Methods, systems, equipment and media for revising and compiling tender documents for road and bridge engineering projects

Abstract

This invention provides a method, system, device, and medium for revising and compiling road and bridge engineering tender documents, belonging to the field of computer science. The method includes: receiving original road and bridge engineering tender documents and the tenderer's format requirements data, and synchronizing them to a road and bridge engineering scenario-specific format sub-library; extracting basic elements and identifying road and bridge industry-specific elements from the received original road and bridge engineering tender documents, outputting a parsing result containing industry element information; applying rules to correct the parsing result using road and bridge industry-specific formats, outputting a first preliminary correction result; using a machine learning model to correct the parsing result, obtaining a second preliminary correction result; merging the first and second preliminary correction results to obtain conflict-free data, and merging the conflict-free data according to the standard logical order of road and bridge engineering tender documents to generate the target road and bridge engineering tender document. The dual correction method for industry elements improves the bid compliance rate and shortens the revision cycle.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically to a method, system, equipment, and medium for revising and compiling tender documents for road and bridge engineering projects. Background Technology

[0002] In the procurement of goods, works and services, the process by which the tendering party attracts numerous bidders to compete on equal terms by publishing the procurement and requirements in advance, and then organizes experts in technology, economics and law to conduct a comprehensive review of the numerous bidders in accordance with the prescribed procedures, and selects the best bidder for the project.

[0003] The format requirements for special elements in the road and bridge industry are unique, and manual adjustments are far more difficult than for general documents: the bill of quantities in the tender documents must have a fixed column order of "bill number-project name-unit-quantity-unit price", the construction process flowchart must use industry-specific symbols such as "drilled pile" and "box girder casting", and the safety production special plan table must have a dedicated column of "risk level-prevention and control measures". When making manual adjustments, both the format specifications and industry terminology must be taken into account, which can easily lead to the problem of "the format is correct but the industry label is wrong".

[0004] The lag in responding to dynamic changes in the tendering party's format requirements can easily lead to missed bidding opportunities: Tendering parties for road and bridge projects often temporarily supplement the format requirements 3-5 days before the bid deadline (such as adding font specifications for the "Smart Construction Site" chapter). Traditional manual adjustments are required page by page, especially for bill of quantities tables, which need to be recalibrated in terms of column order. Due to insufficient time, format omissions are likely to occur.

[0005] Moreover, existing tools lack adaptability to the road and bridge industry and have insufficient special correction capabilities: general document processing tools (such as Word format painter) cannot identify industry-specific problems such as "incorrect order of bill of quantities" and "non-standard symbols in construction flowcharts", requiring manual verification one by one, which leads to road and bridge companies always facing the dilemma of "difficulty in format compliance and low compilation efficiency". Summary of the Invention

[0006] The purpose of this invention is to provide a method, system, equipment, and medium for revising and compiling tender documents for road and bridge engineering projects. It allows for customized revisions of industry elements such as bill of quantities and construction flowcharts, thereby improving the compliance rate of tenders. When the tendering party temporarily supplements the format requirements, the method can complete the rule update and document revision in a short time, significantly shortening the response cycle compared to manual adjustments.

[0007] To achieve the above objectives, embodiments of the present invention provide a method for revising and compiling tender documents for road and bridge engineering projects, including:

[0008] Receive the original tender documents and the format requirements data of the tendering party for road and bridge engineering projects, and synchronize the format requirements data of the tendering party to the sub-library of road and bridge engineering scenarios.

[0009] The system extracts basic elements and identifies industry-specific elements from the received original tender documents for road and bridge engineering projects, and outputs parsing results containing industry element information.

[0010] The rules of the road and bridge engineering scenario-specific format sub-library are called to correct the parsing results for road and bridge industry-specific formats, and the first preliminary correction result is output.

[0011] A machine learning model is used to correct the analysis results, resulting in a second preliminary corrected result;

[0012] By integrating the first preliminary correction result and the second preliminary correction result, conflict-free data is obtained, and the conflict-free data is merged according to the standard logical order of road and bridge engineering tender documents to generate the target tender document for road and bridge engineering.

[0013] Optionally, the rule of calling the sub-library of road and bridge engineering scenario formats is used to correct the parsing results according to the road and bridge industry-specific format, and output the first preliminary correction result, including:

[0014] Based on the analysis results, the location information of basic elements and road and bridge industry-specific elements in the original tender documents of road and bridge projects is determined;

[0015] Based on the type of basic elements and road and bridge industry-specific elements and the engineering scenario, the corresponding rules are called from the road and bridge engineering scenario-specific format sub-library;

[0016] Compare the actual format of basic elements and road and bridge industry-specific elements with the standard format according to the judgment logic in the rules, and output the compliance or non-compliance result and the type of non-compliance;

[0017] Select a correction scheme that matches the type of violation, correct the violation result, and obtain the first preliminary correction result.

[0018] Optionally, the step of using a machine learning model to correct the analysis results and obtain a second preliminary corrected result includes:

[0019] For the quantity table element type in the original tender documents of road and bridge projects, the Transformer model is selected to perform semantic recognition of table column names to avoid omissions due to different column name expressions in the sub-library of road and bridge project scenarios.

[0020] For the types of construction image elements in the original tender documents of road and bridge projects, a target detection model is selected to identify whether the images contain numbered watermarks and areas with insufficient resolution.

[0021] For the process flow chart element type in the original tender documents of road and bridge engineering: use graph neural networks to learn the logic of road and bridge construction processes and help identify process logic errors.

[0022] Optionally, for the quantity table element type in the original tender documents of road and bridge engineering, the Transformer model is selected to perform semantic recognition of table column names to avoid omissions due to different column name expressions in the sub-library of road and bridge engineering scenario formats, including:

[0023] For the input quantity table, use pandas to read the header row of the quantity table and get the column name text;

[0024] The obtained column name text is input into the Transformer model, which outputs the confidence scores of multiple categories and selects the category with the highest confidence score as the predicted category. The Transformer model has the following structure: input layer, BERT pre-trained layer (freezing the first 6 layers to retain the underlying semantic features), pooling layer (outputting a 768-dimensional vector representing the overall semantics of the column name), dropout layer, first fully connected layer (compressing the feature dimension), second fully connected layer (corresponding to 12 standard column name categories), and output layer (using the Softmax function to output the confidence score of each category).

[0025] After the Transformer model outputs the predicted category, it calls the road and bridge engineering scenario-specific format sub-library to perform column name correction. Specifically, if the Transformer model identifies the project code as a list number, and the standard format requirement for the list number in the road and bridge engineering scenario-specific format sub-library is 12-digit encoding, it checks whether the current project code format meets the 12-digit encoding requirement. If it does not meet the requirement, a correction prompt will automatically pop up, indicating that the column name project code belongs to the list number category and needs to be supplemented with a 12-digit encoding. If it does meet the requirement, it is marked as semantically compliant with column name, and proceeds to the subsequent table column order correction stage.

[0026] Optionally, for construction image elements in the original tender documents for road and bridge engineering projects, a target detection model is selected to identify whether the images contain numbered watermarks and areas with insufficient resolution, including:

[0027] The input construction images are normalized in size, standardized in pixels, and noise suppressed to obtain the target construction images.

[0028] The ResNet-50 network is used to extract deep visual features from the target construction image, and a sliding window is applied to the feature map to generate candidate regions containing the construction target.

[0029] For the generated candidate regions containing construction targets, first align the features, then classify them as either numbered watermarked or insufficient resolution regions and fine-tune the bounding boxes to adapt to the precise positioning requirements of the construction targets.

[0030] For the classification results, retain the target boxes whose category probability is greater than the preset probability threshold, and remove overlapping boxes using non-maximum suppression. Output the bounding boxes and categories of the numbered watermark and the region with insufficient resolution.

[0031] Optionally, for the process flow diagram element type in the original tender documents for road and bridge engineering: use graph neural networks to learn the logic of road and bridge construction processes and assist in identifying process logic errors, including:

[0032] Use OCR to identify the process text boxes in the process flow diagram and use them as the node set of the target diagram, and identify the arrow lines between the process text boxes and use them as the edge set of the target diagram.

[0033] By integrating the node set and edge set, the process flow diagram is transformed into structured data consisting of nodes and edges;

[0034] The structured data of nodes and edges is input into the input layer of the GAT graph neural network, and a linear transformation is performed on the node features to enhance the expression of node features.

[0035] Calculate the attention weight of a node to its neighboring nodes, and based on this attention weight, aggregate the features of the neighboring nodes and update the node's features;

[0036] The updated final features of all nodes are processed by global mean pooling to obtain the graph-level features of the entire process flow diagram. The graph-level features are then input into the fully connected layer to output the probability distribution of logical error types.

[0037] Optionally, the attention weights of a node to its neighboring nodes are calculated, and based on these attention weights, the features of the neighboring nodes are aggregated to update the node's features, including:

[0038] The formula for calculating the correlation of node pairs is as follows:

[0039] ;

[0040] In the formula, a represents the attention weight vector, and Z i Z j Representing nodes respectively and neighboring nodes Features after linear transformation Z represents i and Z j splicing;

[0041] The formula for normalizing attention weights is as follows:

[0042] ;

[0043] In the formula, N( i ) is a node v i The set of all neighboring nodes;

[0044] The feature update formula for a node is as follows:

[0045] ;

[0046] In the formula, This is the activation function.

[0047] Secondly, the present invention also provides a system for revising and compiling tender documents for road and bridge engineering projects, comprising:

[0048] The input module is used to receive the original tender documents and the format requirements of the tendering party for road and bridge engineering projects, and to synchronize the format requirements of the tendering party to the sub-library of road and bridge engineering scenarios.

[0049] The element parsing module is used to extract basic elements and identify industry-specific elements from the received original tender documents for road and bridge engineering projects, and output parsing results containing industry element information.

[0050] The correction module is used to call the rules of the road and bridge engineering scenario-specific format sub-library, correct the parsing results for road and bridge industry-specific formats, and output the first preliminary correction result; a machine learning model is used to correct the parsing results to obtain the second preliminary correction result.

[0051] The merge module is used to merge the first preliminary correction result and the second preliminary correction result to obtain conflict-free data, and merge the conflict-free data according to the standard logical order of road and bridge engineering tender documents to generate the target tender document for road and bridge engineering.

[0052] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method for revising and compiling road and bridge engineering tender documents.

[0053] Fourthly, the present invention also provides a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for revising and compiling road and bridge engineering tender documents.

[0054] The above technical solution enables the integrated processing of road and bridge engineering tender documents through "dynamic compliance + collaborative pre-resolution + special correction", ensuring that the documents meet industry standards and the dynamic requirements of the tendering party, shortening the preparation cycle by more than 60%, and supporting integration with the road and bridge enterprise tender document management system, with strong industry adaptability and practicality.

[0055] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0056] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0057] Figure 1 This is a flowchart illustrating the implementation of a method for revising and compiling tender documents for road and bridge engineering, as provided in an embodiment of the present invention.

[0058] Figure 2 This is a schematic diagram of the structure of a road and bridge engineering tender document revision and compilation system provided in an embodiment of the present invention;

[0059] Figure 3 This is an example diagram of a road and bridge engineering tender document revision and compilation system provided in an embodiment of the present invention;

[0060] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0061] Various embodiments of this disclosure will be described more fully in the following detailed description. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0062] In the following, the terms “comprising” or “may include”, which may be used in various embodiments of this disclosure, indicate the presence of the disclosed functions or operations and do not limit the addition of one or more functions or operations. Furthermore, as used in various embodiments of this disclosure, the terms “comprising,” “having,” and their cognates are intended only to indicate a specific feature, number, step, operation, or combination of the foregoing and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, or combinations of the foregoing, or the possibility of adding one or more features, numbers, steps, operations, or combinations of the foregoing.

[0063] In various embodiments of this disclosure, the expression "or" or "at least one of A and / or B" includes any combination or all combinations of the words listed simultaneously. For example, the expression "A or B" or "at least one of A and / or B" may include A, may include B, or may include both A and B.

[0064] 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, and 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.

[0065] See Figure 1 The diagram shows a flowchart of a method for revising and finalizing tender documents for road and bridge engineering projects in a specific embodiment, including the following execution steps:

[0066] Step 100: Receive the original tender documents and the tenderer's format requirements data for the road and bridge project, and synchronize the tenderer's format requirements data to the road and bridge project scenario-specific format sub-library.

[0067] Specifically, in addition to supporting Word / PDF, it adds import of road and bridge industry-specific formats (Excel, cost estimation software exported files) and automatically converts them into parsable formats; it also provides an "upload portal for bidding party format requirements" to support users to upload temporary supplementary format documents (such as a Word version of "Green Construction Chapter Requirements").

[0068] The tender documents for road and bridge projects include the official documents submitted by road and bridge companies when participating in bidding for projects such as highways, bridges, and tunnels. In addition to the basic format, they must meet the specific requirements of the road and bridge industry (such as the column order of the bill of quantities table and the symbols of the construction process flowchart), and cover core modules such as construction organization design, bill of quantities, and special safety production plan.

[0069] For example, the sub-libraries of road and bridge engineering scene formats are shown in Table 1:

[0070] Table 1. Format Sub-libraries for Road and Bridge Engineering Scenarios

[0071]

[0072] Step 101: Extract basic elements and identify industry-specific elements from the received original tender documents for road and bridge engineering projects, and output the parsing results containing industry element information.

[0073] Step 102: Call the rules of the road and bridge engineering scenario-specific format sub-library, correct the parsing results for the road and bridge industry-specific format, and output the first preliminary correction result.

[0074] Specifically, when performing step 102, the following steps can be performed:

[0075] S1020: Based on the analysis results, determine the location information of basic elements and road and bridge industry-specific elements in the original tender documents for road and bridge engineering.

[0076] S1021: Based on the type of basic elements and road and bridge industry-specific elements and the engineering scenario, call the corresponding rules from the road and bridge engineering scenario format sub-library.

[0077] Examples include: basic rules such as margins, headers and footers; industry rules such as: tables with fixed column order for bill of quantities and "safety production table" including a "risk level" column; images with construction image numbers in the format of "chapter number-list number-image name"; flowcharts with road and bridge-specific symbols for construction procedures and red dotted arrows (indicating construction process); and dynamic updates such as: supporting users to upload temporary requirements from the bidding party and completing rule entry and sub-database updates within 10 minutes.

[0078] S1022: Compare the actual format of basic elements and road and bridge industry-specific elements with the standard format according to the judgment logic in the rules, and output the compliance or non-compliance result and the type of non-compliance.

[0079] S1023: Select a correction scheme that matches the violation type, correct the violation result, and obtain the first preliminary correction result.

[0080] Specifically, basic corrections include: title numbering, page margins, etc.; industry-specific corrections include: tables: automatically calibrating the column order of the bill of quantities and supplementing missing "bill of quantities number" columns; images: generating industry standard numbers for construction images, automatically prompting "high-resolution drawings are needed" when resolution is insufficient; flowcharts: replacing non-industry symbols, such as changing "general arrows" to "construction procedure arrows" and adjusting the color scheme to industry standards (light gray background + black text); machine learning optimization: optimizing the accuracy of "bill of quantities column order calibration" and "flowchart symbol replacement" through historical road and bridge tender document cases.

[0081] For example, the parsing results are first called to determine the specific locations of "quantity tables, construction pictures, and flowcharts" in the tender documents; based on the element type and engineering scenario, the corresponding rule set is called from the rule base; the actual format of the element is compared with the standard format according to the "judgment logic" in the rule, and the "compliance / violation" result and violation type are output; for the violation items, the "correction plan" is executed, such as "incorrect column order" calling the table processing interface to adjust the column order, and "missing number" generating a standard number based on the chapter number and list number.

[0082] Step 103: Use a machine learning model to correct the analysis results and obtain a second preliminary correction result.

[0083] Specifically, when executing step 103, the following steps can be performed:

[0084] S1030: For the quantity table element type in the original tender documents of road and bridge projects, the Transformer model is selected to perform semantic recognition of table column names, so as to avoid omissions due to different column name expressions in the sub-library of road and bridge engineering scenario formats.

[0085] Specifically, when executing step S1030, the following sub-steps can be executed:

[0086] SA: For the input quantity table, use pandas to read the header row of the quantity table and get the column name text.

[0087] SB: Input the obtained column name text into the Transformer model, output the confidence scores of multiple categories, and select the category with the highest confidence score as the predicted category. The Transformer model has the following structure: input layer, BERT pre-trained layer, freeze the first 6 layers to retain the underlying semantic features, pooling layer, output 768-dimensional vector representing the overall semantics of the column name, dropout layer, first fully connected layer to compress the feature dimension, second fully connected layer corresponding to 12 standard column name categories, and output layer, using the Softmax function to output the confidence score of each category.

[0088] SC: After the Transformer model outputs the predicted category, it calls the road and bridge engineering scenario-specific format sub-library to perform column name correction. If the Transformer model identifies the project code as a list number, and the standard format requirement for the list number in the road and bridge engineering scenario-specific format sub-library is 12-digit encoding, it checks whether the current project code format meets the 12-digit encoding requirement. If it does not meet the requirement, a correction prompt will automatically pop up, indicating that the column name project code belongs to the list number category and needs to be supplemented with a 12-digit encoding. If it meets the requirement, it is marked as semantically compliant with column name and proceeds to the subsequent table column order correction stage.

[0089] In one specific implementation, let the text of the column name in the quantity table to be identified be x, and after word segmentation, a token sequence is obtained: x =[ t 1, t 2,..., t T ], where T is the sequence length. Initial features are obtained through BERT's embedding layers: The sequence is The L-layer Transformer encoder in BERT outputs context-aware features: ,in h t For the first t The contextual features of each token; the features used to aggregate the semantics of the entire sentence are taken as the overall semantic representation of the column name: Through the weighted connection layer Mapping to the standard class space, the output predicted probability is calculated by the fully connected layer: In the formula, Let c be the weight vector of class c. This is a bias. The class probabilities are obtained by normalizing using the Softmax function: in, Let x be the predicted probability that column name x belongs to category c. The quantification formula for misidentification is as follows:

[0090]

[0091] In the formula, p y For the real category y The predicted probability, Represents the true category y The one-hot encoding representation. The total loss function for the batch of samples is:

[0092]

[0093] In the formula, Let be the predicted probability of the true class of the i-th sample. Let the weight of the true class of the i-th sample be . This is the focusing factor.

[0094] S1031: For the construction image element type in the original tender documents of road and bridge projects, select the target detection model to identify whether the image contains numbered watermarks and areas with insufficient resolution.

[0095] Specifically, when executing step S1031, the following sub-steps can be executed:

[0096] Sa: Performs size normalization, pixel standardization, and noise suppression on the input construction image to obtain the target construction image.

[0097] Specifically, the input image is resized to a fixed size, and the scaling ratio is recorded for use in restoring the bounding box representation; the RGB three channels are standardized separately using the following formula: in, This represents the pixel value of the c-th channel, h-th row, and ω-th column of the original image. This represents the mean value of the channel on the training set of construction images. The standard deviation is denoted as .

[0098] Sb: The ResNet-50 network is used to extract deep visual features of the target construction image, and a sliding window is used on the feature map to generate candidate regions containing the construction target.

[0099] Specifically, ResNet-50 consists of 4 residual blocks, and the formula for each residual block is as follows: in, x Input feature maps for residual blocks. W 1, W2 represents the convolution kernel weights, BN is batch normalization, ReLU is the activation function, and shortcut ( x () is a shortcut connection. After passing through ResNet-50, the original image is downsampled to a feature map F of 16×16×2048.

[0100] Sc: For the generated candidate regions containing construction targets, first align the features, then classify them as either numbered watermarked or insufficient resolution regions and fine-tune the bounding boxes to adapt to the precise positioning requirements of the construction targets.

[0101] Specifically, at each pixel location in feature map F, anchor boxes of 5 scales and 3 aspect ratios are generated to cover small-sized watermarks and large-sized low-resolution regions, with a total number of anchor boxes N= HF × WF × K , K The number of anchor boxes per pixel. For each anchor box, the RPN outputs two results: a classification score and a feature map convolution. F Mapped to 2× Nanchor The vector is processed by the Sigmoid function to obtain the foreground confidence score of each anchor box. p rpn : in, For classification weights, For bias. Mapped to 4× using a 1×1 convolution. N anchor The vector is used to correct the anchor box position, making the candidate box closer to the true target. Regression parameters are calculated as follows: ,in, For regression weights, This is a bias. The class probabilities are obtained after applying the Softmax function: in, , For the weights of the fully connected layer, , This is for bias. The boundary fine-tuning correction formula is as follows:

[0102]

[0103] In the formula, The center coordinates, width, and height of the RPN candidate box. These are the coordinates of the final target bounding box.

[0104] Sd: For the classification results, retain the target boxes whose category probability is greater than the preset probability threshold, and remove overlapping boxes using non-maximum suppression. Output the bounding boxes and categories of the numbered watermark and the region with insufficient resolution.

[0105] S1032: For the process flow chart element type in the original tender documents of road and bridge engineering: use graph neural networks to learn the logic of road and bridge construction processes and help identify process logic errors.

[0106] Specifically, when executing step S1032, the following sub-steps can be executed:

[0107] S1: Use OCR to identify the process text boxes in the process flow diagram and use them as the node set of the target diagram, and identify the arrow connections between the process text boxes and use them as the edge set of the target diagram.

[0108] S2: Integrate the node set and edge set to obtain structured data that transforms the process flow diagram into node-edge data.

[0109] S3: Input the structured data of nodes and edges into the input layer of the GAT graph neural network, and perform a linear transformation on the node features to enhance the expression of node features.

[0110] Specifically, the linear transformation formula is as follows:

[0111] ,in, For learnable weight matrix, These are node features.

[0112] S4: Calculate the attention weight of a node to its neighboring nodes, and based on this attention weight, aggregate the features of the neighboring nodes and update the node's features.

[0113] Specifically, the formula for calculating the correlation of node pairs is as follows:

[0114] ;

[0115] In the formula, a represents the attention weight vector, and Z i Z j Representing nodes respectively and neighboring nodes Features after linear transformation Z represents i and Z j splicing;

[0116] The formula for normalizing attention weights is as follows:

[0117] ;

[0118] In the formula, N( i ) is a node v i The set of all neighboring nodes;

[0119] The feature update formula for a node is as follows:

[0120] ;

[0121] In the formula, This is the activation function.

[0122] S5: Perform global mean pooling on the updated final features of all nodes to obtain the graph-level features of the entire process flow diagram, and input the graph-level features into the fully connected layer to output the probability distribution of logical error types.

[0123] In practical applications, the final node features are either a concatenation or mean of the features from multiple heads:

[0124]

[0125] In the formula, K For the number of attention heads, Let be the attention weight for the k-th head, and || denote feature concatenation.

[0126] In some implementations, the graph-level misclassification loss formula is as follows:

[0127]

[0128] In the formula, For real labels, For category weights, To predict probabilities.

[0129] The node-level error localization loss function is as follows:

[0130]

[0131] In the formula, N abn This represents the number of abnormally associated nodes. y i For nodes v i The true label, To predict probabilities.

[0132] It should be noted that the model in this application does not replace the rule base, but rather serves as a "supplement and optimization" to address the limitations of the rule base: 1. Improve the recognition accuracy of ambiguous scenarios: For example, the rule base may have difficulty judging cases where "the table column names are not standard," but the model, through semantic understanding, can accurately identify cases where "the column names are not standard but are substantively compliant," avoiding misjudgment by the rule base as "missing columns"; 2. Optimize correction priority and adaptive adjustment: When multiple violations exist simultaneously, the model can learn the priority of "correcting the column order first, then verifying the data format" through historical amendment examples; at the same time, the model will count "cases that were still rejected by the tendering party after correction" and automatically provide feedback to the rule base with suggestions to "add regional rules," forming a closed loop of "correction-feedback-optimization."

[0133] Step 104: Merge the first preliminary correction result and the second preliminary correction result to obtain conflict-free data, and merge the conflict-free data according to the standard logical order of road and bridge engineering tender documents to generate the target tender document for road and bridge engineering.

[0134] Specifically, when multiple writers are editing, the rules of the sub-libraries for different scenarios are synchronized in real time; conflict warnings: such as "the cost group is using a table with a blue border (violation, should be black)" or "the safety group is missing a 'risk level' column," will immediately pop up a warning and suggest a correction plan; access control: only the "format administrator" can modify industry rules to avoid accidental operation by ordinary writers. Logical merging: sub-files are merged according to the standard order of road and bridge engineering tender documents (project overview → construction organization design → bill of quantities → safety production plan); secondary correction: the "cross-module format consistency" of the merged document is finally calibrated.

[0135] Collaboration Relationship: Receives conflict-free data from the collaborative conflict pre-resolution module, completes merging and secondary correction, and then transmits it to the output module.

[0136] In one specific implementation, the collaborative mechanism between the rule base and the machine learning model is as follows: They do not work independently, but rather collaborate through a process of "rule screening first, model fine-tuning, and result feedback" to ensure the comprehensiveness and accuracy of identification and correction: Step 1: Initial screening by the rule base: The rule base first processes "deterministic violations," quickly resolving 80% of basic formatting issues and reducing the model's computational load. Step 2: Fine-tuning by the model: For ambiguous scenarios that the rule base "cannot judge," the model is invoked for secondary identification, outputting accurate results. Step 3: Result feedback: "New violation types" identified by the model automatically generate "rule addition suggestions," which are then entered into the rule base after administrator confirmation. "High-frequency violation rules" from the rule base are input into the model as "key training samples," improving the model's speed in recognizing such issues.

[0137] In some implementations, the hardware environment for implementing this solution is as follows: PC: Recommended configuration includes Intel Core i7 CPU, 16GB RAM, and 500GB SSD, supporting single-parse of engineering drawing PDFs larger than 100MB; Server: Dual-socket Intel Xeon E5 processors, 64GB RAM, and RAID 5 array (2TB or more), supporting simultaneous collaborative editing by 50 people (e.g., "10 people in the construction team + 20 people in the cost team + 20 people in the safety team"); ​​Cloud: Deployed on Alibaba Cloud ECS (4 cores, 8GB RAM or more), supporting elastic scaling of computing resources during peak bidding periods (e.g., the end of a quarter). Software Environment: Operating System: PC supports Windows 10 / 11 (must be compatible with Glodon software), server supports Windows Server 2019; Supporting Software: JRE 1.8 and Python 3.9 are installed (for industry element recognition models), and a new Excel format conversion component and a road and bridge engineering OCR terminology recognition plugin are added (supporting accurate extraction of terms such as "list number" and "construction procedure"); Integration and Compatibility: Through API interface, it integrates with commonly used systems of road and bridge enterprises (tender management system, cost estimation software) to achieve seamless connection of the entire process of "cost estimation software export → format correction → tender system storage".

[0138] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0139] The beneficial effects achieved by this application are as follows:

[0140] 1. Addressing pain points in road and bridge industry-specific formats and improving bid compliance rates: Customized modifications to industry elements such as bill of quantities and construction flowcharts increase the compliance rate of bid document formats from 70% for manual processing to over 98%, avoiding the invalidation of bid qualifications due to industry format deviations.

[0141] 2. Dynamic response to tenderer requirements and shortened preparation cycle: When the tenderer temporarily supplements the format requirements, the rules are updated and the documents are revised within 10 minutes. Compared with the manual adjustment time of 2-3 days, the response cycle is greatly shortened, ensuring timely submission.

[0142] 3. Pre-resolve collaboration conflicts and reduce compilation costs: Real-time warning of format deviations during multi-group editing, no need for dedicated personnel to make unified corrections after compilation, reducing labor costs by 70% and shortening the compilation cycle from an average of 10 days to less than 4 days.

[0143] 4. Strong industry adaptability and expanded application scenarios: Supports importing files from road and bridge cost estimation software such as Glodon and Tongwang, and can be adapted to various engineering bidding scenarios such as highways, bridges, and tunnels. At the same time, it can be integrated with the company's existing tender document system. Its practicality and scalability far exceed those of general document processing tools.

[0144] like Figure 2 As shown, the following are embodiments of the road and bridge engineering tender document revision and compilation system provided in this disclosure. The revision and compilation methods of road and bridge engineering tender documents in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the road and bridge engineering tender document revision and compilation system, please refer to the embodiments of the above-mentioned road and bridge engineering tender document revision and compilation methods.

[0145] The system for revising and finalizing road and bridge engineering tender documents includes:

[0146] The input module is used to receive the original tender documents and the format requirements of the tendering party for road and bridge engineering projects, and to synchronize the format requirements of the tendering party to the sub-library of road and bridge engineering scenarios.

[0147] The element parsing module is used to extract basic elements and identify industry-specific elements from the received original tender documents for road and bridge engineering projects, and output parsing results containing industry element information.

[0148] The correction module is used to call the rules of the road and bridge engineering scenario-specific format sub-library, correct the parsing results for road and bridge industry-specific formats, and output the first preliminary correction result; a machine learning model is used to correct the parsing results to obtain the second preliminary correction result.

[0149] The merge module is used to merge the first preliminary correction result and the second preliminary correction result to obtain conflict-free data, and merge the conflict-free data according to the standard logical order of road and bridge engineering tender documents to generate the target tender document for road and bridge engineering.

[0150] In one example, see Figure 3The diagram shown is an example of a system for revising and compiling road and bridge engineering tender documents according to an embodiment of the present invention. The input module supports single file uploads and multiple sub-file uploads. Template selection: users can choose specific templates. Document parsing module: parses the uploaded single file or multiple sub-files. Format knowledge base: provides document format-related knowledge to guide the processing. Intelligent revision engine: intelligently revises documents using information from specification configurations and version libraries. Compilation module: merges multiple files or templates. Specification configuration and version library: provides specification configurations and version management for different tenderers / projects, supporting document processing and revision. Output module: exports the processed document in Word or PDF format. System integration interface: integrates with other systems via API or internal platform. Review and logging: provides revision rules and difference visualization for easy review and tracking of modification records.

[0151] Figure 4 This is a schematic diagram of the hardware structure of an electronic device that implements various embodiments of the present invention.

[0152] The method for revising and compiling road and bridge engineering tender documents provided in this application can be applied to electronic devices. Those skilled in the art will understand that the electronic device structure involved in the embodiments of this invention does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this invention, the electronic device includes, but is not limited to, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0153] Electronic devices may include processors, external memory interfaces, internal memory, universal serial bus (USB) interfaces, charging management modules, power management modules, batteries, wireless communication modules, audio modules, speakers, microphones, sensor modules, buttons, cameras, displays, and SIM card interfaces, etc.

[0154] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the electronic device. In other embodiments of this application, the electronic device may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0155] A processor may include one or more processing units, such as: a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors.

[0156] The processor can serve as the nerve center and command center of an electronic device. The controller can generate operation control signals based on the instruction opcode and timing signals to control the fetching and execution of instructions.

[0157] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from this memory. This avoids repeated accesses, reduces processor latency, and thus improves system efficiency.

[0158] An external storage interface (ESI) can be used to connect external memory cards, such as microSD cards, to expand the storage capacity of electronic devices. The external memory card communicates with the processor through the ESI to perform data storage functions, such as saving music and video files on the external memory card.

[0159] Internal memory can be used to store computer executable program code, which includes instructions. The processor executes various functional applications and data processing of electronic devices by running the instructions stored in internal memory. Internal memory can include a program storage area and a data storage area. Internal memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one disk storage device, flash memory device, universal flash storage (UFS), etc.

[0160] Wireless communication functionality in electronic devices can be achieved through antennas, wireless communication modules, modem processors, and baseband processors.

[0161] Wireless communication modules can provide solutions for wireless communication applications in electronic devices, including wireless local area networks (WLANs) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), and infrared (IR) technologies.

[0162] Electronic devices can implement audio functions through audio modules, speakers, receivers, microphones, headphone jacks, and application processors.

[0163] Electronic devices can achieve shooting functions through ISPs, cameras, video codecs, GPUs, displays, and application processors.

[0164] Electronic devices can achieve display functions through GPUs, displays, and application processors.

[0165] A GPU is a microprocessor for image processing, connected to the display screen and application processor. GPUs perform mathematical and geometric calculations for graphics rendering. A processor may include one or more GPUs, which execute program instructions to generate or modify display information.

[0166] A display screen is used to display images, videos, etc. A display screen includes a display panel.

[0167] The storage medium provided in this application stores a program product capable of implementing methods for revising and compiling tender documents for road and bridge engineering projects.

[0168] The method for revising and compiling road and bridge engineering tender documents includes: receiving the original tender documents and the tenderer's format requirements data, and synchronizing the tenderer's format requirements data to the road and bridge engineering scenario-specific format sub-library; extracting basic elements and identifying road and bridge industry-specific elements from the received original tender documents, and outputting parsing results containing industry element information; calling the rules of the road and bridge engineering scenario-specific format sub-library to revise the parsing results for road and bridge industry-specific formats, and outputting a first preliminary revision result; using a machine learning model to revise the parsing results to obtain a second preliminary revision result; merging the first preliminary revision result and the second preliminary revision result to obtain conflict-free data, and merging the conflict-free data according to the standard logical order of road and bridge engineering tender documents to generate the target tender document for road and bridge engineering.

[0169] In some possible implementations, the subject matter of this disclosure, namely, "Method and System for Revising and Compiling Bidding Documents for Road and Bridge Engineering," can be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure.

[0170] The storage medium disclosed herein may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0171] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for revising and merging a road and bridge engineering tender document, characterized in that, The method comprises the following steps: Receiving the original road and bridge engineering bidding file and the format requirement data of the tendering party, and synchronizing the format requirement data of the tendering party to the road and bridge engineering sub-scene format library; Extracting the basic elements from the received original road and bridge engineering bidding file and identifying the road and bridge industry-specific elements, and outputting the analysis result containing the industry element information; Calling the rules of the road and bridge engineering sub-scene format library to correct the analysis result, and outputting the first preliminary correction result; Using a machine learning model to correct the analysis result to obtain a second preliminary correction result; Fusing the first preliminary correction result and the second preliminary correction result to obtain conflict-free data, and merging the conflict-free data according to the logical order of the road and bridge engineering bidding file standard to generate the target road and bridge engineering bidding file; Wherein, the step of using a machine learning model to correct the analysis result to obtain a second preliminary correction result comprises: For the engineering quantity table element type in the original road and bridge engineering bidding file, a Transformer model is selected to perform semantic recognition on the table column name to avoid missing judgment due to different column name expressions in the road and bridge engineering sub-scene format library; For the construction picture element type in the original road and bridge engineering bidding file, a target detection model is selected to identify whether the picture contains a numbered watermark and a resolution insufficient area; For the process flowchart element type in the original road and bridge engineering bidding file: a graph neural network is used to learn the road and bridge construction process logic to assist in identifying the flow logic error.

2. The method of claim 1, wherein the method further comprises: The step of calling the rules of the road and bridge engineering sub-scene format library to correct the analysis result to output the first preliminary correction result comprises: Based on the analysis result, the position information of the basic elements and the road and bridge industry-specific elements in the original road and bridge engineering bidding file is determined; According to the types of the basic elements and the road and bridge industry-specific elements and the engineering scene, the corresponding rules are called from the road and bridge engineering sub-scene format library; According to the judgment logic in the rules, the actual format of the basic elements and the road and bridge industry-specific elements is compared with the standard format to output the compliance or violation result and the violation type; Selecting a correction scheme matched with the violation type to correct the violation result to obtain the first preliminary correction result.

3. The method of claim 1, wherein the method further comprises: For the engineering quantity table element type in the original road and bridge engineering bidding file, a Transformer model is selected to perform semantic recognition on the table column name to avoid missing judgment due to different column name expressions in the road and bridge engineering sub-scene format library, comprising: For the input engineering quantity table, the header row of the engineering quantity table is read by pandas to obtain the column name text; The obtained column name text is input into a Transformer model, and the confidence of multiple categories is output, and the category with the highest confidence is selected as the predicted category, wherein the structure of the Transformer model is an input layer, a BERT pre-training layer, the first 6 layers are frozen to retain the bottom semantic features, a pooling layer, an output of 768-dimensional vectors representing the overall semantics of the column name, a dropout layer, a first fully connected layer to compress the feature dimension, a second fully connected layer corresponding to 12 standard column name categories, and an output layer using a Softmax function to output the confidence of each category; After the Transformer model outputs the predicted category, a road and bridge engineering sub-scene format library is called to perform column name correction. If the Transformer model identifies the project code as a list number category, the standard format of the list number category in the road and bridge engineering sub-scene format library requires 12-bit coding, then the format of the current project code is checked to see if it meets the 12-bit coding requirement. If not, a correction prompt is automatically popped up that the column name project code belongs to the list number category and needs to be supplemented to 12-bit coding. If it meets the requirement, it is marked as column name semantic compliance and enters the subsequent table column sequence correction link.

4. The method of claim 1, wherein the method further comprises: For the construction picture element type in the original road and bridge engineering bidding document, a target detection model is selected to identify whether the picture contains a numbered watermark and a resolution insufficient area, including: The input construction picture is subjected to size normalization, pixel standardization, and noise suppression processing to obtain a target construction picture; ResNet-50 network is used to extract deep visual features of the target construction picture, and a sliding window is generated on the feature map to generate a candidate area containing the construction target; For the generated candidate area containing the construction target, the features are first aligned, and then classification and boundary box fine-tuning are performed to determine whether it is a numbered watermark or a resolution insufficient area to adapt to the precise positioning needs of the construction target; For the classification result, the target box with a class probability greater than a preset probability threshold is retained, and non-maximum suppression is used to remove overlapping boxes to output the boundary box and category of the numbered watermark and resolution insufficient area.

5. The method of claim 1, wherein the method further comprises: For the process flowchart element type in the original road and bridge engineering bidding document: a graph neural network is used to learn the road and bridge construction process logic to assist in identifying flow logic errors, including: OCR is used to identify the process text box in the process flowchart, and the process text box is used as the node set of the target graph, and the arrow connection between the process text boxes is identified and used as the edge set of the target graph; The node set and edge set are integrated to obtain structured data that converts the process flowchart into nodes-edges; The input layer of the GAT graph neural network is input with the structured data of nodes-edges, and the node features are linearly transformed to enhance the node feature expression; The attention weight of the node to its neighbor nodes is calculated, and the features of the neighbor nodes are aggregated according to the attention weight to update the features of the nodes; The updated final features of all nodes are subjected to global mean pooling processing to obtain the graph-level features of the entire process flowchart, and the graph-level features are input into the fully connected layer to output the probability distribution of the logic error type.

6. The method of claim 5, wherein the method further comprises: The computing node pays attention to the weight of its neighbor nodes, and according to the attention weight, aggregates the features of the neighbor nodes, updates the features of the node, including: The correlation calculation formula of the node pair is as follows: ； In the formula, a represents an attention weight vector, Z i , Z j respectively represent the features of the node and the neighbor node after linear transformation, and represents the concatenation of Z i and Z j . The attention weight normalization formula is as follows: ； where N i is the set of all neighbor nodes of node v i ​ The feature update formula of the node is as follows: ； In the formula, is an activation function.

7. A system for revising and integrating a road and bridge engineering tender document, applied to the method for revising and integrating a road and bridge engineering tender document according to any one of claims 1 to 6, characterized in that, Including: The input module is used for receiving the original bidding file of road and bridge engineering and the format requirement data of the tendering party, and synchronizing the format requirement data of the tendering party to the road and bridge engineering sub-scene format library; The element analysis module is used for extracting basic elements and identifying road and bridge industry specific elements from the received original bidding file of road and bridge engineering, and outputting analysis results containing industry element information; The correction module is used for calling the rules of the road and bridge engineering sub-scene format library to correct the analysis results in the road and bridge industry special format, and output the first preliminary correction result; The machine learning model is used to correct the analysis results to obtain the second preliminary correction result; The joint module is used to fuse the first preliminary correction result and the second preliminary correction result, obtain conflict-free data, and merge the conflict-free data according to the logical sequence of the road and bridge engineering bidding file standard to generate the target bidding file of road and bridge engineering.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the steps of the road and bridge engineering bidding file correction and joint method according to any one of claims 1 to 6.

9. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the steps of the road and bridge engineering bidding file correction and joint method according to any one of claims 1 to 6.

Images