A new auxiliary bidding and tendering review system

Abstract

The present invention relates to the technical field of information data processing. The present invention provides a novel auxiliary bidding and tendering review system, which includes a server end, a bidder client and a background control end. Through modules such as bidding material acquisition, classification, matching, inspection and review, the automatic processing and review of bidding documents are realized. The system can receive bidding materials in various formats, perform image recognition and classification, match project demand information, simulate and test the compliance of bidding materials, and feedback the correction results. Finally, the system reviews the corrected bidding materials. By integrating word embedding technology, random forest model, neural network model and other technologies, the present invention improves the review efficiency and accuracy and reduces human intervention.

Description

Technical Field

[0001] The present invention relates to the technical field of information data processing, and in particular to a novel auxiliary bidding and tendering review system. Background Art

[0002] The bidding and tendering review system is a key tool to ensure the fairness, transparency and efficiency of bidding and tendering activities. It integrates the dual functions of expert information management and review process management. The bidding and tendering review system can automate the core links such as bid submission, review and result announcement. The bidding and tendering review system realizes the intelligent allocation of review tasks, progress tracking and comprehensive analysis of results. The bidding and tendering review system significantly shortens the project preparation cycle, improves the market response speed, and performs well in information security and confidentiality, effectively preventing unfair competition and maintaining market fairness and order.

[0003] In the traditional bidding and tendering review process, bidders need to manually prepare bids, and the form and content of bid documents are often complex and diverse. After entering the offline review stage, experts and owners need to conduct a large amount of independent review work, which is inefficient and easily affected by human factors. In order to improve the working efficiency of the bidding and tendering inquiry and evaluation system, simplify the work process, and reduce human intervention, the present invention provides a new auxiliary bidding and tendering review system. Summary of the invention

[0004] In view of the shortcomings of the existing technology, the present invention provides a new auxiliary bidding and tendering review system to solve the problem that the form and content of the bidding documents are often complex and diverse, and after entering the offline review stage, experts and owners need to carry out a large amount of independent review work, which is inefficient and easily affected by human factors.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0006] The present invention discloses a novel auxiliary bidding and tendering review system, comprising: a server end, a bidder client end and a background control end, wherein the server end is connected to the bidder client end and the background control end respectively, and the server end comprises:

[0007] The bidding material acquisition module receives the bidding material data from the bidding client on the server side, performs image recognition on the received bidding material data, and obtains the bidding material image recognition result.

[0008] The bidding material classification module extracts data features from the bidding material image recognition results to obtain image feature data, substitutes the image feature data into a preset bidding material classification model to obtain classified bidding data;

[0009] A bidding material matching module obtains bidding project demand information, matches the bidding project demand information with the classified bidding data, obtains bidding data matching results, the bidding data matching results include successfully matched data and unsuccessfully matched data, establishes a corresponding relationship between the successfully matched data and the bidding project demand information, retrieves the bidding material data corresponding to the unsuccessfully matched data, converts the bidding material data corresponding to the unsuccessfully matched data into text data, performs cluster analysis on the text data, obtains text data cluster analysis results, matches the text data cluster analysis results in a bidding data knowledge base, and obtains bidding project categories corresponding to the text data cluster analysis results;

[0010] The bidding material inspection module substitutes the bidding project category corresponding to the text data cluster analysis result and the matching success data into the preset bidding material simulation test model to obtain the first bidding material simulation test result, which includes the detection success data and the detection failure data, and transmits the detection failure data to the bidder client and the backend control end;

[0011] The bidding material review module receives feedback data from the bidder client after correcting the unsuccessful detection data, obtains the repaired bidding data, substitutes the repaired bidding data and the successfully matched data into the preset bidding material simulation test model, and obtains the second bidding material simulation test result. If the second bidding material simulation test result is qualified, the second bidding material simulation test result is marked as information to be reviewed and sent to the background control end.

[0012] Furthermore, in the novel auxiliary bidding and tendering review system of the present invention, the bidding materials acquisition module is also used for:

[0013] The server receives the bidding material data uploaded by the bidder’s client, including documents, pictures and PDF;

[0014] Preprocess the received bidding materials, including file format conversion, image clarity optimization and image denoising;

[0015] Use the image recognition algorithm of optical character recognition technology to perform image recognition on the pre-processed bidding materials, extract the text, numbers and symbol information in the image, and convert it into editable and searchable text data;

[0016] After the image recognition is completed, the image recognition results of the bidding materials are generated. The image recognition results of the bidding materials include the recognized text content, text location information and elements in the image. The elements in the image include tables and charts.

[0017] Furthermore, in the novel auxiliary bidding and tendering review system of the present invention, the bidding material classification module is also used for:

[0018] Performing text preprocessing on the text data in the bidding material image recognition results, the text preprocessing includes word segmentation, stop word removal and stemming text to obtain preprocessed text data;

[0019] The pre-processed text data is converted into numerical feature vectors using word embedding technology, and the pre-trained bidding material classification model is loaded from the server's model library;

[0020] The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model automatically classifies the bidding materials according to the input feature data and the learning mechanism within the bidding material classification model.

[0021] The bidding material classification model outputs the classification results, which are the category labels to which the bidding materials are classified. The category labels to which the bidding materials are classified include technical solutions, financial statements, and project plans.

[0022] The classification results are associated with the original bidding material data to form classified bidding data; the classified bidding data is stored in the database of the server, so that the classification results can be provided to the bidder client and the backend control terminal through the API interface for viewing and confirmation.

[0023] Furthermore, in the novel auxiliary bidding and tendering review system of the present invention, the bidding material matching module is also used for:

[0024] Obtain demand information of the current bidding project from the backend control terminal or related database, including project background, technical requirements, performance indicators and qualification requirements;

[0025] The acquired bidding project demand information is matched with the previously classified bidding data one by one, matching similarity calculation, keyword matching and semantic analysis are performed, and the bidding data is divided into successfully matched data and unsuccessfully matched data. The successfully matched data refers to the part of the bidding materials that matches the similarity with the bidding project demand information or meets the similarity requirements, and the unsuccessfully matched data refers to the part of the bidding materials that does not match the similarity with the bidding project demand information or does not meet the similarity requirements. A corresponding relationship is established between the successfully matched data and the corresponding bidding project demand information;

[0026] The original bidding material data corresponding to the unsuccessful matching part is retrieved from the classified bidding data. If the original bidding material data is in non-text form, it is converted into editable text data.

[0027] Apply cluster analysis algorithm to process the converted text data to obtain patterns, themes and categories. After the cluster analysis is completed, the cluster analysis results of the text data are generated. The cluster analysis results include different clusters, representative words or phrases in each cluster, and similarities and differences between clusters;

[0028] Matching the text data cluster analysis results with the information in the bidding data knowledge base, which includes historical bidding data, successful cases, and industry norms;

[0029] According to the matching process, the bidding project category corresponding to the text data clustering analysis result is determined, and the matching results, corresponding relationships, clustering analysis results and project categories are stored in the server's database and provided to the bidder's client or backend control terminal for viewing and confirmation via an API interface.

[0030] Furthermore, in the novel auxiliary bidding and tendering review system of the present invention, the bidding materials inspection module is also used for:

[0031] The bidding project categories corresponding to the text data cluster analysis results and the matching successful data are used as input information to load the preset bidding material simulation test model:

[0032] Loading a bidding material simulation test model from a model library on a server, the bidding material simulation test model simulates the actual review process based on the input bidding data and outputs a test result;

[0033] Substitute the bidding project category and the matching success data into the bidding material simulation test model, and input the text data and numerical data into the corresponding input layer of the model;

[0034] The bidding material simulation test model is started, and the bidding material simulation test model processes the input bidding data, and the bidding material simulation test model generates simulation test results;

[0035] After the bidding materials simulation test model is completed, the first bidding materials simulation test result is output. The first bidding materials simulation test result includes successful test data and unsuccessful test data. The successful test data refers to the part of the bidding materials that meets the preset standards of the bidding materials simulation test model, and the unsuccessful test data refers to the part of the bidding materials that does not meet the preset standards of the bidding materials simulation test model.

[0036] Furthermore, in the novel auxiliary bidding and tendering review system of the present invention, the bidding material classification module is also used for:

[0037] Acquire historical data of bidding material image recognition results, perform word segmentation, remove stop words, and perform stem extraction on text data in the historical data of bidding material image recognition results to obtain pre-processed historical text data;

[0038] The preprocessed text history data is converted into numerical feature vectors using word embedding technology, and the random forest model is trained using the numerical feature vectors to obtain a bidding material classification model;

[0039] The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model automatically classifies the bidding materials based on the input image feature data and the random forest learning algorithm within the model.

[0040] The bidding material classification model outputs classification results, which are the category labels into which the bidding materials are classified. The category labels include technical solutions, financial statements, and project plans.

[0041] The classification results are associated with the original bidding material data to form classified bidding data, which are stored in the database of the server.

[0042] Furthermore, in the novel auxiliary bidding and tendering review system of the present invention, the bidding materials inspection module is also used for:

[0043] Obtain the bidding project categories and matching success data corresponding to the historical text data clustering analysis results, start the bidding material simulation test model, and input the historical text data and numerical data into the input layer of the neural network model. The neural network model trains the input bidding data to obtain the bidding material simulation test model.

[0044] Beneficial effects of the present invention:

[0045] The system significantly reduces the independent review work of experts and owners during the review stage through automated processing of bidding documents, intelligent classification, automatic matching and simulation testing, thereby greatly improving the review efficiency. The system adopts modeling processing methods, such as random forest models and neural network models, to reduce the subjective influence of human factors on the review results and enhance the fairness and transparency of the review.

[0046] The system can receive and process bidding materials in various formats, including documents, images, and PDFs, which improves the system's adaptability and flexibility and meets the needs of different bidders. Through automated review processes and modeled processing methods, the system reduces human intervention, thereby ensuring the fairness of the review process and the transparency of the results. The system can generate detailed review result reports, including successful and unsuccessful test data, to help reviewers quickly understand the compliance and accuracy of bidding materials and provide support for decision-making.

[0047] Bidders can correct the data that failed to be tested and re-enter the system for testing. This function ensures the final qualification of the bidding materials and improves the accuracy and effectiveness of the review. The system integrates advanced technical means such as word embedding technology, random forest model, neural network model, and uses historical data for model training to continuously improve the system's intelligence level and processing capabilities, providing more efficient and accurate support for bidding and tendering reviews.

[0048] Through the automated review process, the system significantly shortens the project preparation cycle, improves the market response speed, and provides a strong guarantee for the smooth progress of bidding activities. The system performs well in information security and confidentiality, effectively prevents the occurrence of unfair competition, maintains market fairness and order, and provides strong support for the healthy development of bidding activities. The system provides classification results, matching results and other information to the bidder client and backend control end for viewing and confirmation through the API interface, which enhances the transparency and operability of the system and improves user experience and satisfaction.

[0049] To sum up, the new auxiliary bidding and tendering review system provided by the present invention effectively solves the pain points in the traditional bidding and tendering review process through a series of innovative technical means and functional designs, significantly improves the review efficiency and transparency, and provides a strong guarantee for the smooth progress of bidding and tendering activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0051] Figure 1 A schematic diagram of a unit module in a novel auxiliary bidding and tendering review system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0053] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0054] The present invention discloses a novel auxiliary bidding and tendering review system, comprising: a server end, a bidder client end and a background control end, wherein the server end is connected to the bidder client end and the background control end respectively, and the server end comprises:

[0055] The bidding material acquisition module receives the bidding material data from the bidding client on the server side, performs image recognition on the received bidding material data, and obtains the bidding material image recognition result.

[0056] The server is responsible for receiving the bidding material data uploaded by the bidder’s client, which covers multiple formats such as documents, pictures, and PDF. A series of pre-processing work is performed on the received bidding materials, including file format conversion, image clarity optimization, and image denoising, to ensure the accuracy and efficiency of subsequent processing.

[0057] Using advanced optical character recognition (OCR) technology, the pre-processed bidding materials are image-recognized. The core task of this step is to accurately extract the text, numbers, and symbol information in the image and convert them into editable and searchable text data.

[0058] After completing the image recognition, the system will generate the image recognition results of the bidding materials. These results include not only the recognized text content, but also the text location information and other elements, tables and charts in the image.

[0059] As an important component of the server side, the bidding material acquisition module works closely with other modules (such as the bidding material classification module, bidding material matching module, etc.) to jointly promote the smooth progress of the automated process of bidding and tendering review.

[0060] The bidding material classification module extracts data features from the bidding material image recognition results to obtain image feature data, substitutes the image feature data into a preset bidding material classification model to obtain classified bidding data;

[0061] The bidding material classification module is responsible for accurately classifying the bidding material image recognition results, laying a solid foundation for subsequent matching and processing work.

[0062] The bidding material classification module first extracts the data features from the bidding material image recognition results to obtain image feature data, and then substitutes these image feature data into the preset bidding material classification model for classification processing.

[0063] Before classification, the bidding material classification module preprocesses the text data in the bidding material image recognition results, including word segmentation, stop word removal, and stem extraction, etc., in order to improve the accuracy of classification. Using advanced word embedding technology, the preprocessed text data is converted into a numerical feature vector so that the model can process it efficiently.

[0064] A pre-trained bidding material classification model is loaded from the model library of the server. The bidding material classification model has the ability to automatically classify bidding materials.

[0065] The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model uses its internal learning mechanism to automatically and accurately classify the bidding materials.

[0066] The classification model outputs the classification results, which are the specific category labels into which the bidding materials are classified. The category labels include technical solutions, financial statements, and project plans, etc.

[0067] The classification results are associated with the original bidding material data to form classified bidding data, which is then stored in the server database. The classification results can be provided to the bidder client and backend control terminal through the API interface for easy viewing and confirmation.

[0068] The bidding material classification model is also responsible for obtaining historical data of bidding material image recognition results and preprocessing these historical data such as word segmentation, stop word removal, and stem extraction. The preprocessed historical data is used to train the random forest model to continuously improve the accuracy and performance of the bidding material classification model.

[0069] The bidding material classification module is closely integrated with other modules in the system (such as bidding material acquisition module, bidding material matching module, etc.), together forming a complete and efficient functional system of the new auxiliary bidding and tendering review system.

[0070] A bidding material matching module obtains bidding project demand information, matches the bidding project demand information with the classified bidding data, obtains bidding data matching results, the bidding data matching results include successfully matched data and unsuccessfully matched data, establishes a corresponding relationship between the successfully matched data and the bidding project demand information, retrieves the bidding material data corresponding to the unsuccessfully matched data, converts the bidding material data corresponding to the unsuccessfully matched data into text data, performs cluster analysis on the text data, obtains text data cluster analysis results, matches the text data cluster analysis results in a bidding data knowledge base, and obtains bidding project categories corresponding to the text data cluster analysis results;

[0071] Obtain detailed demand information of the current bidding project from the backend control terminal or related database, including key contents such as project background, technical requirements, performance indicators and qualification requirements.

[0072] The acquired bidding project demand information is matched with the previously classified bidding data one by one. The matching process involves technologies such as similarity calculation, keyword matching, and semantic analysis. According to the matching results, the bidding data is divided into successfully matched data and unsuccessfully matched data. Successfully matched data refers to the part of the bidding materials that matches or meets the similarity requirements with the bidding project demand information, while unsuccessfully matched data refers to the part of the bidding materials that does not match or does not meet the similarity requirements with the bidding project demand information.

[0073] Establish a clear correspondence between the successfully matched data and the corresponding bidding project demand information to facilitate subsequent review work.

[0074] The original bidding material data corresponding to the unsuccessful matching part is retrieved from the classified bidding data. If the original bidding material data exists in non-text form (such as pictures, PDF, etc.), relevant technologies are used to convert it into editable text data for subsequent processing.

[0075] The transformed text data is processed in depth using cluster analysis algorithms to reveal its underlying patterns, themes, and categories. After the cluster analysis is completed, the cluster analysis results of the text data are generated, including different clusters, representative words or phrases in each cluster, and similarities and differences between clusters.

[0076] The text data clustering analysis results are matched with the information in the bidding data knowledge base. The bidding data knowledge base covers historical bidding data, successful cases, industry norms and other aspects, providing a rich reference for the matching process.

[0077] According to the results of the matching process, the bidding project category corresponding to the text data cluster analysis results is determined to provide clear classification information for the reviewers.

[0078] The key information such as matching results, corresponding relationships, cluster analysis results and project categories are stored in the server database and provided to the bidder’s client or backend control terminal for review and confirmation through the API interface. This step ensures the traceability and accessibility of the information and facilitates the subsequent review work.

[0079] The bidding materials matching module can efficiently match bidding materials with project requirements, helping reviewers quickly identify bidding materials that meet the requirements, thereby improving the efficiency of bidding review.

[0080] The bidding material inspection module substitutes the bidding project category corresponding to the text data cluster analysis result and the matching success data into the preset bidding material simulation test model to obtain the first bidding material simulation test result, which includes the detection success data and the detection failure data, and transmits the detection failure data to the bidder client and the backend control end;

[0081] The bidding project category corresponding to the text data clustering analysis results and the matching successful data are used as input information, and the preset bidding material simulation test model is loaded to prepare for subsequent simulation testing.

[0082] The bidding material simulation test model simulates the actual review process based on the input bidding data. This process is designed to conduct a comprehensive inspection and evaluation of the bidding materials through the model's preset standards and logic.

[0083] Substitute the bidding project categories and successful matching data into the bidding material simulation test model, including inputting text data and numerical data into the corresponding input layer of the model to ensure that the model can accurately process these data.

[0084] Start the bidding materials simulation test model to process the input bidding data. The bidding materials simulation test model will check the bidding materials one by one according to the preset algorithms and logics and generate simulation test results.

[0085] After the bidding material simulation test model is completed, the first bidding material simulation test result is output. The first bidding material simulation test result includes successful test data and unsuccessful test data, providing a basis for subsequent processing and correction.

[0086] Successfully tested data refers to the portion of bidding materials that meets the preset standards of the bidding materials simulation test model. This portion of data will be deemed qualified and may enter the subsequent review process.

[0087] Unsuccessful detection data refers to the part of the bidding materials that does not meet the preset standards of the bidding materials simulation test model. This part of the data needs to be further processed and corrected.

[0088] The unsuccessful detection data is transmitted to the bidder's client and the back-end control terminal. This step is intended to let the bidder and the back-end control terminal know which parts of the data do not meet the requirements so as to carry out subsequent processing and correction.

[0089] After the bidder client corrects the unsuccessful detection data, it receives the corrected feedback data. The corrected bidding data and the successfully matched data are again substituted into the preset bidding material simulation test model for a second simulation test. This step is to ensure that the corrected data meets the preset standards and requirements of the model.

[0090] If the simulated test result of the second bidding material is qualified, the simulated test result of the second bidding material will be marked as pending information and sent to the back-end control terminal for final review. This step is to ensure that all simulated test data are finally reviewed and confirmed to ensure their accuracy and compliance.

[0091] The bidding material review module receives feedback data from the bidder client after correcting the unsuccessful detection data, obtains the repaired bidding data, substitutes the repaired bidding data and the successfully matched data into the preset bidding material simulation test model, and obtains the second bidding material simulation test result. If the second bidding material simulation test result is qualified, the second bidding material simulation test result is marked as information to be reviewed and sent to the background control end.

[0092] Through the received feedback data, the module can obtain the repaired bidding data, which has been revised by the bidder to solve the problems found in the previous detection.

[0093] Substitute the repaired bidding data and the previously successfully matched data into the preset bidding material simulation test model. This step is to re-test the repaired bidding data to make it meet the evaluation criteria.

[0094] The bidding material simulation test model processes the repaired bidding data and the successfully matched data, and outputs the second bidding material simulation test result, which will reflect whether the repaired bidding data meets the evaluation requirements.

[0095] The bidding material simulation test model will judge the second bidding material simulation test results to determine whether the bidding material simulation test model is qualified. This step is the key link to ensure that the bidding materials can meet the evaluation standards after being revised.

[0096] If the second bidding material simulation test result is qualified, the bidding material review module will mark the result as pending review information and send it to the backend control end. This step is to enable the backend control end to understand which bidding materials have been revised and meet the evaluation criteria in order to carry out subsequent review work.

[0097] The bidding materials review module plays a role in the entire system to ensure that the bidding materials meet the review standards after being revised. By receiving the revised feedback data, substituting it into the simulation test model, obtaining the test results and sending the information to be reviewed, the bidding materials review module improves the integrity and accuracy of the bidding review process.

[0098] Specifically, in the novel auxiliary bidding and tendering review system of the present invention, the bidding material acquisition module is also used for:

[0099] The server receives the bidding material data uploaded by the bidder’s client, including documents, pictures and PDF;

[0100] Preprocess the received bidding materials, including file format conversion, image clarity optimization and image denoising;

[0101] Use the image recognition algorithm of optical character recognition technology to perform image recognition on the pre-processed bidding materials, extract the text, numbers and symbol information in the image, and convert it into editable and searchable text data;

[0102] After the image recognition is completed, the image recognition results of the bidding materials are generated. The image recognition results of the bidding materials include the recognized text content, text location information and elements in the image. The elements in the image include tables and charts.

[0103] Specifically, the bidding materials acquisition module in the novel auxiliary bidding and tendering review system of the present invention also has the following functions:

[0104] The server can receive various bidding materials data uploaded by the bidder’s client, including but not limited to documents, pictures and PDF files. This function ensures that the system can handle a variety of bidding materials and meet different evaluation requirements.

[0105] The received bidding materials are preprocessed, and the preprocessing process includes file format conversion, image clarity optimization, and image denoising. File format conversion ensures that all bidding materials can be processed uniformly by the system, while image clarity optimization and denoising improve the accuracy of image recognition, laying a solid foundation for subsequent processing.

[0106] The image recognition algorithm of advanced optical character recognition (OCR) technology is used to perform image recognition on the pre-processed bidding materials. This step can accurately extract the text, numbers and symbol information in the image and convert it into editable and searchable text data. This greatly improves the processing efficiency of bidding materials and makes subsequent review work more convenient.

[0107] After image recognition is completed, the system generates image recognition results of the bidding materials. These results include not only the recognized text content, but also the text location information and elements in the image, such as tables and charts. These rich recognition results provide strong support for subsequent classification, matching and inspection of bidding materials.

[0108] Specifically, in the novel auxiliary bidding and tendering review system of the present invention, the bidding material classification module is also used for:

[0109] Performing text preprocessing on the text data in the bidding material image recognition results, the text preprocessing includes word segmentation, stop word removal and stemming text to obtain preprocessed text data;

[0110] The pre-processed text data is converted into numerical feature vectors using word embedding technology, and the pre-trained bidding material classification model is loaded from the server's model library;

[0111] The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model automatically classifies the bidding materials according to the input feature data and the learning mechanism within the bidding material classification model.

[0112] The bidding material classification model outputs the classification results, which are the category labels to which the bidding materials are classified. The category labels to which the bidding materials are classified include technical solutions, financial statements, and project plans.

[0113] The classification results are associated with the original bidding material data to form classified bidding data; the classified bidding data is stored in the database of the server, so that the classification results can be provided to the bidder client and the backend control terminal through the API interface for viewing and confirmation.

[0114] The text data in the bidding material image recognition results are preprocessed. This process includes word segmentation, stop word removal, and stem extraction, aiming to obtain cleaner and more accurate text data, laying the foundation for subsequent processing.

[0115] Using word embedding technology, the preprocessed text data is converted into numerical feature vectors. These feature vectors can represent the semantic information of the text data and provide effective input for the classification model.

[0116] Load the pre-trained bidding material classification model from the server's model library. This model is built based on machine learning or deep learning algorithms and has the ability to automatically classify bidding materials.

[0117] The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The model automatically classifies the bidding materials based on the input feature data using its internal learning mechanism to achieve efficient and accurate classification processing.

[0118] The bidding materials classification model outputs the classification results, that is, the category labels into which the bidding materials are classified. These category labels include technical solutions, financial statements, and project plans, etc., providing a clear classification basis for subsequent review work.

[0119] The classification results are associated with the original bidding material data to form the classified bidding data. Subsequently, the classified bidding data is stored in the server database for subsequent processing and query. The classification results are provided to the bidder client and the backend control terminal through the API interface for viewing and confirmation. This function enhances the transparency and operability of the system, allowing users to easily obtain the classification results and perform subsequent processing.

[0120] Obtain the historical data of the bidding materials image recognition results, preprocess the text data and convert them into feature vectors. Then, use these feature vectors to train the random forest model to obtain a more accurate bidding materials classification model. This process helps to continuously improve the accuracy and performance of the classification model.

[0121] Substitute the extracted image feature data into the trained bidding material classification model to automatically classify new bidding materials. This function improves the accuracy and efficiency of classification and provides strong support for bidding review work.

[0122] As the bidding material data continues to increase and update, the bidding material classification module can continuously train and optimize the model. This function enables the system to adapt to the ever-changing bidding environment, improve overall performance and meet the changing needs of users.

[0123] In summary, the bidding material classification module plays a vital role in the new auxiliary bidding and tendering review system. Through efficient and accurate classification processing and rich functional support, it provides strong guarantee and convenient operation experience for bidding and tendering review work.

[0124] Specifically, the bidding materials matching module of the novel auxiliary bidding and tendering review system of the present invention is also used for:

[0125] Obtain demand information of the current bidding project from the backend control terminal or related database, including project background, technical requirements, performance indicators and qualification requirements;

[0126] The acquired bidding project demand information is matched with the previously classified bidding data one by one, matching similarity calculation, keyword matching and semantic analysis are performed, and the bidding data is divided into successfully matched data and unsuccessfully matched data. The successfully matched data refers to the part of the bidding materials that matches the similarity with the bidding project demand information or meets the similarity requirements, and the unsuccessfully matched data refers to the part of the bidding materials that does not match the similarity with the bidding project demand information or does not meet the similarity requirements. A corresponding relationship is established between the successfully matched data and the corresponding bidding project demand information;

[0127] The original bidding material data corresponding to the unsuccessful matching part is retrieved from the classified bidding data. If the original bidding material data is in non-text form, it is converted into editable text data.

[0128] Apply cluster analysis algorithm to process the converted text data to obtain patterns, themes and categories. After the cluster analysis is completed, the cluster analysis results of the text data are generated. The cluster analysis results include different clusters, representative words or phrases in each cluster, and similarities and differences between clusters;

[0129] Matching the text data cluster analysis results with the information in the bidding data knowledge base, which includes historical bidding data, successful cases, and industry norms;

[0130] According to the matching process, the bidding project category corresponding to the text data clustering analysis result is determined, and the matching results, corresponding relationships, clustering analysis results and project categories are stored in the server's database and provided to the bidder's client or backend control terminal for viewing and confirmation via an API interface.

[0131] The bidding material matching module can obtain the demand information of the current bidding project from the background control terminal or related database. The demand information includes but is not limited to project background, technical requirements, performance indicators and qualification requirements, providing an accurate basis for subsequent matching analysis.

[0132] The acquired bidding project demand information is matched with the previously classified bidding data one by one. The matching process involves a variety of technical means such as similarity calculation, keyword matching and semantic analysis to improve the accuracy of matching. Through this process, the bidding data is divided into two categories: successfully matched data and unsuccessfully matched data.

[0133] Successfully matched data refers to the part of the bidding materials that matches or meets the similarity requirements of the bidding project demand information. This module will establish a corresponding relationship between this part of the data and the corresponding bidding project demand information to facilitate subsequent review work.

[0134] For unsuccessful matching data, that is, the part of the bidding materials that does not match the similarity of the bidding project demand information or does not meet the similarity requirements, the module will retrieve the corresponding original bidding material data from the classified bidding data. If this data exists in non-text form, the module will convert it into editable text data for subsequent processing.

[0135] The converted text data is processed by applying cluster analysis algorithms to reveal its inherent patterns, themes, and categories. After the cluster analysis is completed, detailed cluster analysis results will be generated, including different clusters, representative words or phrases in each cluster, and similarities and differences between clusters. This information helps to deeply understand the content characteristics of the bidding materials.

[0136] Knowledge base matching matches the text data clustering analysis results with the information in the bidding data knowledge base. The bidding data knowledge base covers historical bidding data, successful cases, industry norms and other aspects, providing rich references for the matching process.

[0137] Determine the bidding project category. Based on the results of the matching process, this module can determine the bidding project category corresponding to the text data cluster analysis results. This step is of great significance for the classification and positioning of subsequent review work.

[0138] Finally, the module will store the matching results, corresponding relationships, cluster analysis results, project categories and other information in the server database. At the same time, this information will be provided to the bidder’s client or backend control terminal for review and confirmation through the API interface, which enhances the transparency and operability of the system.

[0139] Specifically, the bidding materials inspection module of the novel auxiliary bidding and tendering review system of the present invention is also used for:

[0140] The bidding project categories corresponding to the text data cluster analysis results and the matching successful data are used as input information to load the preset bidding material simulation test model:

[0141] Loading a bidding material simulation test model from a model library on a server, the bidding material simulation test model simulates the actual review process based on the input bidding data and outputs a test result;

[0142] Substitute the bidding project category and the matching success data into the bidding material simulation test model, and input the text data and numerical data into the corresponding input layer of the model;

[0143] The bidding material simulation test model is started, and the bidding material simulation test model processes the input bidding data, and the bidding material simulation test model generates simulation test results;

[0144] After the bidding materials simulation test model is completed, the first bidding materials simulation test result is output. The first bidding materials simulation test result includes successful test data and unsuccessful test data. The successful test data refers to the part of the bidding materials that meets the preset standards of the bidding materials simulation test model, and the unsuccessful test data refers to the part of the bidding materials that does not meet the preset standards of the bidding materials simulation test model.

[0145] The bidding material inspection module is mainly responsible for substituting the bidding project categories corresponding to the text data clustering analysis results and the matching successful data into the preset bidding material simulation test model, and obtaining the first bidding material simulation test results by simulating the actual review process.

[0146] The input information received by the bidding material checking module includes the bidding project category and the matching success data corresponding to the text data clustering analysis result.

[0147] Load the bidding material simulation test model from the server's model library. The bidding material simulation test model can simulate the actual review process based on the input bidding data and output the test results. Substitute the bidding project category and the matching success data into the bidding material simulation test model, and input the text data and numerical data into the corresponding input layer of the model to prepare data for simulation testing.

[0148] The bidding material simulation test model is started to process the input bidding data. The model will generate simulation test results based on its internal learning mechanism to evaluate the compliance and accuracy of the bidding materials. After the bidding material simulation test model is completed, the first bidding material simulation test result is output. The result includes successful test data and unsuccessful test data. Successful test data refers to the part of the bidding materials that meets the preset standards of the model, while unsuccessful test data refers to the part of the bidding materials that does not meet the preset standards of the model.

[0149] The unsuccessful detection data is transmitted to the bidder's client and the backend control terminal. In this way, the bidder can understand the existing problems in a timely manner and make corrections to improve the compliance and accuracy of the bidding materials. By simulating the actual review process, the bidding material inspection module can identify possible problems in advance, thereby reducing the workload of the offline review stage, improving the review efficiency, and reducing the impact of human factors. This lays a solid foundation for subsequent review work and ensures the compliance and accuracy of the bidding materials.

[0150] Specifically, in the novel auxiliary bidding and tendering review system of the present invention, the bidding material classification module is also used for:

[0151] Acquire historical data of bidding material image recognition results, perform word segmentation, remove stop words, and perform stem extraction on text data in the historical data of bidding material image recognition results to obtain pre-processed historical text data;

[0152] The preprocessed text history data is converted into numerical feature vectors using word embedding technology, and the random forest model is trained using the numerical feature vectors to obtain a bidding material classification model;

[0153] The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model automatically classifies the bidding materials based on the input image feature data and the random forest learning algorithm within the model.

[0154] The bidding material classification model outputs classification results, which are the category labels into which the bidding materials are classified. The category labels include technical solutions, financial statements, and project plans.

[0155] The classification results are associated with the original bidding material data to form classified bidding data, which are stored in the database of the server.

[0156] The bidding material classification module obtains the image recognition results of the bidding materials and extracts data features from them to obtain image feature data. This is the basis for subsequent classification processing. Using advanced word embedding technology, the pre-processed text data (including word segmentation, stop word removal, and stem extraction) is converted into a numerical feature vector. This conversion process enables the text data to be effectively processed by the machine learning model.

[0157] Substitute the extracted image feature data into the preset bidding material classification model, and obtain the classified bidding data through the automatic classification function of the model. Load the pre-trained bidding material classification model from the server's model library. This model is based on the random forest algorithm and can accurately and automatically classify bidding materials.

[0158] The bidding material classification model outputs the classification results, that is, the category labels to which the bidding materials are classified. This is one of the key information in the review process. Category labels include technical solutions, financial statements, and project plans, etc. These labels help reviewers quickly understand the content and type of bidding materials.

[0159] The classification results are associated with the original bidding material data to form the classified bidding data. This step ensures the integrity and consistency of the data. The classified bidding data is stored in the server database for subsequent query and use. At the same time, the classification results are provided to the bidder client and the backend control terminal through the API interface for viewing and confirmation, which enhances the transparency and operability of the system.

[0160] In order to obtain an accurate bidding material classification model, it is necessary to obtain the historical data of bidding material image recognition results and use the data to train the random forest model.

[0161] First, the text data in the bidding material image recognition result historical data is preprocessed by word segmentation, stop word removal, and stem extraction to obtain preprocessed historical text data. Then, the random forest model is trained using the preprocessed data to finally obtain the bidding material classification model.

[0162] In summary, the bidding materials classification module classifies and processes bidding materials in an automated and intelligent way, improving the efficiency of the bidding and tendering review system. At the same time, it provides strong support for reviewers through rich functional support and convenient operation experience.

[0163] Specifically, the bidding materials inspection module of the novel auxiliary bidding and tendering review system of the present invention is also used for:

[0164] Obtain the bidding project categories and matching success data corresponding to the historical text data clustering analysis results, start the bidding material simulation test model, and input the historical text data and numerical data into the input layer of the neural network model. The neural network model trains the input bidding data to obtain the bidding material simulation test model.

[0165] The main purpose of training the neural network model is to obtain a bidding material simulation test model. The bidding material simulation test model can simulate the actual review process based on the input bidding data and output the test results, thereby helping reviewers quickly identify the parts of the bidding materials that need attention.

[0166] The training data mainly includes the bidding project categories corresponding to the cluster analysis results of historical text data and the matching success data. These data are the basis for the training of the neural network model. By training the bidding material simulation test model, the relationship between different bidding material characteristics and the evaluation results can be learned.

[0167] During the training process, the bidding material simulation test model is first started, and the historical text data and numerical data are input into the input layer of the neural network model. Then, the neural network model is trained on the input bidding data, and the model parameters are continuously adjusted and optimized so that it can accurately simulate the actual review process.

[0168] After the training is completed, the bidding material simulation test model can be used to test new bidding data. Through the prediction and evaluation of the model, it can quickly determine whether the new bidding data meets the preset evaluation criteria, providing strong support for the reviewers.

[0169] After the bidding material simulation test model is run, the first bidding material simulation test result is output. This result includes successful test data and unsuccessful test data, which respectively represent the parts of the bidding materials that meet and do not meet the preset review criteria. This helps reviewers quickly locate and focus on bidding materials that need further review.

[0170] As new bidding data and review results accumulate, the neural network model can be further trained and optimized. By continuously updating training data and adjusting model parameters, the accuracy and generalization ability of the model can be improved, making it better adapted to the bidding review needs in different scenarios.

[0171] The technical solution of the present invention effectively solves the problems that the form and content of the bidding documents are complex and diverse, and that experts and owners need to do a lot of independent review work during the offline review stage, which is inefficient and easily affected by human factors. The specific solution is as follows:

[0172] The bidding material acquisition module receives the bidding material data (including documents, pictures and PDF) uploaded by the bidder’s client, and performs pre-processing such as file format conversion, image clarity optimization and image denoising. Then, optical character recognition technology is used to extract text, numbers and symbols in the image and convert them into editable and searchable text data.

[0173] The bidding materials classification module performs word segmentation, removes stop words, and extracts stems on the preprocessed text data, and converts them into numerical feature vectors. Then, the bidding materials are automatically classified using the pre-trained bidding materials classification model, and the classification labels include technical solutions, financial statements, and project plans.

[0174] The bidding material matching module is responsible for obtaining bidding project demand information and matching it with the classified bidding data. The successfully matched data is associated with the bidding project demand information. For the unsuccessful matching data, it is converted into text data for cluster analysis and then matched with the information in the bidding data knowledge base to determine the bidding project category.

[0175] The bidding material inspection module substitutes the bidding project categories and matching successful data corresponding to the cluster analysis results into the preset bidding material simulation test model, performs simulation detection, and generates detection results, which include successful detection data and unsuccessful detection data. The bidding material review module receives the feedback data of the bidder client after correcting the unsuccessful detection data, and re-substitutes it into the simulation test model for detection. If the result is qualified, it is marked as information to be reviewed and sent to the background control end.

[0176] By automating the processing of bidding documents, intelligent classification, automatic matching and simulation testing, the present invention significantly reduces the independent review work of experts and owners in the review stage, thereby improving the review efficiency. The modeling processing method reduces the impact of human factors on the review results and enhances the fairness and transparency of the review.

[0177] This system integrates advanced technical means such as word embedding technology, random forest model, neural network model, etc., and trains the model through historical data to continuously improve the intelligence level and processing capabilities of the system.

[0178] In summary, the synergy of these modules and functions enables the present invention to efficiently process complex and diverse bidding documents and automatically complete most of the review work, thereby greatly improving the efficiency and fairness of bidding reviews and effectively reducing interference from human factors.

Claims

1. A new type of auxiliary bidding and tendering review system, characterized in that: include: The server side, the bidder client side and the backend control side. The server side is connected to the bidder client side and the backend control side respectively. The server side includes: The bidding material acquisition module receives the bidding material data from the bidding client on the server side, performs image recognition on the received bidding material data, and obtains the bidding material image recognition result. The bidding material classification module extracts data features from the bidding material image recognition results to obtain image feature data, substitutes the image feature data into a preset bidding material classification model to obtain classified bidding data; A bidding material matching module obtains bidding project demand information, matches the bidding project demand information with the classified bidding data, obtains bidding data matching results, the bidding data matching results include successfully matched data and unsuccessfully matched data, establishes a corresponding relationship between the successfully matched data and the bidding project demand information, retrieves the bidding material data corresponding to the unsuccessfully matched data, converts the bidding material data corresponding to the unsuccessfully matched data into text data, performs cluster analysis on the text data, obtains text data cluster analysis results, matches the text data cluster analysis results in a bidding data knowledge base, and obtains bidding project categories corresponding to the text data cluster analysis results; The bidding material inspection module substitutes the bidding project category corresponding to the text data cluster analysis result and the matching success data into the preset bidding material simulation test model to obtain the first bidding material simulation test result, which includes the detection success data and the detection failure data, and transmits the detection failure data to the bidder client and the backend control end; The bidding material review module receives feedback data from the bidder client after correcting the unsuccessful detection data, obtains the repaired bidding data, substitutes the repaired bidding data and the successfully matched data into the preset bidding material simulation test model, and obtains the second bidding material simulation test result. If the second bidding material simulation test result is qualified, the second bidding material simulation test result is marked as information to be reviewed and sent to the background control end.

2. The novel auxiliary bidding and tendering review system as claimed in claim 1 is characterized in that: The bidding material acquisition module is also used for: The server receives the bidding material data uploaded by the bidder’s client, including documents, pictures and PDF; Preprocess the received bidding materials, including file format conversion, image clarity optimization and image denoising; Use the image recognition algorithm of optical character recognition technology to perform image recognition on the pre-processed bidding materials, extract the text, numbers and symbol information in the image, and convert it into editable and searchable text data; After the image recognition is completed, the image recognition results of the bidding materials are generated. The image recognition results of the bidding materials include the recognized text content, text location information and elements in the image. The elements in the image include tables and charts.

3. The novel auxiliary bidding and tendering review system as claimed in claim 1 is characterized in that: The bidding material classification module is also used for: Performing text preprocessing on the text data in the bidding material image recognition results, the text preprocessing includes word segmentation, stop word removal and stemming text to obtain preprocessed text data; The pre-processed text data is converted into numerical feature vectors using word embedding technology, and the pre-trained bidding material classification model is loaded from the server's model library; The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model automatically classifies the bidding materials according to the input feature data and the learning mechanism within the bidding material classification model. The bidding material classification model outputs the classification results, which are the category labels to which the bidding materials are classified. The category labels to which the bidding materials are classified include technical solutions, financial statements, and project plans. Associating the classification results with the original bidding material data to form classified bidding data; The classified bidding data is stored in the server's database, and the classification results are provided to the bidder's client and the backend control terminal through an API interface for viewing and confirmation.

4. The novel auxiliary bidding and tendering review system as claimed in claim 1 is characterized in that: The bidding material matching module is also used for: Obtain demand information of the current bidding project from the backend control terminal or related database, including project background, technical requirements, performance indicators and qualification requirements; The acquired bidding project demand information is matched with the previously classified bidding data one by one, matching similarity calculation, keyword matching and semantic analysis are performed, and the bidding data is divided into successfully matched data and unsuccessfully matched data. The successfully matched data refers to the part of the bidding materials that matches the similarity with the bidding project demand information or meets the similarity requirements, and the unsuccessfully matched data refers to the part of the bidding materials that does not match the similarity with the bidding project demand information or does not meet the similarity requirements. A corresponding relationship is established between the successfully matched data and the corresponding bidding project demand information; Retrieving the original bidding material data corresponding to the unsuccessful matching part from the classified bidding data, and if the original bidding material data is in non-text form, converting it into editable text data; Apply cluster analysis algorithm to process the converted text data to obtain patterns, themes and categories. After the cluster analysis is completed, the cluster analysis results of the text data are generated. The cluster analysis results include different clusters, representative words or phrases in each cluster, and similarities and differences between clusters; Matching the text data cluster analysis results with the information in the bidding data knowledge base, which includes historical bidding data, successful cases, and industry norms; According to the matching process, the bidding project category corresponding to the text data clustering analysis result is determined, and the matching results, corresponding relationships, clustering analysis results and project categories are stored in the server's database and provided to the bidder's client or backend control terminal for viewing and confirmation via an API interface.

5. The novel auxiliary bidding and tendering review system as claimed in claim 1 is characterized in that: The bidding material inspection module is also used for: The bidding project categories corresponding to the text data cluster analysis results and the matching successful data are used as input information to load the preset bidding material simulation test model: Loading a bidding material simulation test model from a model library on a server, the bidding material simulation test model simulates the actual review process based on the input bidding data and outputs a test result; Substitute the bidding project category and the matching success data into the bidding material simulation test model, and input the text data and numerical data into the corresponding input layer of the model; The bidding material simulation test model is started, and the bidding material simulation test model processes the input bidding data, and the bidding material simulation test model generates simulation test results; After the bidding materials simulation test model is completed, the first bidding materials simulation test result is output. The first bidding materials simulation test result includes successful test data and unsuccessful test data. The successful test data refers to the part of the bidding materials that meets the preset standards of the bidding materials simulation test model, and the unsuccessful test data refers to the part of the bidding materials that does not meet the preset standards of the bidding materials simulation test model.

6. The novel auxiliary bidding and tendering review system as claimed in claim 1, characterized in that: The bidding material classification module is also used for: Acquire historical data of bidding material image recognition results, perform word segmentation, remove stop words, and perform stem extraction on text data in the historical data of bidding material image recognition results to obtain pre-processed historical text data; The preprocessed text history data is converted into numerical feature vectors using word embedding technology, and the random forest model is trained using the numerical feature vectors to obtain a bidding material classification model; The extracted image feature data is used as input and substituted into the loaded bidding material classification model. The bidding material classification model automatically classifies the bidding materials based on the input image feature data and the random forest learning algorithm within the model. The bidding material classification model outputs classification results, which are the category labels to which the bidding materials are classified. The category labels include technical solutions, financial statements, and project plans. The classification results are associated with the original bidding material data to form classified bidding data, which are stored in the database of the server.

7. The novel auxiliary bidding and tendering review system as claimed in claim 1 is characterized in that: The bidding material inspection module is also used for: Obtain the bidding project categories and matching success data corresponding to the historical text data clustering analysis results, start the bidding material simulation test model, and input the historical text data and numerical data into the input layer of the neural network model. The neural network model trains the input bidding data to obtain the bidding material simulation test model.

Images