Automatic review method for construction project bidding technical proposals

Abstract

The present application discloses a method for automatically evaluating construction engineering bidding technical solutions, which includes: obtaining multiple historical bidding technical solutions and their corresponding scores, adjusting at least part of the content in the copy of the historical bidding technical solution whose score exceeds the score threshold and reducing its corresponding score, and obtaining the adjusted bidding technical solution and its corresponding score; generating training labels corresponding to all historical bidding technical solutions and the adjusted bidding technical solution; wherein the training label at least includes the score of the bidding technical solution; using all historical bidding technical solutions and the adjusted bidding technical solution and their respective corresponding training labels, supervised training is performed on the bid evaluation model to obtain the trained bid evaluation model; wherein the trained bid evaluation model is used to automatically evaluate the bid technical solution to be scored. The above scheme can improve the bid evaluation efficiency.

Description

Technical Field

[0001] The present application relates to the technical field of bidding and tendering, and in particular to a method for automatically evaluating technical proposals for construction project bidding. Background Art

[0002] Bidding is a competitive procurement method commonly used in many industries such as construction projects. This method conforms to the laws of market operation, helps fair competition, and makes it possible to select the best at a low price. Therefore, it is widely used in the fields of bulk goods trading, contracting and subcontracting of engineering construction projects, and procurement and provision of service projects. In the bidding process, bidders need to provide bidding technical solutions to the tenderer, and the tenderer needs to score the bidding technical solutions. In the existing technology, the bid evaluation experts hired by the tenderer usually need to score the bidding technical solutions provided by the bidders. However, this scoring method is inefficient and it is difficult to ensure the stability of the quality of the bid evaluation and scoring, and it is also difficult to ensure the fairness and justice of the bid evaluation work. Therefore, how to improve the efficiency of bid evaluation has become an urgent problem to be solved. Summary of the invention

[0003] The present application provides a method for automatically evaluating construction project bidding technical solutions, which can improve the efficiency of bid evaluation.

[0004] A first aspect of an embodiment of the present application provides a method for automatically evaluating construction project bidding technical schemes, the method comprising: obtaining multiple historical bidding technical schemes and their corresponding scores, adjusting at least part of the contents in the copies of the historical bidding technical schemes whose scores exceed a score threshold and reducing their corresponding scores, to obtain adjusted bidding technical schemes and their corresponding scores; generating training labels corresponding to all the historical bidding technical schemes and the adjusted bidding technical schemes; wherein the training labels at least include the scores of the bidding technical schemes; using all the historical bidding technical schemes and the adjusted bidding technical schemes and their respective corresponding training labels, to perform supervised training on a bid evaluation model, to obtain the trained bid evaluation model; wherein the trained bid evaluation model is used to automatically evaluate the bidding technical schemes to be scored.

[0005] A second aspect of an embodiment of the present application provides an analysis device, which includes a processor, a memory, and a communication circuit. The processor is coupled to the memory and the communication circuit, respectively. Program data is stored in the memory. The processor implements the steps in any of the above methods by executing the program data in the memory.

[0006] A third aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer-readable storage medium can be executed by a processor to implement the steps in any of the methods described above.

[0007] The beneficial effects of the present application are as follows: the automatic evaluation method for construction project bidding technical solutions of the present application sets a score threshold, screens out historical bidding documents with higher scores, and then adjusts part of the contents of the copies of the historical bidding documents and reduces their corresponding scores to obtain the adjusted bidding technical solutions and their corresponding scores; then generates training labels corresponding to all historical bidding technical solutions and adjusted bidding technical solutions, and finally uses all historical bidding technical solutions and adjusted bidding technical solutions and their respective corresponding training labels to supervise and train the bid evaluation model to obtain the trained bid evaluation model, which is used to automatically evaluate the bidding technical solutions to be scored. The above method adds the scores and bidding technical solutions to the training process together, and the training model is scored. The trained model reduces the scores of bidding technical solutions that are more similar to historical bidding technical solutions. Therefore, the trained bid evaluation model can achieve end-to-end output of the scores of the bidding technical proposals to be scored, improve the bid evaluation efficiency, and when the current bidding technical proposals to be scored and the historical bidding technical proposals of the same bidder are input, it can also detect whether the current bidding technical proposal to be scored is just a simple modification of the historical bidding technical proposal before participating in the current bidding, so as to determine whether to lower the score and achieve a more reasonable scoring of the bidding technical proposals. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0009] Figure 1 This is a flow chart of an implementation method of an automatic review method for a construction project bidding technical solution for this application;

[0010] Figure 2 This is a flow chart of another implementation method of the automatic review method for construction project bidding technical solutions for this application;

[0011] Figure 3 This is a flow chart of another implementation method of the automatic review method for construction project bidding technical solutions for this application;

[0012] Figure 4 This is a flow chart of another implementation method of the automatic review method for construction project bidding technical solutions for this application;

[0013] Figure 5 This is a flow chart of another implementation method of the automatic review method for construction project bidding technical solutions for this application;

[0014] Figure 6 This is a schematic diagram of the structure of an embodiment of the analysis device of the present application;

[0015] Figure 7 This is a schematic diagram of the structure of an embodiment of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0017] The terms "system" and "network" are often used interchangeably in this article. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship. In addition, "many" in this article means two or more than two.

[0018] See also Figure 1 In one embodiment, the method for automatically reviewing a construction project bidding technical solution includes:

[0019] S110: Acquire multiple historical bidding technical solutions and their corresponding scores, adjust at least part of the content in the copies of the historical bidding technical solutions whose scores exceed the score threshold and reduce their corresponding scores, to obtain adjusted bidding technical solutions and their corresponding scores.

[0020] Specifically, historical bidding technical proposals are bound to their corresponding scores. Multiple historical bidding technical proposals and their corresponding scores can be pre-saved in the same bid evaluation system, or obtained by the bid evaluation system from other related terminals. Historical bidding technical proposals include construction organization design, supervision outline, etc.

[0021] Furthermore, by setting a score threshold, historical bidding technical solutions with higher scores are screened out, at least copies of historical bidding technical solutions with scores exceeding the score threshold are generated, and part of the content in the copy is adjusted to obtain an adjusted bidding technical solution, and at the same time, the score of the corresponding adjusted bidding technical solution is reduced based on the adjusted content.

[0022] In one application method, a copy of a historical bidding technical solution with a score exceeding a score threshold is generated, wherein the original score of the copy is the same as the corresponding historical bidding technical solution, part of the content in the copy is adjusted and the score of the copy is lowered to obtain an adjusted bidding technical solution and its corresponding score. The adjusted bidding technical solution only corresponds to the historical bidding technical solution with a score exceeding the score threshold, thereby improving the efficiency of subsequent training in the subsequent training process.

[0023] In another application method, copies of all historical bidding technical solutions are generated, wherein the original scores of the copies are the same as the corresponding historical bidding technical solutions. Part of the content in the copies is adjusted and the scores of the copies are lowered to obtain adjusted bidding technical solutions and their corresponding scores. The adjusted bidding technical solutions correspond to all historical bidding technical solutions, thereby improving the accuracy of subsequent training in the subsequent training process.

[0024] In another application method, a copy of the historical bidding technical solution with a score exceeding the score threshold and a partial score not exceeding the score threshold is generated, wherein the original score of the copy is the same as the corresponding historical bidding technical solution, and part of the content in the copy is adjusted and the score of the copy is lowered to obtain an adjusted bidding technical solution and its corresponding score. The adjusted bidding technical solution is related to all historical bidding technical solutions with a score exceeding the score threshold and a partial score not exceeding the score threshold, thereby taking into account both training efficiency and training accuracy in the subsequent training process.

[0025] In one application scenario, the scores of historical bidding technical solutions correspond to multiple preset score intervals, among which the one exceeding the score threshold is the highest interval, and then at least a copy of the historical bidding technical solution corresponding to the highest interval is generated.

[0026] Optionally, adjusting part of the contents of the copy includes at least one of modifying numbers, replacing keywords, and changing word order, and the reduced score corresponding to the copy is related to the adjusted content, thereby improving the rationality of the score corresponding to the adjusted bidding technical proposal.

[0027] S120: Generate training labels corresponding to all historical bidding technical solutions and adjusted bidding technical solutions; wherein the training labels at least include the scores of the bidding technical solutions.

[0028] Specifically, at least the scores of the bidding technical solutions are used to generate training labels corresponding to the bidding technical solutions, thereby obtaining training labels corresponding to all historical bidding technical solutions and adjusted bidding technical solutions.

[0029] In one application, the training labels only include the scores of the bidding technical solutions, which are used to supervise the bid evaluation model to reasonably score the bidding technical solutions.

[0030] In another application, the training labels include scores and adjustment flags. The adjustment flags indicate whether the historical bidding technical solutions have been adjusted, thereby supervising the training model to score the bidding technical solutions and determine whether the bidding technical solutions are directly modified from other materials.

[0031] S130: Using all historical bidding technical solutions and adjusted bidding technical solutions and their corresponding training labels, supervise the bid evaluation model to obtain a trained bid evaluation model; wherein the trained bid evaluation model is used to automatically evaluate the bidding technical solutions to be scored.

[0032] Specifically, the samples in the supervised training process are all historical bidding technical solutions and adjusted bidding technical solutions. Combined with the training labels, the trained bid evaluation model can be obtained. In supervised training, some common algorithms are usually used, such as linear regression, decision tree, support vector machine, etc.

[0033] In one application method, the training labels only include scores. The historical bidding technical solutions and the adjusted bidding technical solutions are input into the bid evaluation model to obtain the predicted scores of the bidding technical solutions by the bid evaluation model. The estimated losses are determined based on the deviation between the predicted scores and the scores included in the training labels. The parameters of the bid evaluation model are adjusted based on the estimated losses until the preset convergence conditions are met to obtain the trained bid evaluation model.

[0034] In another application method, the training labels include scores and adjustment marks. The historical bidding technical solutions and the adjusted bidding technical solutions are input into the bid evaluation model to obtain the predicted scores and predicted marks of the bidding technical solutions by the bid evaluation model. The score loss is determined based on the deviation between the predicted score and the score included in the training labels. The mark loss is determined based on the deviation between the predicted mark and the adjustment mark included in the training labels. The estimated loss is determined based on the score loss and the mark loss. The parameters of the bid evaluation model are adjusted based on the estimated loss until the preset convergence conditions are met to obtain the trained bid evaluation model.

[0035] In one application scenario, the preset convergence condition includes that the estimated loss is less than or equal to the loss threshold. Therefore, when the estimated loss is greater than the loss threshold, the process returns to the step of inputting the historical bidding technical solution and the adjusted bidding technical solution into the bid evaluation model until the preset convergence condition is met.

[0036] In another application scenario, the preset convergence conditions include that the number of training times reaches the number threshold and the estimated loss is less than or equal to the loss threshold. Therefore, when the estimated loss is greater than the loss threshold or the number of training times does not reach the number threshold, it returns to the step of inputting the historical bidding technical solution and the adjusted bidding technical solution into the evaluation model, and increases the number of training times until the preset convergence conditions are met.

[0037] It can be understood that the trained bid evaluation model can achieve end-to-end output of the scores of the bidding technical schemes to be scored, thereby improving the bid evaluation efficiency. Moreover, when the current bidding technical scheme to be scored and the historical bidding technical scheme of the same bidder are input, it can also detect whether the current bidding technical scheme to be scored is simply a simple modification of the historical bidding technical scheme before participating in the current bidding, thereby determining whether to lower the score and achieve a more reasonable score for the bidding technical scheme. The above model can assist the bid evaluation experts in reference, and can also be used as a basis for analyzing the actual scoring deviations of the bid evaluation experts, such as whether there are malicious low scores or abnormally high scores, etc., to measure the evaluation quality of the bid evaluation experts.

[0038] From the above content, it can be seen that the automatic evaluation method of the construction project bidding technical scheme of this application sets a score threshold to screen out historical bidding documents with higher scores, and then adjusts part of the content of the copy of the historical bidding document and reduces its corresponding score to obtain the adjusted bidding technical scheme and its corresponding score; then generates training labels corresponding to all historical bidding technical schemes and adjusted bidding technical schemes, and finally uses all historical bidding technical schemes and adjusted bidding technical schemes and their respective corresponding training labels to supervise the evaluation model to obtain the trained evaluation model, and the trained evaluation model is used to automatically evaluate the bidding technical schemes to be scored. The above method adds the score and the bidding technical scheme to the training process, and the training model is scored. The trained model reduces the score of the bidding technical scheme that is more similar to the historical bidding technical scheme. Therefore, the trained bid evaluation model can achieve end-to-end output of the scores of the bidding technical proposals to be scored, improve the bid evaluation efficiency, and when the current bidding technical proposals to be scored and the historical bidding technical proposals of the same bidder are input, it can also detect whether the current bidding technical proposal to be scored is just a simple modification of the historical bidding technical proposal before participating in the current bidding, so as to determine whether to lower the score and achieve a more reasonable scoring of the bidding technical proposals.

[0039] In one embodiment, step S130 uses all historical bidding technical solutions and adjusted bidding technical solutions and their corresponding training labels to supervise the bid evaluation model, and before obtaining the trained bid evaluation model, it also includes:

[0040] Obtain historical bidding conditions corresponding to historical bidding technical solutions, and use the historical bidding conditions as reference bidding conditions for the historical bidding technical solutions and their corresponding adjusted bidding technical solutions.

[0041] Specifically, the reference bidding conditions are historical bidding conditions, and the reference bidding conditions may be materials pre-saved in the same bid evaluation system, or materials acquired by the bid evaluation system from other associated terminals.

[0042] It should be noted that all the bidding conditions in this patent (historical bidding conditions, reference bidding conditions) do not refer to bidding qualification conditions, but refer to the existing objective conditions of the bidding project (such as the site conditions of the bidding project, water, electricity, access, network conditions, etc.), as well as the bidding project's personalized requirements for the bidding technical plan (construction organization design or supervision outline), mainly content related to project construction needs, such as whether the quality target is qualified or Lu Ban Award; whether the technical requirement adopts BIM modeling technology; whether the construction requirement adopts prefabricated construction; and whether the material requirement adopts some new energy-saving and environmentally friendly materials.

[0043] Bidding projects have different objective conditions and different personalized requirements, which have a great impact on the bidding technical plan. Therefore, each supervision training process of this application should be based on the same bidding conditions.

[0044] Step S130 uses all historical bidding technical solutions and adjusted bidding technical solutions and their corresponding training labels to supervise and train the bid evaluation model to obtain a trained bid evaluation model, including:

[0045] S140: Using all historical bidding technical solutions and their corresponding reference bidding conditions and training labels, as well as all adjusted bidding technical solutions and their corresponding reference bidding conditions and training labels, supervise the bid evaluation model and obtain a trained bid evaluation model.

[0046] Specifically, in addition to historical bidding technical plans and adjusted bidding technical plans, samples in the supervised training process also include reference bidding conditions. The reference bidding conditions are associated with historical bidding technical plans and adjusted bidding technical plans. The higher the degree of compliance between the bidding technical plan and the bidding conditions, the higher the score of the bidding technical plan. The bid evaluation model learns these associated samples, thereby improving the accuracy and generalization ability of the model scoring.

[0047] In one embodiment, the bid evaluation model includes a scoring network and an adjustment network.

[0048] See also Figure 2 Step S140 uses all historical bidding technical solutions and their corresponding reference bidding conditions and training labels, as well as all adjusted bidding technical solutions and their corresponding reference bidding conditions and training labels, to supervise and train the bid evaluation model to obtain a trained bid evaluation model, including:

[0049] S141: All historical bidding technical solutions and their corresponding reference bidding conditions, as well as all adjusted bidding technical solutions and their corresponding reference bidding conditions, are input into the scoring network to obtain the first scores corresponding to the historical bidding technical solutions and the adjusted bidding technical solutions respectively.

[0050] Specifically, the scoring network scores the conformity relationship between the historical bidding technical solution and the reference bidding conditions, and obtains a first score corresponding to the historical bidding technical solution; the scoring network scores the conformity relationship between the adjusted bidding technical solution and the reference bidding conditions, and obtains a first score corresponding to the adjusted bidding technical solution. The conformity relationship includes the matching degree between the materials and the bidding conditions, and may also include the understanding degree of the materials to the bidding conditions.

[0051] S142: taking each historical bidding technical solution and all its corresponding adjusted bidding technical solutions as training sample sets, inputting each training sample set into the adjustment network, and obtaining a second score corresponding to the adjusted bidding technical solution in each training sample set.

[0052] Specifically, the adjustment network adjusts the scores of the historical bidding technical solutions and the adjusted bidding technical solutions, trains the bid evaluation model, and outputs a second score corresponding to the adjusted bidding technical solution.

[0053] S143: Determine the estimated loss based on the first score and training label corresponding to the historical bidding technical solution and the second score and training label corresponding to the adjusted bidding technical solution.

[0054] Specifically, the estimated loss includes the difference between the training label and the first score corresponding to the historical bidding technical solution, and also includes the difference between the second score and the training label corresponding to the adjusted bidding technical solution. The estimated loss is an indicator to measure the gap between the model prediction result and the actual result. The smaller the estimated loss, the closer the model prediction result is to the actual result.

[0055] S144: Adjust the parameters of the bid evaluation model based on the estimated loss until the preset convergence conditions are met to obtain a trained bid evaluation model.

[0056] Specifically, after obtaining the estimated loss, the parameters of the bid evaluation model are adjusted based on the preset loss, and the process returns to the step of inputting all historical bidding technical solutions and their corresponding reference bidding conditions, as well as all adjusted bidding technical solutions and their corresponding reference bidding conditions, into the scoring network. By repeatedly calculating the loss and optimizing the parameters, the preset convergence conditions are met and the prediction performance is improved. The commonly used parameter adjustment is to optimize the model parameters based on the estimated loss using the gradient descent method to minimize the estimated loss.

[0057] The above method first uses the scoring network to score the adjusted bidding technical solution, and then uses the adjustment network to adjust the score of the adjusted bidding technical solution. The smaller the estimated loss of the training model between the training label and the second score, the more accurate the prediction score of the bid evaluation model.

[0058] In one embodiment, the scoring network performs the following steps on the input bidding technical solution and reference bidding conditions:

[0059] (a) Performing semantic analysis on the input bidding technical solution and reference bidding conditions to obtain a first semantic result corresponding to the bidding technical solution and a second semantic result corresponding to the reference bidding conditions.

[0060] Specifically, natural language semantic analysis is used on the bidding technical proposal and reference bidding conditions, including text classification, sentiment analysis and intent recognition, so as to parse and analyze the meanings of words, phrases, paragraphs and chapters in the bidding technical proposal and reference bidding conditions.

[0061] (b) Determine a first score corresponding to the input bidding technical solution based on the first semantic result and the matching degree between the first semantic result and the second semantic result.

[0062] Specifically, the quality of the bidding technical solution is analyzed based on the first semantic result, and the matching degree between the bidding technical solution and reference bidding conditions is analyzed based on the matching degree of the first semantic result and the second semantic result, wherein the first score is obtained based on the first semantic result and the matching degree, that is, the first score is positively correlated with the quality of the bidding technical solution itself and the matching degree between the bidding technical solution and the bidding conditions, thereby improving the accuracy of the first score.

[0063] In other implementations, you can also choose to conduct a topic analysis, that is, analyze whether the topics of the bidding technical solution and the bidding document correspond to each other as the basis for model scoring.

[0064] See also Figure 3 In one embodiment, step S142 adjusts the network to perform the following steps on the input training sample set:

[0065] S150: Obtain a first keyword set corresponding to each adjusted bidding technical solution in the training sample set, and a second keyword set corresponding to the historical bidding technical solution; wherein the first keyword set and the second keyword set include at least text keywords and numeric keywords.

[0066] Specifically, numbers related to the construction requirements of the project in the adjusted bidding technical scheme are extracted as digital keywords, and based on the first semantic result corresponding to each adjusted bidding technical scheme, text keywords corresponding to each adjusted bidding technical scheme are obtained to obtain a first keyword set. Similarly, numbers related to the construction requirements of the project in historical bidding technical schemes are extracted as digital keywords, and based on the first semantic result corresponding to each historical bidding technical scheme, text keywords corresponding to each historical bidding technical scheme are obtained to obtain a second keyword set.

[0067] S151: Based on the first keyword set and the second keyword set, and the first semantic result and the second semantic result, determine the similarity of each adjusted bidding technical solution in the training sample set to the historical bidding technical solution.

[0068] Specifically, the representation form of the first semantic result and the second semantic result is a vector, and based on the overlap rate of the first keyword set and the second keyword set, the similarity of each adjusted bidding technical solution in the training sample set to the historical bidding technical solution is determined.

[0069] Furthermore, when the matching degree between the first keyword set and the second keyword set is higher, the matching degree between the first semantic result and the second semantic result is higher, and the similarity between the adjusted bidding technical solution and the historical bidding technical solution is higher. At the same time, the bidding technical solution and the reference bidding conditions, that is, the first semantic result and the second semantic result, are introduced to train the model with relevant materials, so that a more accurate bid evaluation model can be obtained.

[0070] S152: In response to the similarity exceeding the similarity threshold, determining an adjustment score corresponding to adjusting the bidding technical solution based on the similarity; wherein the adjustment score is positively correlated with the similarity.

[0071] Specifically, a similarity threshold is set. The higher the similarity, the more closely the adjusted bidding technical solution matches the historical bidding materials, and the higher the adjustment score. In other words, the higher the similarity, the greater the score reduction for the adjusted bidding technical solution.

[0072] It can be understood that when the similarity does not exceed the similarity threshold, the corresponding adjusted bidding technical scheme will be regarded as the bidding technical scheme that appears for the first time, and the corresponding adjusted bidding technical scheme can directly use the score output by the scoring network, so as to objectively score the bidding technical scheme based on the quality of the bidding technical scheme itself and the matching degree between the bidding technical scheme and the bidding conditions. When the similarity exceeds the similarity threshold, the corresponding adjusted bidding technical scheme will be regarded as a bidding technical scheme that does not appear for the first time, that is, it is obtained by simply modifying the corresponding historical bidding technical scheme, so as to use the adjustment network to determine the adjustment score based on the similarity, thereby reducing the score of the corresponding adjusted bidding technical scheme, so that the evaluation model can score more accurately.

[0073] S153: Determine a second score corresponding to the adjusted bidding technical solution based on the difference between the first score corresponding to the adjusted bidding technical solution and the adjustment score.

[0074] Specifically, the second score is obtained by subtracting the adjustment score from the first score. The introduction of the adjustment score can make the bid evaluation model more accurate.

[0075] In the above method, the adjustment network determines the similarity of each adjusted bidding technical solution in the training sample set relative to the historical bidding technical solution, and reduces the score of the materials whose similarity exceeds the similarity threshold, so as to make the bid evaluation model more accurate and reduce the probability that the same bidder can still get a high score after having a historical bidding technical solution that has obtained a high score and continuing to use it after making simple modifications on the basis of the corresponding historical bidding technical solution.

[0076] In one embodiment, determining the estimated loss based on the first score and training label corresponding to the historical bidding technical solution and the second score and training label corresponding to the adjusted bidding technical solution includes:

[0077] (c) A first loss is obtained based on a first score and training label corresponding to the historical bidding technical solution, and a second loss is obtained based on a second score and training label corresponding to the adjusted bidding technical solution.

[0078] Specifically, the first loss is the difference between the training label and the first score, the second loss is the difference between the second score and the training label, and the loss is the difference between the true value and the predicted value.

[0079] (d) Obtain weights corresponding to the first loss and the second loss, perform weighted summation on the first loss and the second loss, and obtain an estimated loss.

[0080] Specifically, the estimated loss can be obtained by directly adding the first loss and the second loss, or by weighting the first loss and the second loss. By setting corresponding weights for the first loss and the second loss, the focus of model training can be selected, thereby more accurately evaluating the estimated loss.

[0081] In one application scenario, the sum of the weights corresponding to the first loss and the second loss is 1, and the first loss and the second loss correspond to the same weight value, so that the first loss and the second loss are weightedly summed to obtain the estimated loss, so as to coordinately adjust the scoring network and the adjustment network, so that the objective scoring and score adjustment capabilities of the trained scoring model are more balanced.

[0082] In another application scenario, the sum of the weights corresponding to the first loss and the second loss is 1, and the weight value corresponding to the first loss is greater than the weight value corresponding to the second loss, so that the first loss and the second loss are weightedly summed to obtain the estimated loss, so as to adjust the scoring network and the adjustment network based on the estimated loss, so that the trained scoring model has accurate and objective scoring capabilities, while taking into account the score adjustment capabilities.

[0083] In another application scenario, the sum of the weights corresponding to the first loss and the second loss is 1, and the weight value corresponding to the first loss is less than the weight value corresponding to the second loss, so that the first loss and the second loss are weightedly summed to obtain the estimated loss, so as to adjust the scoring network and the adjustment network based on the estimated loss, so that the trained scoring model has objective scoring capabilities, and the corresponding similarity judgment and score adjustment capabilities are more accurate.

[0084] See also Figure 4 In one embodiment, step S110 obtains a plurality of historical bidding technical solutions and their corresponding scores, and at least partially adjusts the contents of the copies of the historical bidding technical solutions whose scores exceed the score threshold and reduces their corresponding scores, to obtain the adjusted bidding technical solutions and their corresponding scores, including:

[0085] S111: Acquire multiple historical bidding technical solutions and their corresponding scores, and generate at least one copy bidding technical solution corresponding to the historical bidding technical solution whose score exceeds the score threshold.

[0086] Specifically, for historical bidding technical proposals, the system will identify them and convert them into readable text. For the text, charts and other contents containing certain information displayed in picture format, they can first be converted into text format through optical character recognition technology (OCR, Optical Character Recognition).

[0087] Furthermore, at least one copy of the bidding technical solution corresponding to the historical bidding technical solution with a score exceeding the score threshold is generated.

[0088] In one application scenario, multiple copies of bidding technical solutions corresponding to historical bidding technical solutions with scores exceeding a score threshold are generated, so that the historical bidding technical solutions with scores exceeding the score threshold and their corresponding multiple adjusted bidding technical solutions are used to train the bid evaluation model, so as to focus on training the bid evaluation model to judge whether the scores of high-scoring bidding technical solutions need to be adjusted, thereby improving the training efficiency of the model and reducing the probability that the same bidder can still get a high score when he continues to use a historical bidding technical solution that has obtained a high score after making simple modifications to the corresponding historical bidding technical solution.

[0089] In another application scenario, a copy of the bidding technical scheme corresponding to all historical bidding technical schemes is generated to reduce the number of adjusted bidding technical schemes generated, so as to use all historical bidding technical schemes and their corresponding adjusted bidding technical schemes to train the bid evaluation model, thereby training the bid evaluation model to objectively score the bidding technical schemes in all score ranges and then adjust the scores. While improving the model training efficiency, the trained bid evaluation model has the ability to reasonably adjust the scores of all ranges.

[0090] S112: Adjust part of the content of each copy bidding technical solution corresponding to the historical bidding technical solution in different adjustment methods to obtain at least one adjusted bidding technical solution; wherein the adjustment method includes adjusting at least one of the words and numbers in the copy bidding technical solution.

[0091] Specifically, all duplicate bidding technical solutions corresponding to the same historical bidding technical solution correspond to different adjustment methods, and the adjusted bidding technical solutions corresponding to multiple historical bidding technical solutions correspond to multiple adjustment methods, thereby improving the diversity of the adjusted bidding technical solutions.

[0092] Optionally, the adjustment method may also include semantic adjustment, generating an adjustment bid technical plan through text, number and semantic adjustment methods.

[0093] S113: Using the score corresponding to the historical bidding technical solution as the initial score of the adjusted bidding technical solution, and reducing the initial score based on the adjustment method corresponding to the adjusted bidding technical solution to obtain the score of the adjusted bidding technical solution; wherein each adjustment method corresponds to its own preset score reduction value.

[0094] Specifically, the initial score of each adjusted bidding technical solution is the same as its corresponding historical bidding technical solution, so the initial score is reduced according to the adjustment method corresponding to the adjusted bidding technical solution to obtain the score of the adjusted bidding technical solution, wherein each adjustment method corresponds to its own preset reduction score.

[0095] In one application scenario, the preset score reduction values ​​corresponding to the adjustment method of adjusting only numbers, the adjustment method of adjusting only texts, and the adjustment method of adjusting both texts and numbers decrease in descending order, so that the adjustment bidding technical solution that only adjusts numbers related to the construction needs of the project has the highest score reduction, thereby improving the rationality of the score of the adjustment bidding technical solution.

[0096] In a specific application scenario, each adjustment method corresponds to a score range, and the minimum value of the score range of the adjustment method that only adjusts numbers is greater than the maximum value of the score range of the adjustment method that only adjusts text; the minimum value of the score range of the adjustment method that only adjusts text is greater than the maximum value of the score range of the adjustment method that adjusts text and numbers, and a preset score reduction value is obtained from the corresponding score range based on the adjustment content corresponding to the adjustment method, wherein the preset score reduction value is negatively correlated with the number of adjusted numbers and positively correlated with the semantic similarity of the adjusted text, wherein the semantic similarity is obtained based on the semantics of the adjusted text and the semantics of the original text in the corresponding historical bidding technical proposal.

[0097] The above method can obtain a more accurate adjustment network by classifying the adjustment methods of the bidding technical solutions and setting corresponding scoring methods.

[0098] See also Figure 5 In one embodiment, in response to the trained bid evaluation model, the above-mentioned automatic evaluation method for construction project bidding technical solutions further includes:

[0099] S210: In response to obtaining the bidder's to-be-scored technical bidding solution for the current project, obtaining all scored technical bidding solutions of the bidder for historical projects within at least a preset time period before the current project.

[0100] S220: Input the bidding technical proposal to be scored and all scored bidding technical proposals into the bid evaluation model to obtain the target score corresponding to the bidding technical proposal to be scored; wherein the bid evaluation model is obtained after training based on the method in any of the above-mentioned implementation modes.

[0101] Specifically, the materials to be scored and the scored bidding technical solutions may be pre-stored in the bid evaluation system, or may be materials obtained by the bid evaluation system from other associated terminals.

[0102] By analyzing the bidder's technical proposal to be scored by the bid evaluation model obtained after training using the method in any of the above implementation modes, a more reasonable target score for the technical proposal to be scored can be obtained, and the scoring efficiency can be improved.

[0103] See also Figure 6 In one embodiment, the analysis device 200 includes a processor 210, a memory 220 and a communication circuit 230. The processor 210 is coupled to the memory 220 and the communication circuit 230 respectively. The memory 220 stores program data. The processor 210 implements a method as in any of the above embodiments by executing the program data in the memory 220. The detailed steps can be found in the above implementation mode and will not be repeated here.

[0104] See also Figure 7 In one embodiment, a computer-readable storage medium 301 stores a computer program 302, and the computer program 302 can be executed by a processor to implement a method as described in any of the above embodiments.

[0105] Among them, the computer-readable storage medium 301 can specifically be a device that can store the computer program 302, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, or it can also be a server that stores the computer program 302. The server can send the stored computer program 302 to other devices for execution, or it can also execute the stored computer program 302 by itself.

[0106] It should be noted that the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present implementation scheme.

[0107] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0108] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the various implementation methods of the present application.

[0109] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for automatically evaluating a construction project bidding technical solution, characterized in that: include: Acquire multiple historical bidding technical solutions and their corresponding scores, adjust at least part of the content in the copy of the historical bidding technical solution whose score exceeds the score threshold and reduce its corresponding score, to obtain an adjusted bidding technical solution and its corresponding score; wherein the score of the historical bidding technical solution corresponds to a plurality of preset score intervals, wherein the interval that exceeds the score threshold is the highest interval; Generate training labels corresponding to all the historical bidding technical solutions and the adjusted bidding technical solutions; wherein the training labels at least include the scores of the bidding technical solutions; Obtaining historical bidding conditions corresponding to the historical bidding technical solution, and using the historical bidding conditions as reference bidding conditions for the historical bidding technical solution and its corresponding adjusted bidding technical solution; Using all the historical bidding technical solutions and the adjusted bidding technical solutions and their corresponding training labels, the bid evaluation model is supervised and trained to obtain the trained bid evaluation model; specifically comprising: using all the historical bidding technical solutions and their corresponding reference bidding conditions and training labels, as well as all the adjusted bidding technical solutions and their corresponding reference bidding conditions and training labels, the bid evaluation model is supervised and trained to obtain the trained bid evaluation model; wherein the trained bid evaluation model is used to automatically evaluate the bidding technical solutions to be scored; The obtaining of multiple historical bidding technical solutions and their corresponding scores, adjusting at least part of the contents in the copies of the historical bidding technical solutions whose scores exceed the score threshold and reducing the corresponding scores, to obtain the adjusted bidding technical solution and its corresponding score, includes: obtaining multiple historical bidding technical solutions and their corresponding scores, generating at least a copy bidding technical solution corresponding to the historical bidding technical solution whose scores exceed the score threshold; adjusting part of the contents of each copy bidding technical solution corresponding to the historical bidding technical solution according to different adjustment methods, to obtain the adjusted bidding technical solution; wherein the adjustment method includes adjusting at least one of the characters and numbers in the copy bidding technical solution; using the score corresponding to the historical bidding technical solution as the initial score of the adjusted bidding technical solution, reducing the initial score based on the adjustment method corresponding to the adjusted bidding technical solution, to obtain the score of the adjusted bidding technical solution; wherein each of the adjustment methods corresponds to its own preset score reduction value; The bid evaluation model includes a scoring network and an adjustment network; the bid evaluation model is supervised and trained by using all the historical bidding technical solutions and their corresponding reference bidding conditions and training labels, as well as all the adjusted bidding technical solutions and their corresponding reference bidding conditions and training labels to obtain the trained bid evaluation model, including: Input all the historical bidding technical solutions and their corresponding reference bidding conditions, as well as all the adjusted bidding technical solutions and their corresponding reference bidding conditions, into the scoring network to obtain first scores corresponding to the historical bidding technical solutions and the adjusted bidding technical solutions; use each of the historical bidding technical solutions and all of the corresponding adjusted bidding technical solutions as training sample sets, input each of the training sample sets into the adjustment network to obtain a second score corresponding to the adjusted bidding technical solution in each of the training sample sets; determine an estimated loss based on the first score and the training label corresponding to the historical bidding technical solution, as well as the second score and the training label corresponding to the adjusted bidding technical solution; adjust the parameters of the bid evaluation model based on the estimated loss until a preset convergence condition is met, and obtain the trained bid evaluation model; The adjustment network determines the adjustment score corresponding to the adjusted bidding technical solution based on the semantics of the bidding technical solution and the reference bidding conditions, and the text keywords and numeric keywords corresponding to the adjusted bidding technical solution and the historical bidding technical solution; The determining of the estimated loss based on the first score and the training label corresponding to the historical bidding technical solution, and the second score and the training label corresponding to the adjusted bidding technical solution, includes: obtaining a first loss based on the first score and the training label corresponding to the historical bidding technical solution, and obtaining a second loss based on the second score and the training label corresponding to the adjusted bidding technical solution; obtaining weights corresponding to the first loss and the second loss, and performing weighted summation of the first loss and the second loss to obtain the estimated loss; In response to obtaining the trained bid evaluation model, the method also includes: in response to obtaining the bidder's bid technical scheme to be scored for this project, obtaining all the scored bid technical schemes of the bidder for historical projects within a preset time length before the current project; inputting the bid technical scheme to be scored and all the scored bid technical schemes into the bid evaluation model to obtain the target score corresponding to the bid technical scheme to be scored.

2. The method according to claim 1, characterized in that The scoring network performs the following steps on the input bidding technical solution and reference bidding conditions: Performing semantic analysis on the input bidding technical solution and reference bidding conditions to obtain a first semantic result corresponding to the bidding technical solution and a second semantic result corresponding to the reference bidding conditions; Based on the first semantic result and the matching degree between the first semantic result and the second semantic result, the first score corresponding to the input bidding technical solution is determined.

3. The method according to claim 2, characterized in that The adjustment network performs the following steps on the input training sample set: Obtaining a first keyword set corresponding to each of the bid adjustment technical solutions in the training sample set, and a second keyword set corresponding to the historical bid technical solutions; wherein the first keyword set and the second keyword set at least include text keywords and numeric keywords; Based on the first keyword set and the second keyword set, and the first semantic result and the second semantic result, determine the similarity of each of the bid adjustment technical solutions in the training sample set to the historical bid technical solutions; In response to the similarity exceeding a similarity threshold, determining an adjustment score corresponding to the adjusted bidding technical solution based on the similarity; wherein the adjustment score is positively correlated with the similarity; The second score corresponding to the adjusted bidding technical solution is determined based on the difference between the first score corresponding to the adjusted bidding technical solution and the adjustment score.

4. An automatic review device for construction project bidding technical solutions, characterized in that: The device includes a processor, a memory and a communication circuit, the processor is coupled to the memory and the communication circuit respectively, the memory stores program data, and the processor implements the steps in the method as described in any one of claims 1 to 3 by executing the program data in the memory.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program can be executed by a processor to implement the steps in the method according to any one of claims 1 to 3.

Images