closed-loop processing flow.

[0160] The above solution, by deeply integrating multimodal artificial intelligence, structured modeling of engineering projects, and dynamic rule calculation into the payment management process, has the following significant technical effects and advantages compared to existing technologies:

[0161] 1. Achieve high-precision semantic parsing of unstructured engineering documents: By integrating a multimodal AI model that combines optical character recognition, natural language processing and computer vision, the system can accurately identify the four-tuple information of region-profession-process-completion status in complex documents such as progress reports and on-site photos. The accuracy of structured extraction is improved, and manual input errors and omissions are greatly reduced.

[0162] 2. Establish a precise mapping mechanism between the physical space of the project and the contract terms: Based on a multi-level region tree model, the system can automatically associate the actual construction progress with the contract payment nodes, solving the problem of the disconnect between the completed content and the payment conditions caused by the lack of engineering semantic modeling in traditional systems, and ensuring that the basis for payment is objective, verifiable and more accurate.

[0163] 3. Supports fine-grained and dynamic automatic calculation of progress payments: The system can match the contract unit price in real time according to the smallest completed project unit, realizing automated pricing as each unit is completed, eliminating the need for manual calculation by cost estimators, improving calculation efficiency, and ensuring that the results are fully traceable and auditable.

[0164] 4. The early warning mechanism has context awareness and risk prediction capabilities: By integrating multi-dimensional data such as the legal term of the contract, historical approval time and current process load to build a dynamic risk scoring model, the system can predict the risk of overdue payment in advance and effectively avoid compliance default and cash flow crisis.

[0165] 5. Achieve real-time collaboration and transparency of fund information among multiple parties: All participants can view payment status based on the same data platform, eliminating information silos and cognitive biases, reducing disputes caused by poor communication, and improving overall collaboration efficiency.

[0166] This invention not only solves the long-standing technical problems in the field of engineering payment, such as low automation, poor accuracy, and difficulty in collaboration, but also achieves a leap from process digitization to intelligent decision-making automation through technological innovation, providing reliable technical support for the digital transformation of the engineering construction industry.

[0167] Based on the same inventive concept, this invention also provides a compliance risk control and payment management method for the engineering construction industry, applied to the compliance risk control and payment management system for the engineering construction industry described in the first aspect, with reference to... Figure 4 The method includes the following steps:

[0168] S101, receive unstructured engineering files uploaded by users, and use a pre-built multimodal AI model to identify and output structured information including a quadruple of region, specialty, process and completion status; wherein, the multimodal AI model includes fused optical character recognition, natural language processing and computer vision;

[0169] S102, Based on the project contract and drawings, construct a multi-level regional tree model, and associate each leaf node in the tree model with the unit price, weight and payment triggering conditions stipulated in the project contract;

[0170] S103, store and manage payment rules in project contracts, support parsing natural language terms into executable logical rules, and bind them to nodes in the region tree model; wherein, the model nodes are leaf nodes in the region tree model, each leaf node corresponds to an independent minimum pricing unit, and is associated with the unit price, weight and payment triggering conditions stipulated in the project contract.

[0171] S104, based on the output list of completed areas, traverse the multi-level area tree model, automatically accumulate the corresponding unit price and quantity of work, generate the amount payable for this payment, and generate detailed calculation basis for audit traceability and verification; wherein, the list of completed areas comes from the completion status information in the quadruple;

[0172] S105 collects multi-dimensional process data from each approval stage in real time and dynamically judges the probability of delinquency through a pre-built risk scoring model. When the threshold is exceeded, multi-level warnings are triggered according to preset strategies.

[0173] Furthermore, the method also includes:

[0174] Provide a unified data dashboard to display the payment application status, paid / unpaid amount, approval progress and warning information within the scope of their authority to different roles, so as to ensure that all parties have a consistent understanding of the dynamics of funds.

[0175] The multi-level regional tree model is constructed by organizing the data according to the levels of project, section, building, floor, and profession.

[0176] The specific steps for associating each leaf node in the tree model with the unit price, weight, and payment trigger conditions stipulated in the project contract are as follows:

[0177] During the system initialization phase, the Bill of Quantities (BQ) and payment terms in the project contract are parsed, and the construction area, professional category, and contract unit price corresponding to each smallest pricing unit are extracted. Based on the construction drawings and the component attributes and spatial topology relationships in the pre-built corresponding BIM model, the physical space division structure is determined, and tree nodes for projects, sections, buildings, floors, and professions are generated level by level. The BIM model provides the geometric location, floor, professional category, and unique identifier at the component level for calibrating the area division.

[0178] During the operation phase, the actual construction completion status is obtained by identifying the collected panoramic images of the site; the regional location and professional type in the identification results are matched with the leaf nodes of the regional tree model; when the physical area corresponding to a certain leaf node is identified by the panoramic image as having completed the specified process and meeting the acceptance conditions agreed in the contract, it is considered to trigger the payment conditions of that node; the BQ unit price and quantity of work bound to that node are then used for subsequent pricing calculations.

[0179] In this embodiment, the step of parsing natural language terms into executable logical rules and binding them to region tree model nodes specifically includes:

[0180] A semantic parser based on domain dictionary enhancement is used to segment and extract keywords from the contract payment terms, identifying the subject of the condition, unit of measurement, payment ratio or amount, and preconditions.

[0181] The above elements are converted into structured rule expressions, expressed as IF[region plus process plus completion degree] AND [acceptance status] THEN trigger payment [percentage / amount]; wherein, the real-time determination of [region plus process plus completion degree] depends on the AI ​​recognition results of the on-site panoramic image;

[0182] The identified component completion status is mapped to the corresponding node in the region tree and used as the input variable of the rule engine; the rule expression is finally bound to the corresponding pricing unit node in BQ to ensure that the payment logic is consistent with the contract pricing system; it supports the automatic association of physical construction progress (such as the topping out of the 5th floor of building x) to the payment node agreed in the contract, such as paying XX yuan for each standard floor completed.

[0183] The process of traversing the multi-level region tree model, automatically accumulating the corresponding unit price and quantity of work, and generating the amount due for this payment specifically includes:

[0184] Based on the panoramic image recognition results output by the multimodal AI model, the status of completed areas and their professional processes is extracted to form a structured list of quadruplets.

[0185] For each quadruple, locate the corresponding leaf node in the region tree model and verify whether it meets the bound payment triggering condition;

[0186] If satisfied, the contract unit price, unit of measurement, and quantity coefficient are obtained from the BQ entry associated with this node, and the amount payable for this node is calculated in combination with the actual completion ratio.

[0187] The system aggregates all eligible leaf node payments and generates a total payment amount and a detailed breakdown table. The breakdown table includes node path, BQ code, unit price, quantity, calculation formula, source panoramic image timestamp, and identification confidence level. This allows for the dynamic generation of accurate payment amounts based on the identified completed areas and their corresponding contract unit price lists without manual intervention.

[0188] Furthermore, a dynamic risk scoring model is constructed by integrating multi-dimensional data such as the legal term of the project contract, historical approval time, and current process load. The dynamic determination of the probability of delinquency through this pre-constructed risk scoring model specifically includes:

[0189] Real-time data collection includes the current approval stage of the payment application, the statutory response deadline stipulated in the contract, the historical average processing time for this stage, and the number of pending work orders.

[0190] At the same time, the panoramic image recognition record on which the payment application is based will be linked. If the recognition confidence is lower than the threshold or there is a dispute mark, the risk score weight will be automatically increased.

[0191] Calculate the remaining available time window = statutory deadline - current date;

[0192] Based on the probability distribution of processing time fitted from historical data, and combined with the current backlog of work orders for load correction, the expected processing time of this application in this stage is predicted.

[0193] If the predicted time exceeds the remaining available time window, a risk of overdue payment is identified. Further, a comprehensive risk score is calculated based on the extent of the overdue payment and the reliability of the panoramic data. When the score exceeds a preset threshold, a warning action of the corresponding level is triggered.

[0194] By comprehensively considering multi-dimensional data such as the statutory response period for contracts, historical approval time, and current backlog status, the risk of overdue payments is dynamically assessed and tiered warnings are triggered, rather than relying on static reminders with fixed time offsets.

[0195] It should be noted that for a more detailed description of the workflow of the method embodiments, please refer to the aforementioned system embodiments section, which will not be repeated here.

[0196] Achieve high-precision semantic parsing of unstructured engineering documents: By integrating optical character recognition, natural language processing and computer vision into a multimodal AI model, the system can accurately identify the four-tuple information of region-profession-process-completion status in complex documents such as progress reports and on-site photos, thereby improving the accuracy of structured extraction and significantly reducing manual input errors and omissions.

[0197] Establish a precise mapping mechanism between the physical space of the project and the contract terms: Based on a multi-level region tree model, the system can automatically associate the actual construction progress with the contract payment nodes, solving the problem of the disconnect between the completed content and the payment conditions caused by the lack of engineering semantic modeling in traditional systems, and ensuring that the basis for payment is objective, verifiable and more accurate;

[0198] Supports fine-grained and dynamic automatic calculation of progress payments: The system can match the contract unit price in real time according to the smallest completed engineering unit, realize automated pricing as one unit is completed, eliminate the need for cost estimators to manually calculate, improve calculation efficiency, and the results are fully traceable and auditable.

[0199] The early warning mechanism has context awareness and risk prediction capabilities: by integrating multi-dimensional data such as the legal term of the contract, historical approval time and current process load to build a dynamic risk scoring model, the system can predict the risk of overdue payment in advance and effectively avoid compliance default and cash flow crisis.

[0200] Achieve real-time collaboration and transparency of fund information among multiple parties: All participants can view payment status based on the same data platform, eliminating information silos and cognitive biases, reducing disputes caused by poor communication, and improving overall collaboration efficiency.

[0201] This invention not only solves the long-standing technical problems in the field of engineering payment, such as low automation, poor accuracy, and difficulty in collaboration, but also achieves a leap from process digitization to intelligent decision-making automation through technological innovation, providing reliable technical support for the digital transformation of the engineering construction industry.

[0202] Another embodiment of the present invention differs from the previous two embodiments in that:

[0203] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0204] In the embodiments provided in this application, it should be understood that the disclosed methods and systems can also be implemented in other ways. The system embodiments described above are merely illustrative. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the figures. For example, two consecutive blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or using a combination of dedicated hardware and computer instructions.

[0205] Furthermore, the functional modules in the various embodiments of this invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. When using each module, information is collected and stored only with the full authorization of the relevant user or organization and in compliance with relevant laws and regulations, and the security and privacy of the data are protected. Unauthorized access is strictly prohibited. Data processing will be carried out within the scope stipulated by law and will not exceed the authorized purpose and scope. At the same time, the authorizing party has the right to access, correct, delete, restrict processing, refuse, etc., of its personal data, and strictly comply with applicable laws and regulations and conduct compliance reviews.

[0206] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions should all be covered within the scope of protection of the present invention.

Claims

1. A compliance risk control and payment management system for the engineering construction industry, characterized in that, The system includes: The intelligent engineering document parsing module is used to receive unstructured engineering files uploaded by users, and uses a pre-built multimodal AI model to identify and output structured information including a quadruple of region, specialty, process and completion status; wherein, the multimodal AI model includes a fusion of optical character recognition, natural language processing and computer vision. The project structured modeling module is used to build a multi-level regional tree model based on the project contract and drawings, and associate each leaf node in the tree model with the unit price, weight and payment triggering conditions stipulated in the project contract. The mapping engine is used to store and manage payment rules in project contracts. It supports parsing natural language terms into executable logical rules and binding them to nodes in the region tree model. The model nodes are leaf nodes in the region tree model. Each leaf node corresponds to an independent minimum pricing unit and is associated with the unit price, weight, and payment triggering conditions stipulated in the project contract. The dynamic progress payment calculation module is used to traverse the multi-level region tree model based on the list of completed regions output by the intelligent parsing module of the project document, automatically accumulate the corresponding unit price and quantity of work, generate the amount payable for this payment, and generate detailed calculation basis for audit traceability and verification; wherein, the list of completed regions is derived from the completion status information in the quadruple; The intelligent early warning and risk control module is used to collect multi-dimensional process data of each approval stage in real time, and dynamically judge the probability of delinquency through a pre-built risk scoring model. When the threshold is exceeded, multi-level early warnings are triggered according to preset strategies.

2. The system as described in claim 1, characterized in that, It also includes a multi-party collaborative funds view module, which provides a unified data dashboard to display the payment application status, paid / unpaid amount, approval progress and warning information within the scope of their authority to different roles, so as to ensure that all parties have a consistent understanding of the dynamics of funds.

3. The system as described in claim 1, characterized in that, The multi-level regional tree model is constructed by organizing the data according to the levels of project, section, building, floor, and profession. The specific steps for associating each leaf node in the tree model with the unit price, weight, and payment trigger conditions stipulated in the project contract are as follows: During the system initialization phase, the Bill of Quantities (BQ) and payment terms in the project contract are parsed, and the construction area, professional category, and contract unit price corresponding to each smallest pricing unit are extracted. Based on the construction drawings and the component attributes and spatial topology relationships in the pre-built corresponding BIM model, the physical space division structure is determined, and tree nodes for projects, sections, buildings, floors, and professions are generated level by level. The BIM model provides the geometric location, floor, professional category, and unique identifier at the component level for calibrating the area division. During the operation phase, the actual construction completion status is obtained by identifying the collected panoramic images of the site; the regional location and professional type in the identification results are matched with the leaf nodes of the regional tree model; when the physical area corresponding to a certain leaf node is identified by the panoramic image as having completed the specified process and meeting the acceptance conditions agreed in the contract, it is considered to trigger the payment conditions of that node; the BQ unit price and quantity of work bound to that node are then used for subsequent pricing calculations.

4. The system as described in any one of claims 1 to 3, characterized in that, The process of parsing natural language terms into executable logical rules and binding them to region tree model nodes specifically includes: A semantic parser based on domain dictionary enhancement is used to segment and extract keywords from the contract payment terms, identifying the subject of the condition, unit of measurement, payment ratio or amount, and preconditions. The above elements are converted into structured rule expressions, expressed as IF[region plus process plus completion degree] AND [acceptance status] THEN trigger payment [percentage / amount]; wherein, the real-time determination of [region plus process plus completion degree] depends on the AI ​​recognition results of the on-site panoramic image; The identified component completion status is mapped to the corresponding node in the region tree and used as the input variable of the rule engine; the rule expression is finally bound to the corresponding pricing unit node in BQ to ensure that the payment logic is consistent with the contract pricing system.

5. The system as described in claim 4, characterized in that, The process of traversing the multi-level region tree model, automatically accumulating the corresponding unit price and quantity of work, and generating the amount due for this payment specifically includes: Based on the panoramic image recognition results output by the multimodal AI model, the status of completed areas and their professional processes is extracted to form a structured list of quadruplets. For each quadruple, locate the corresponding leaf node in the region tree model and verify whether it meets the bound payment triggering condition; If satisfied, the contract unit price, unit of measurement, and quantity coefficient are obtained from the BQ entry associated with this node, and the amount payable for this node is calculated in combination with the actual completion ratio. Summarize all leaf node payables that meet the conditions, and generate a total payable amount and itemized details table. The details table includes node path, BQ code, unit price, quantity, calculation formula, source panoramic image timestamp, and identification confidence level.

6. The system as described in claim 5, characterized in that, A dynamic risk scoring model is constructed by integrating multi-dimensional data such as the legal term of project contracts, historical approval duration, and current process load. The dynamic determination of the probability of delinquency using this pre-built risk scoring model specifically includes: Real-time data collection includes the current approval stage of the payment application, the statutory response deadline stipulated in the contract, the historical average processing time for this stage, and the number of pending work orders. At the same time, the panoramic image recognition record on which the payment application is based will be linked. If the recognition confidence is lower than the threshold or there is a dispute mark, the risk score weight will be automatically increased. Calculate the remaining available time window = statutory deadline - current date; Based on the probability distribution of processing time fitted from historical data, and combined with the current backlog of work orders for load correction, the expected processing time of this application in this stage is predicted. If the predicted time exceeds the remaining available time window, a risk of overdue payment is identified. Further, a comprehensive risk score is calculated based on the extent of the overdue payment and the reliability of the panoramic data. When the score exceeds a preset threshold, a warning action of the corresponding level is triggered.

7. A compliance risk control and payment management method for the engineering construction industry, characterized in that, The method applied to the compliance risk control and payment management system for the engineering construction industry as described in claim 1 includes the following steps: The system receives unstructured engineering files uploaded by users, uses a pre-built multimodal AI model to identify and output structured information including a quadruple of region, specialty, process, and completion status; wherein, the multimodal AI model includes fused optical character recognition, natural language processing, and computer vision. Based on the project contract and drawings, construct a multi-level regional tree model, and associate each leaf node in the tree model with the unit price, weight and payment triggering conditions stipulated in the project contract; Store and manage payment rules in project contracts, support parsing natural language terms into executable logical rules, and bind them to nodes in the region tree model; wherein, the model nodes are leaf nodes in the region tree model, each leaf node corresponds to an independent minimum pricing unit, and is associated with the unit price, weight and payment triggering conditions stipulated in the project contract; Based on the output list of completed areas, the multi-level area tree model is traversed, the corresponding unit price and quantity of work are automatically accumulated, the amount payable for this period is generated, and detailed calculation basis is generated for audit traceability and verification; wherein, the list of completed areas is derived from the completion status information in the quadruple; The system collects multi-dimensional process data from each approval stage in real time and dynamically judges the probability of delinquency through a pre-built risk scoring model. When the threshold is exceeded, multi-level warnings are triggered according to preset strategies.

8. The method as described in claim 7, characterized in that, The method further includes: Provide a unified data dashboard to display the payment application status, paid / unpaid amount, approval progress and warning information within the scope of their authority to different roles, so as to ensure that all parties have a consistent understanding of the dynamics of funds.

9. The method as described in claim 7, characterized in that, The multi-level regional tree model is constructed by organizing the data according to the levels of project, section, building, floor, and profession. The specific steps for associating each leaf node in the tree model with the unit price, weight, and payment trigger conditions stipulated in the project contract are as follows: During the system initialization phase, the Bill of Quantities (BQ) and payment terms in the project contract are parsed, and the construction area, professional category, and contract unit price corresponding to each smallest pricing unit are extracted. Based on the construction drawings and the component attributes and spatial topology relationships in the pre-built corresponding BIM model, the physical space division structure is determined, and tree nodes for projects, sections, buildings, floors, and professions are generated level by level. The BIM model provides the geometric location, floor, professional category, and unique identifier at the component level for calibrating the area division. During the operation phase, the actual construction completion status is obtained by identifying the collected panoramic images of the site; the regional location and professional type in the identification results are matched with the leaf nodes of the regional tree model; when the physical area corresponding to a certain leaf node is identified by the panoramic image as having completed the specified process and meeting the acceptance conditions agreed in the contract, it is considered to trigger the payment conditions of that node; the BQ unit price and quantity of work bound to that node are then used for subsequent pricing calculations.

10. The method as described in claim 9, characterized in that, The process of parsing natural language terms into executable logical rules and binding them to region tree model nodes specifically includes: A semantic parser based on domain dictionary enhancement is used to segment and extract keywords from the contract payment terms, identifying the subject of the condition, unit of measurement, payment ratio or amount, and preconditions. The above elements are converted into structured rule expressions, expressed as IF[region plus process plus completion degree] AND [acceptance status] THEN trigger payment [percentage / amount]; wherein, the real-time determination of [region plus process plus completion degree] depends on the AI ​​recognition results of the on-site panoramic image; The identified component completion status is mapped to the corresponding node in the region tree and used as the input variable of the rule engine; the rule expression is finally bound to the corresponding pricing unit node in BQ to ensure that the payment logic is consistent with the contract pricing system.

Images