Bid evaluation method and device, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention relates to the field of computer-aided bid evaluation, and discloses a bid evaluation method and device, electronic equipment, a storage medium and a program product, and the method comprises the steps: obtaining a pre-evaluation chain corresponding to a bid invitation file, the structure of the pre-evaluation chain is a directed acyclic graph composed of nodes and edges, the nodes represent evaluation elements in the bid invitation file, and the edges represent evaluation elements in the bid invitation file; the edge represents a logic dependency relationship among the review elements; obtaining an expert completion chain corresponding to the pre-review chain, wherein the structure of the expert completion chain is a directed acyclic graph; consistency verification is carried out on the expert completion chain to obtain a verification result, and the consistency verification comprises at least one of causal consistency verification, logic consistency verification and cross-chain consistency verification; and when the test result shows that the test is passed, solidifying the expert completion chain into a strong consistent review chain, and applying the strong consistent review chain to the bid evaluation process of the bid document. The method at least can achieve the effects of ensuring the internal consistency and cross-dimension coordination of the review standard and improving the review quality.

Description

Technical Field

[0001] This application relates to the field of computer-aided bid evaluation, and particularly to a bid evaluation method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] Bid evaluation is a crucial step in the bidding process for major projects such as engineering construction and government procurement. Traditional bid evaluation methods rely heavily on manual labor, requiring experts to thoroughly read the bidding documents, extract key evaluation points item by item, and compare them with the submitted bids. This process is time-consuming, inefficient, and prone to subjective bias and omissions.

[0003] In recent years, artificial intelligence technology has been applied to assist in bid evaluation, such as automatic extraction of key evaluation points based on natural language processing and comparison of bid documents based on knowledge graphs. However, the bid evaluation process usually involves multiple dimensions of review, including qualifications, technology, and business, with complex logical dependencies between these dimensions. Existing intelligent bid evaluation systems struggle to guarantee the internal consistency and cross-dimensional coordination of evaluation standards, making it difficult to ensure the quality of the evaluation. Summary of the Invention

[0004] The purpose of this application is to provide at least one evaluation method, device, electronic equipment, storage medium, and program product that can at least solve the technical problem of difficulty in ensuring evaluation quality, and at least achieve the effect of ensuring the internal consistency and cross-dimensional coordination of evaluation standards, thereby improving the evaluation quality.

[0005] To address the aforementioned technical problems, at least one embodiment of this application provides a bid evaluation method, comprising: obtaining a pre-review chain corresponding to a tender document, wherein the structure of the pre-review chain is a directed acyclic graph composed of nodes and edges, wherein the nodes represent review elements in the tender document, and the edges represent logical dependencies between review elements; obtaining an expert completion chain corresponding to the pre-review chain, wherein the structure of the expert completion chain is a directed acyclic graph; performing a consistency check on the expert completion chain to obtain a check result, wherein the consistency check includes at least one of causal consistency check, logical consistency check, and cross-chain consistency check; and when the check result indicates that the check has passed, solidifying the expert completion chain into a strongly consistent review chain, and applying the strongly consistent review chain to the bid evaluation process of the tender document.

[0006] By establishing a complete process—from pre-review chain to expert supplementation chain, consistency verification, and solidified execution—a standardized and traceable system has been built, from tender document parsing to final bid evaluation, systematically improving the quality of bid evaluation. Consistency verification avoids logical errors and circular dependencies caused by human negligence in traditional bid evaluation, ensuring the internal consistency and cross-dimensional coordination of review standards.

[0007] Furthermore, once solidified into a strongly consistent review chain, the strongly consistent review chain can be reused as a standardized review template, improving the efficiency of subsequent bid evaluation for different bid documents.

[0008] In some optional embodiments, obtaining the expert completion chain corresponding to the pre-review chain includes: providing an interactive interface to receive editing operations of the pre-review chain by review experts to generate the expert completion chain. The editing operations include at least one of adding or deleting nodes, modifying the logical relationship type or direction of edges, adding trigger conditions to nodes, splitting nodes, and merging multiple nodes.

[0009] With an intuitive interactive interface, experts can perform structured editing without understanding complex technical implementations. It supports a variety of editing operations, including common operations such as adding and deleting nodes, adjusting logical relationships, splitting and merging, to meet the different review habits of experts.

[0010] In some optional embodiments, the method further includes: during the editing operation, detecting whether a closed loop exists in the pre-review chain being edited; if the closed loop is detected, displaying prompts for the closed loop through the interactive interface.

[0011] Real-time loop detection during the editing process prevents experts from unintentionally introducing circular dependencies, reducing the workload of subsequent verification. Real-time prompts are displayed through an interactive interface, providing instant feedback to help experts quickly locate and correct problems, shortening the time required for experts to complete the chain of dependencies.

[0012] In some optional embodiments, obtaining the pre-review chain corresponding to the tender document includes: extracting review information from the tender document using n different language models and generating explanatory content, obtaining review information and explanatory content corresponding to each language model, wherein the review information includes review elements and logical dependencies between review elements, and the explanatory content describes the reasons for extracting the review information, where n is an integer greater than 1; performing multiple rounds of update processing on the review information and explanatory content corresponding to the n language models until the review information corresponding to the n language models is the same after the update, and taking the review information that the n language models reach consensus after the last round of update as the target review information; each The update process for the review information and explanatory content corresponding to the language model is the same. The update process for the i-th language model is as follows: obtain n-1 check results corresponding to the i-th language model, combine the n-1 check results, use the i-th language model to re-extract the review information from the tender document, and generate new explanatory content; where 1≤i≤n, each check result corresponds to another language model, and for each other language model, combined with the explanatory content corresponding to the i-th language model, perform error correction checks on the review information corresponding to the i-th language model to obtain check results including error correction results and explanations; construct the pre-review chain based on the target review information.

[0013] By cross-validating and iteratively updating multiple language models, the potential biases and errors of a single model are reduced, improving the accuracy of the extracted target review information. Furthermore, each model not only outputs review information but also generates explanatory content, making the extraction process more transparent, trustworthy, and interpretable, and facilitating error correction checks by other language models. Through multiple rounds of updates until the review information from all models is consistent, the final result ensures that it represents the shared understanding of multiple models, thus enhancing reliability.

[0014] In some optional embodiments, the method further includes optimizing the n language models respectively based on the strong consensus review chain.

[0015] By using high-quality review chains, which have been reviewed and solidified by experts, as training data, the business understanding capability of the language model is continuously improved, and closed-loop optimization forms a positive feedback loop. In addition, there is no need to specially label the training data; the annotations that are naturally generated in the business process can be used for model optimization, realizing a "business is annotation, and annotation is business" model.

[0016] In some optional embodiments, the causal consistency check is used to detect whether there is a circular dependency loop in the expert completion chain; the logical consistency check is used to detect whether there is a logical conflict in the logical dependencies in the expert completion chain; the cross-chain consistency check is used to check whether there is a decision conflict between different types of review chains, and the different types of review chains include at least two of the qualification review chain, technical review chain, business review chain, and additional review chain in the expert completion chain.

[0017] Three verification methods ensure the quality of the expert-completed chain from different dimensions, achieving comprehensive quality assurance. Among them, causal consistency prevents circular dependencies and ensures the feasibility of the review process; logical consistency prevents rule conflicts and ensures the rationality of review conclusions; and cross-chain consistency prevents dimensional contradictions and ensures overall review coordination.

[0018] At least one embodiment of this application also provides a bid evaluation device, comprising: a first acquisition module, configured to acquire a pre-review chain corresponding to a tender document, wherein the structure of the pre-review chain is a directed acyclic graph composed of nodes and edges, wherein the nodes represent review elements in the tender document, and the edges represent logical dependencies between review elements; a second acquisition module, configured to acquire an expert completion chain corresponding to the pre-review chain, wherein the structure of the expert completion chain is a directed acyclic graph; a verification module, configured to perform consistency verification on the expert completion chain and obtain a verification result, wherein the consistency verification includes at least one of causal consistency verification, logical consistency verification, and cross-chain consistency verification; and a bid evaluation module, configured to solidify the expert completion chain into a strongly consistent review chain and apply the strongly consistent review chain to the bid evaluation process of the tender document when the verification result indicates that the verification has passed.

[0019] In some optional embodiments, the second acquisition module is used to provide an interactive interface to receive editing operations from review experts on the pre-review chain in order to generate the expert completion chain. The editing operations include at least one of adding or deleting nodes, modifying the logical relationship type or direction of edges, adding trigger conditions to nodes, splitting nodes, and merging multiple nodes.

[0020] In some optional embodiments, the second acquisition module is further configured to detect whether there is a closed loop in the pre-review chain being edited during the editing operation; if the closed loop is detected, display prompts for the closed loop through the interactive interface.

[0021] In some optional embodiments, the first acquisition module is used to extract review information from the tender document using n different language models, and generate explanatory content, obtaining review information and explanatory content corresponding to each language model. The review information includes review elements and logical dependencies between review elements, and the explanatory content describes the reasons for extracting the review information, where n is an integer greater than 1. Multiple rounds of update processing are performed on the review information and explanatory content corresponding to the n language models until the updated review information corresponding to the n language models is the same. The review information that the n language models agree on after the last round of updates is taken as the target review information. Each language model... The update process for the corresponding review information and explanatory content is the same. The update process for the i-th language model is as follows: obtain n-1 check results corresponding to the i-th language model, combine the n-1 check results, use the i-th language model to re-extract the review information from the tender document, and generate new explanatory content; where 1≤i≤n, each check result corresponds to another language model, and for each other language model, combined with the explanatory content corresponding to the i-th language model, perform error correction checks on the review information corresponding to the i-th language model to obtain check results including error correction results and explanations; construct the pre-review chain based on the target review information.

[0022] In some optional embodiments, the apparatus further includes an optimization module for optimizing the n language models respectively based on the strong consistency review chain.

[0023] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described evaluation method.

[0024] At least one embodiment of this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described evaluation method.

[0025] At least one embodiment of this application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described evaluation method. Attached Figure Description

[0026] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0027] Figure 1This is a flowchart of a bid evaluation method provided in one embodiment of this application; Figure 2 This is a schematic diagram of a pre-review chain provided in one embodiment of this application; Figure 3 This is a schematic diagram of an expert supplement chain provided in one embodiment of this application; Figure 4 This is a schematic diagram of a bid evaluation device provided in another embodiment of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0029] Example 1: One method of this embodiment can be applied to electronic devices with communication, computing, and data storage capabilities, and its specific process can be as follows: Figure 1 As shown, it includes S101-S104.

[0030] S101, obtain the pre-review chain corresponding to the tender documents.

[0031] The structure of the pre-review chain is a directed acyclic graph consisting of nodes and edges. Nodes represent review elements in the tender documents, and edges represent logical dependencies between review elements.

[0032] Evaluation criteria are indicators set in the bidding documents based on the characteristics and requirements of the bidding project, used to measure the quality of bids. These indicators usually cover multiple dimensions such as technology, commerce, and price, and are the main basis for judging whether bids meet the bidding requirements and whether they are qualified to win the bid.

[0033] Logical dependency refers to the interrelationship and mutual constraint among evaluation elements. This relationship determines that during the bid evaluation process, the evaluation results of some evaluation elements may directly affect the evaluation of other elements, or the evaluation of some evaluation elements may need to be based on the evaluation results of other elements.

[0034] Taking the public bidding and evaluation process of the "Smart Library Construction Project" as an example, the main requirements of the bidding documents include: the bidder must have a Class II professional contracting qualification for building intelligent engineering (mandatory); the technical part includes reviewing the five-dimensional solution of building information model (10 points), intelligent lighting system (8 points), facial recognition access control system (6 points), and environmental monitoring system (6 points); the commercial part includes reviewing the total price (20 points) and construction period (10 points); and providing five years of free operation and maintenance can earn additional points (5 points).

[0035] In the example of the "Smart Library Construction Project", factors such as building intelligent engineering qualifications and building information model five-dimensional solutions are used for evaluation.

[0036] In some examples, obtaining the pre-review chain corresponding to the tender documents may include: extracting review information from the tender documents using a language model, whereby the review information includes review elements and the logical dependencies between review elements; and constructing the pre-review chain based on the review information.

[0037] One way to extract review information from tender documents using a language model is to construct a first hint, which is used to enable the language model to extract review elements and their logical dependencies from the given tender documents. Then, the first hint and the tender documents are input into the language model, and the language model outputs the review information.

[0038] The embodiments of this application do not limit the specific language model used. For example, DeepSeek, Wenxin Yiyan, GPT series, etc.

[0039] The embodiments of this application do not limit the specific text format of the first prompt. For example, the first prompt may be as follows: You are a tender document analysis expert who needs to extract the evaluation elements (i.e., the core indicators examined during bid evaluation) and their logical dependencies (i.e., the logical relationships between the evaluation elements) from a given tender document. Please complete this by following these steps: 1. Extract review elements: Carefully read the tender documents and identify all clauses related to the evaluation (such as technical requirements, commercial terms, pricing rules, etc.).

[0040] Each evaluation element is distilled into a concise phrase or keyword (e.g., "Technical feasibility", "Enterprise qualification level", "Reasonableness of bid price", "Additional requirements").

[0041] Classify by element type (e.g., technical, commercial, qualification, supplementary) and label with category tags (e.g., [technical], [commercial], [qualification], [supplementary]).

[0042] 2. Analyze logical dependencies Determine whether the following relationship types exist between the review elements, and indicate the direction of the relationship (e.g., A->B means A influences B): Sequential dependency: The evaluation of A requires B as a prerequisite (e.g., "evaluate the enterprise's qualifications first, then evaluate the technical solution"). Conditional dependence: The evaluation result of A serves as the evaluation condition for B (e.g., "If the enterprise's qualifications do not meet the standards, the technical solution will not be reviewed"). Influence Dependence: The evaluation result of A directly affects the score of B (e.g., "a high score for the technical solution may improve the cost-effectiveness score"). Complementary dependence: A's shortcomings can be compensated by B's strengths (e.g., "slightly weaker technical capabilities but excellent after-sales service, overall competitiveness is still relatively high").

[0043] Describe the relationships using concise language (e.g., "[Technical] technical solution feasibility -> impact dependence -> [Business] cost-effectiveness").

[0044] 3. Output format requirements Review Element List: List the review elements by category, in the format [Category] Element Name.

[0045] Logical dependency diagram: Connect related elements with arrow symbols (->) and label the relationship type (e.g., A->[influence dependency]->B).

[0046] In other examples, obtaining the pre-review chain corresponding to the tender document includes: extracting review information from the tender document using n different language models and generating explanatory content, obtaining the review information and explanatory content corresponding to each language model. The review information includes review elements and logical dependencies between review elements, and the explanatory content describes the reasons for extracting the review information, where n is an integer greater than 1; performing multiple rounds of updating the review information and explanatory content corresponding to the n language models until the review information corresponding to the n language models is the same after the update, and taking the review information that the n language models agree on after the last round of updates as the target review information; each The update process for the review information and explanatory content corresponding to each language model is the same. The update process for the i-th language model is as follows: obtain n-1 check results corresponding to the i-th language model, combine the n-1 check results, use the i-th language model to re-extract the review information from the tender document, and generate new explanatory content; where 1≤i≤n, each check result corresponds to another language model, and for each other language model, combined with the explanatory content corresponding to the i-th language model, perform error correction checks on the review information corresponding to the i-th language model to obtain check results including error correction results and explanatory content; construct a pre-review chain based on the target review information.

[0047] The method by which the language model first extracts review information from the tender documents and generates explanatory content can be to construct a second prompt. This second prompt is used to enable the language model to extract review information from the given tender documents and generate explanatory content. The specific text form of this second prompt is not limited in the embodiments of this application and can be set based on experience.

[0048] In subsequent iterations, the language model can extract review information from the tender documents and generate explanatory content by constructing a third prompt. This third prompt is used to allow the language model to combine n-1 inspection results to extract review information from the given tender documents and generate explanatory content. The specific text form of this third prompt is not limited in the embodiments of this application and can be set based on experience.

[0049] Taking three language models as an example, the process of obtaining the pre-review chain corresponding to the tender documents can be as follows: 1. Use three different language models (e.g., Model A, Model B, and Model C) to parse the tender documents simultaneously. Each model independently outputs its extracted "review information" (containing review elements and their logical dependencies) and "explanatory content" (explaining why it was extracted in this way).

[0050] 2. Entering multiple rounds of iterative updates, for Model A, the "check results" (containing the correction results and explanations, which explain why the correction results were generated) generated by Models B and C after correcting and checking Model A's "review information" based on Model A's "explanatory content" are fed back to Model A. Model A, based on these feedback check results, rereads the tender documents and outputs updated review information and explanatory content. Models B and C undergo a similar process.

[0051] 3. Repeat the above iterations until the "review information" output by the three models is completely consistent. At this point, a collaborative consensus is considered to have been reached. The "review information" at this point is taken as the "target review information".

[0052] 4. Automatically construct the pre-review chain (directed acyclic graph structure) based on the target review information.

[0053] In some cases, the pre-review chain can include node types such as qualifications, technical, business, and additional.

[0054] In other examples, the pre-review chain includes the type of influence corresponding to the edge, such as sequential dependency.

[0055] By cross-validating and iteratively updating multiple language models, the potential biases and errors of a single model are reduced, improving the accuracy of the extracted target review information. Furthermore, each model not only outputs review information but also generates explanatory content, making the extraction process more transparent, trustworthy, and interpretable, and facilitating error correction checks by other language models. Through multiple rounds of updates until the review information from all models is consistent, the final result ensures that it represents the shared understanding of multiple models, thus enhancing reliability.

[0056] S102, obtain the expert completion chain corresponding to the pre-review chain.

[0057] The structure of the expert-completed chain is a directed acyclic graph.

[0058] In some examples, obtaining the expert completion chain corresponding to the pre-review chain may include: providing an interactive interface to receive editing operations from review experts on the pre-review chain to generate the expert completion chain. The editing operations include at least one of adding or deleting nodes, modifying the logical relationship type or direction of edges, adding trigger conditions to nodes, splitting nodes, and merging multiple nodes.

[0059] The electronic device provides the evaluation experts with a graphical interactive interface that displays the preliminary review chain. Instead of modifying the text, the evaluation experts can edit this structured preliminary review chain.

[0060] Taking the public bidding and evaluation process of the "Smart Library Construction Project" in S101 as an example, the pre-review chain corresponding to this "Smart Library Construction Project" can be as follows: Figure 2 As shown, the editing operations performed by the evaluation experts on the pre-review chain corresponding to this "Smart Library Construction Project" include: Adding a node: Experts believe that network security is of paramount importance, therefore a mandatory node, "Level 3 Certification of Cybersecurity Classification Protection," has been added. Dependency Modification: Experts believed that the review of the "facial recognition system" should depend on the "environmental monitoring system," so the dependency relationship was modified, and a dependency relationship was established between the "facial recognition system" and the newly added "Level 3 Certification of Cybersecurity Classification Protection." Adjusting the logical relationship: Experts believe that the "intelligent lighting system" and the "environmental monitoring system" should be reviewed as a whole, so the two nodes are merged into an "AND" logic group; Additional trigger conditions: Experts set the trigger conditions for the "Five-Year Free Operation and Maintenance" bonus node: "Can only be obtained when the total technical score is greater than or equal to 20 points."

[0061] After the expert completed the editing, the result was as follows: Figure 3 The expert completion chain is shown.

[0062] With an intuitive interactive interface, experts can perform structured editing without understanding complex technical implementations. It supports a variety of editing operations, including common operations such as adding and deleting nodes, adjusting logical relationships, splitting and merging, to meet the different review habits of experts.

[0063] In some examples, this also includes: during the editing process, detecting whether there is a closed loop in the pre-review chain being edited; if a closed loop is detected, displaying prompts about the closed loop through the interactive interface.

[0064] The embodiments of this application do not limit how the prompt content is displayed. For example, the closed loop can be displayed using a preset color in the pre-review chain of the interface. Alternatively, the prompt content can be presented via a pop-up window, with the prompt content describing the closed loop in text format.

[0065] Real-time loop detection during the editing process prevents experts from unintentionally introducing circular dependencies, reducing the workload of subsequent verification. Real-time prompts are displayed through an interactive interface, providing instant feedback to help experts quickly locate and correct problems, shortening the time required for experts to complete the chain of dependencies.

[0066] S103, perform consistency verification on the expert-completed chain to obtain the verification result. The consistency verification includes at least one of causal consistency verification, logical consistency verification and cross-chain consistency verification.

[0067] Among them, causal consistency verification is used to detect whether there are circular dependency loops in the expert completion chain. Logical consistency verification is used to detect whether there are logical conflicts in the logical dependencies in the expert completion chain. Cross-chain consistency verification is used to check whether there are decision conflicts between different types of review chains, which include at least two of the following in the expert completion chain: qualification review chain, technical review chain, business review chain, and additional review chain.

[0068] The expert completion chain typically includes a qualification review section for the bidding entity, which is called the qualification review chain; the expert completion chain may also include a technical content review section, which is called the technical review chain; the expert completion chain also includes a business review section, which is called the business review chain; and the expert completion chain also includes a review section for attachments, which is called the supplementary review chain.

[0069] Three verification methods ensure the quality of the expert-completed chain from different dimensions, achieving comprehensive quality assurance. Among them, causal consistency prevents circular dependencies and ensures the feasibility of the review process; logical consistency prevents rule conflicts and ensures the rationality of review conclusions; and cross-chain consistency prevents dimensional contradictions and ensures overall review coordination.

[0070] In some cases, causal consistency checks can use graph traversal algorithms (such as depth-first search) to detect whether there are circular dependency loops in the expert completion chain.

[0071] In some cases, logical consistency checks can transform the logical relationships in the expert completion chain (such as "AND", "OR", "XOR" and triggering conditions) into satisfiability modular theory problems, which can then be determined using a solver. For example, checking whether there is a contradiction between the triggering condition "total technical score ≥ 20" and the calculation logic of the sum of scores for each technical item.

[0072] In some cases, cross-chain consistency checks examine the logic between chains of different review dimensions. For example, it verifies whether the conclusion (pass / fail) of the "qualification review chain" is correctly used as a prerequisite for the execution of the "technical review chain" and the "business review chain." In this example, it is necessary to ensure that if the qualification review chain determines that the qualification review chain fails, all subsequent scoring chains will not be executed.

[0073] S104 stipulates that when the inspection result indicates that the inspection has passed, the expert supplement chain is solidified into a strong consensus review chain, and the strong consensus review chain is applied to the bid evaluation process of the tender documents.

[0074] By solidifying the expert completion chain, a strongly consistent review chain is obtained. This strongly consistent review chain possesses irreversibility, traceability, enforceability (driving scoring), and can be used to constrain the output space during the bidding process.

[0075] In some cases, optimization can also be performed on each of the n language models based on a strong consistent review chain.

[0076] High-quality review chains, refined and solidified by experts, are used as training data to continuously improve the business understanding capabilities of the language models, creating a positive feedback loop through closed-loop optimization. Furthermore, no special annotation is required for training data; annotations naturally generated during the business process can be used for model optimization, achieving a "business is annotation, annotation is business" model. This enables the n language models to generate more accurate and business-relevant pre-review chains when processing similar tender documents in the future, thus achieving closed-loop optimization.

[0077] In some cases, if the consistency check in S103 fails, the expert completion chain is used as the pre-review chain, and a new expert completion chain for the pre-review chain is obtained.

[0078] Example 2: Another embodiment of this application relates to a bid evaluation device. The implementation details of the bid evaluation device in this embodiment are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this solution. A schematic diagram of the bid evaluation device in this embodiment can be seen as follows: Figure 4As shown, it includes: a first acquisition module 41, used to acquire the pre-review chain corresponding to the bidding documents. The structure of the pre-review chain is a directed acyclic graph composed of nodes and edges. The nodes represent the review elements in the bidding documents, and the edges represent the logical dependencies between the review elements; a second acquisition module 42, used to acquire the expert completion chain corresponding to the pre-review chain. The structure of the expert completion chain is a directed acyclic graph; a verification module 43, used to perform consistency verification on the expert completion chain and obtain the verification result. The consistency verification includes at least one of causal consistency verification, logical consistency verification, and cross-chain consistency verification; and an evaluation module 44, used to solidify the expert completion chain into a strongly consistent review chain when the verification result indicates that the verification has passed, and to apply the strongly consistent review chain to the evaluation process of the bidding documents.

[0079] In some optional embodiments, the second acquisition module 42 is used to provide an interactive interface to receive editing operations of the review experts on the pre-review chain in order to generate an expert completion chain. The editing operations include at least one of adding or deleting nodes, modifying the logical relationship type or direction of edges, adding trigger conditions to nodes, splitting nodes, and merging multiple nodes.

[0080] In some optional embodiments, the second acquisition module 42 is further configured to detect whether there is a closed loop in the pre-review chain being edited during the editing operation; if a closed loop is detected, display prompts for the closed loop through the interactive interface.

[0081] In some optional embodiments, the first acquisition module 41 is used to extract review information from the tender documents using n different language models, and generate explanatory content, thereby obtaining review information and explanatory content corresponding to each language model. The review information includes review elements and logical dependencies between review elements, and the explanatory content describes the reasons for extracting the review information, where n is an integer greater than 1. Multiple rounds of updating are performed on the review information and explanatory content corresponding to the n language models until the updated review information corresponding to the n language models is the same. The review information that the n language models agree on after the last round of updates is taken as the target review information. The update process for the review information and explanatory content corresponding to the language model is the same. The update process for the i-th language model is as follows: obtain n-1 check results corresponding to the i-th language model, combine the n-1 check results, use the i-th language model to re-extract the review information from the tender document, and generate new explanatory content; where 1≤i≤n, each check result corresponds to another language model, and for each other language model, combined with the explanatory content corresponding to the i-th language model, perform error correction checks on the review information corresponding to the i-th language model to obtain check results including error correction results and explanatory content; construct a pre-review chain based on the target review information.

[0082] In some optional embodiments, the apparatus further includes an optimization module 45 for optimizing each of the n language models based on a strong consensus review chain.

[0083] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0084] Example 3: Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the evaluation methods in the above embodiments.

[0085] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0086] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0087] Example 4: Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0088] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0089] Example 5: Another embodiment of this application relates to a computer program product, including a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0090] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A bid evaluation method, characterized in that, include: Obtain the pre-review chain corresponding to the tender document. The structure of the pre-review chain is a directed acyclic graph consisting of nodes and edges. The nodes represent the review elements in the tender document, and the edges represent the logical dependencies between the review elements. Obtain the expert completion chain corresponding to the pre-review chain, wherein the structure of the expert completion chain is a directed acyclic graph; A consistency check is performed on the expert-completed chain to obtain the check result. The consistency check includes at least one of causal consistency check, logical consistency check and cross-chain consistency check. When the test result indicates that the test has passed, the expert completion chain is solidified into a strong consensus review chain, and the strong consensus review chain is applied to the bid evaluation process of the tender documents.

2. The method according to claim 1, characterized in that, The step of obtaining the expert completion chain corresponding to the pre-review chain includes: An interactive interface is provided to receive editing operations from review experts on the pre-review chain in order to generate the expert completion chain. The editing operations include at least one of adding or deleting nodes, modifying the logical relationship type or direction of edges, adding trigger conditions to nodes, splitting nodes, and merging multiple nodes.

3. The method according to claim 2, characterized in that, Also includes: During the editing operation, it is detected whether there is a closed loop in the pre-review chain being edited; If the closed loop is detected, a prompt message for the closed loop will be displayed through the interactive interface.

4. The method according to claim 1, characterized in that, The process of obtaining the pre-review chain corresponding to the tender documents includes: The review information is extracted from the tender document using n different language models, and explanatory content is generated. The review information and explanatory content corresponding to each language model are obtained. The review information includes review elements and logical dependencies between review elements. The explanatory content is used to describe the reasons for extracting the review information. n is an integer greater than 1. The review information and explanatory content corresponding to the n language models are updated in multiple rounds until the review information corresponding to the n language models is the same after the update. The review information that the n language models reach the same conclusion after the last round of updates is taken as the target review information. The update process for the review information and explanatory content corresponding to each language model is the same. The update process for the i-th language model is as follows: obtain n-1 check results corresponding to the i-th language model, combine the n-1 check results, use the i-th language model to re-extract the review information from the tender document, and generate new explanatory content. Wherein, 1≤i≤n, each check result corresponds to another language model, and for each other language model, combined with the explanatory content corresponding to the i-th language model, the review information corresponding to the i-th language model is checked for error correction to obtain a check result including the error correction result and the explanatory content. The pre-review chain is constructed based on the target review information.

5. The method according to claim 4, characterized in that, Also includes: Based on the strong consistency review chain, the n language models are optimized respectively.

6. The method according to any one of claims 1-5, characterized in that, The causal consistency check is used to detect whether there are circular dependency loops in the expert completion chain; The logical consistency check is used to detect whether there are logical conflicts in the logical dependencies in the expert completion chain; The cross-chain consistency check is used to check whether there are decision conflicts between different types of review chains. The different types of review chains include at least two of the following in the expert completion chain: qualification review chain, technical review chain, business review chain, and additional review chain.

7. A bid evaluation device, characterized in that, include: The first acquisition module is used to acquire the pre-review chain corresponding to the tender document. The structure of the pre-review chain is a directed acyclic graph composed of nodes and edges. The nodes represent the review elements in the tender document, and the edges represent the logical dependencies between the review elements. The second acquisition module is used to acquire the expert completion chain corresponding to the pre-review chain, wherein the structure of the expert completion chain is a directed acyclic graph; The verification module is used to perform consistency verification on the expert completion chain and obtain the verification result. The consistency verification includes at least one of causal consistency verification, logical consistency verification and cross-chain consistency verification. The evaluation module is used to solidify the expert completion chain into a strong consensus review chain when the inspection result indicates that the inspection has passed, and to apply the strong consensus review chain to the evaluation process of the tender documents.

8. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the evaluation method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the evaluation method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the evaluation method according to any one of claims 1 to 6.

Images