Intelligent bid evaluation method and system based on large model, and related product

Abstract

The invention discloses an intelligent bid evaluation method and system based on a large model, and a related product, relates to the technical field of purchase bid evaluation systems, and solves the problems that an existing intelligent bid evaluation system is difficult to understand a complex logical relationship and semantic connotation and cannot provide a comprehensive and accurate scoring result. According to the technical scheme, the method comprises the steps that S1, bidding and tendering files are obtained; s2, performing disassembly processing based on the bidding and tendering file to obtain a plurality of bidding and tendering data pairs; s3, performing instruction construction on the bidding and tendering data pair based on a historical bidding and tendering database to obtain a prompt instruction of the bidding and tendering data pair; s4, processing the prompt instruction of the bidding and tendering data pair based on the large model after training and fine tuning to obtain the score of the bidding and tendering data pair, and accumulating the scores to obtain the score of the bidding file; through the method based on the large model, the scoring effect is effectively improved in the intelligent bid evaluation work, the efficiency is improved, manpower and material resources are saved, and meanwhile interference caused by human factors is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of procurement bid evaluation systems, and more specifically, to an intelligent bid evaluation method, system and related products based on a large model. Background Art

[0002] In the current bidding process, the tenderer usually entrusts a third-party organization to publish electronic bidding documents through the bidding system. The bidder prepares electronic bidding materials according to the content of the bidding documents and submits them in the bidding system. The bidding materials usually include the bid and qualification certificate. The third-party organization organizes industry experts to form a bid evaluation committee. The committee reviews the qualifications and bids of the bidders according to the provisions of the bidding documents, scores the bids after ensuring that they meet the bidding conditions, and finally selects the winner from all bidders based on the scores.

[0003] The bid evaluation content can be roughly divided into three categories according to its nature: objective passability review, objective scoring review, and subjective scoring review. During the bid evaluation process, industry experts need to score the bid evaluation documents item by item according to the requirements of the bidding documents, and finally accumulate the scores to get the final score. This manual comparison process requires experts to invest a lot of time and energy, and it is inevitable that there will be wrong evaluations and missed judgments.

[0004] Intelligent bid evaluation integrates cutting-edge technologies such as natural language processing and machine learning to form a complete electronic bidding and tendering support system. The core task of the system is to conduct in-depth comparison and analysis of massive electronic bidding documents and bidding documents, and automatically score them according to the scoring criteria preset in the bidding documents. Intelligent bid evaluation has significant advantages. First, it improves the efficiency of bid evaluation; second, it effectively reduces human errors and biases; in addition, intelligent bid evaluation also improves the transparency of scoring, making the entire bid evaluation process more open and fair, thereby enhancing the trust in the bid evaluation results.

[0005] Although existing intelligent bid evaluation technology has made significant progress in automated processing, it still has certain limitations. At present, intelligent bid evaluation systems are not good enough when dealing with in-depth bid evaluation clauses. Due to the limitations of algorithms and models, existing systems often have difficulty understanding complex logical relationships and semantic connotations, and cannot provide comprehensive and accurate scoring results.

[0006] Therefore, the present application provides a large-model-based intelligent bid evaluation method, system and related products to solve the above problems. Summary of the invention

[0007] The purpose of this application is to provide an intelligent bid evaluation method, system and related products based on a large model to solve the problem that the existing intelligent bid evaluation system has difficulty understanding complex logical relationships and semantic connotations and cannot provide comprehensive and accurate scoring results; this application is based on large model technology to achieve intelligent bid evaluation, enhance the understanding of complex logical relationships and semantic connotations, and achieve more comprehensive and accurate bid evaluation.

[0008] The present application first provides a large-model intelligent bid evaluation method, which includes: S1, obtaining bidding documents; S2, disassembling and processing based on the bidding documents to obtain multiple bidding data pairs; S3, constructing instructions for the bidding data pairs based on the historical bidding database to obtain prompt instructions for the bidding data pairs; S4, processing the prompt instructions of the bidding data pairs based on the large model after training and fine-tuning to obtain the scores of the bidding data pairs, and accumulating the scores to obtain the scores of the bid documents.

[0009] In a possible implementation, S2, performing decomposition processing based on the bidding documents to obtain multiple bidding data pairs; including: S21, extracting all bidding terms and corresponding scoring criteria from the bidding documents based on the bidding document decomposition model; S22, querying the bidding response content corresponding to the bidding terms in the bidding documents based on the retrieval enhancement method; S23, forming bidding data pairs based on the bidding terms and the corresponding bidding response content.

[0010] In a possible implementation, the bid response content corresponding to the tender terms is searched in the bid document based on the retrieval enhancement method; the method includes: dividing the bid document into multiple small text segments according to semantics; converting the small text segments into vector form, and storing them in a database that supports vector retrieval and full-text keyword retrieval; obtaining the tender terms to be queried, and searching the database for candidate bid response content based on vector retrieval and full-text keyword retrieval; and filtering out the bid response content with the highest relevance to the tender terms from the candidate bid response content based on a text relevance model as the bid response content corresponding to the tender terms.

[0011] In a possible implementation, S3, constructing instructions for bidding data pairs based on a historical bidding database to obtain prompt instructions for the bidding data pairs; including: based on the bidding data pairs, querying historical bidding data pairs with a similarity greater than a threshold from the historical bidding database as candidate data pairs; inputting the candidate data pairs into a text similarity evaluation model, and selecting the candidate data pairs with the greatest similarity to the bidding data pairs as examples; generating prompt instructions including role definitions, task descriptions, specific requirements and example content according to the bidding data pairs, the scoring criteria for the bidding data pairs, the examples, and the scoring criteria for the examples.

[0012] In a possible implementation, S4, based on the trained and fine-tuned big model, the prompt instructions of the bidding data pairs are processed to obtain the scores of the bidding data pairs, and the scores are accumulated to obtain the scores of the bidding documents; including: inputting the prompt instructions of the bidding data pairs into the trained and fine-tuned big model to obtain the big model score description text; inputting the big model score description texts of all bidding data pairs into the trained and fine-tuned big model to obtain the scores of all bidding data pairs; accumulating the scores of all bidding data pairs to obtain the scores of the bidding documents.

[0013] In a possible implementation, the historical bidding database is obtained through the following steps: obtaining historical bidding documents; performing disassembly processing based on the historical bidding documents to obtain bidding terms, bidding response contents and corresponding scores; constructing historical bidding data pairs based on the bidding terms and corresponding bidding response contents; storing the historical bidding data pairs and corresponding scores in a standardized format in a database that supports vector retrieval and full-text keyword retrieval to form a historical bidding database.

[0014] In a possible implementation, the trained and fine-tuned big model is obtained through the following steps: selecting a basic big model; based on the historical bidding database as the data set for the big model thinking chain reasoning, using the LoRA fine-tuning technology to fine-tune the basic big model to obtain the trained and fine-tuned big model.

[0015] The present application also provides a big model-based intelligent bid evaluation system, which is used to implement any possible big model-based intelligent bid evaluation method mentioned above, including: a file acquisition module, used to acquire bidding documents; a file disassembly module, used to perform disassembly processing based on the bidding documents to obtain multiple bidding data pairs; a prompt instruction generation module, used to construct instructions based on the bidding data pairs to obtain prompt instructions for the bidding data pairs; a big model scoring module, used to process the prompt instructions of the bidding data pairs based on the trained and fine-tuned big model to obtain the scores of the bidding data pairs, and accumulate the scores to obtain the scores of the bidding documents.

[0016] The present application also provides a computer storage medium, including computer instructions, which, when executed on a computer device, enable the computer device to execute any of the above-mentioned possible large-model-based intelligent bid evaluation methods.

[0017] The present application also provides a computer program product. When the computer program product runs on a computer device, the computer device executes any possible large model-based intelligent bid evaluation method described above.

[0018] Compared with the prior art, the present application has the following beneficial effects: 1. Improved efficiency: The large model can quickly process and analyze a large number of bidding documents, greatly improving the efficiency of scoring; 2. Enhanced accuracy: The large model learns massive Chinese data during training and learns data in the bidding field during fine-tuning, so it can accurately grasp industry norms, terminology and evaluation standards to ensure the accuracy of the scoring results; 3. Enhanced fairness: The large model can make fair judgments and calculations, avoiding bias and injustice caused by human factors, and effectively reducing the occurrence of human errors and bias; 4. Improved transparency: Intelligent bid evaluation improves the transparency of scoring, making the entire bid evaluation process more open and fair, thereby enhancing the trust in the bid evaluation results; 5. Reduced costs: By reducing manual intervention, the labor cost and time cost in the bid evaluation process are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 It is a flowchart of the intelligent bid evaluation method based on the big model; Figure 2 This is the structural diagram of the intelligent bidding evaluation system based on the big model. DETAILED DESCRIPTION

[0020] Hereinafter, the term "include" or "may include" used in various embodiments of the present application indicates the presence of the function, operation or element applied for, and does not limit the increase of one or more functions, operations or elements. In addition, as used in various embodiments of the present application, the terms "include", "have" and their cognates are intended only to indicate specific features, numbers, steps, operations, elements, components or a combination of the foregoing items, and should not be understood as first excluding the presence of one or more other features, numbers, steps, operations, elements, components or a combination of the foregoing items or the possibility of increasing one or more features, numbers, steps, operations, elements, components or a combination of the foregoing items.

[0021] The terms used in the various embodiments of the present application are only used for the purpose of describing specific embodiments and are not intended to limit the various embodiments of the present application. As used herein, the singular form is intended to also include the plural form, unless the context clearly indicates otherwise. Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the meanings commonly understood by ordinary technicians in the field to which the various embodiments of the present application belong. The terms (such as those defined in a generally used dictionary) will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning, unless clearly defined in the various embodiments of the present application.

[0022] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with examples and drawings. The illustrative implementation scheme of the present application and its description are only used to explain the present application and are not intended to limit the present application.

[0023] See also Figure 1 As shown, Figure 1 The flowchart of the intelligent bid evaluation method based on the big model is shown. The method includes: S1, obtaining the bidding documents; S2, disassembling and processing based on the bidding documents to obtain multiple bidding data pairs; S3, constructing instructions for the bidding data pairs based on the historical bidding database to obtain prompt instructions for the bidding data pairs; S4, processing the prompt instructions for the bidding data pairs based on the big model after training and fine-tuning to obtain the scores of the bidding data pairs, and accumulating the scores to obtain the scores of the bidding documents.

[0024] Specifically, the intelligent bid evaluation method based on the big model includes four parts: obtaining bidding documents, disassembling bidding documents, constructing prompt instructions and scoring the big model, and finally obtaining the score of the bidding document. The improvement of the present invention is that it realizes intelligent bid evaluation based on the big model technology, enhances the understanding of complex logical relationships and semantic connotations, realizes more comprehensive and accurate bid evaluation, effectively improves the scoring effect and efficiency in the intelligent bid evaluation work, saves manpower and material resources, and avoids interference caused by human factors, and has broad application prospects and market value.

[0025] In a possible implementation, S2, performing decomposition processing based on the bidding documents to obtain multiple bidding data pairs; including: S21, extracting all bidding terms and corresponding scoring criteria from the bidding documents based on the bidding document decomposition model; S22, querying the bidding response content corresponding to the bidding terms in the bidding documents based on the retrieval enhancement method; S23, forming bidding data pairs based on the bidding terms and the corresponding bidding response content.

[0026] Specifically, step S2 is the disassembly of the bidding documents. Specifically, the bidding documents and their corresponding bidding documents are carefully analyzed to form a series of bidding data pairs, which cover key information such as bidding terms, scoring criteria and bidding response content. Each bidding data pair consists of a bidding term and its corresponding bidding response content. First, it is necessary to extract all bidding terms and corresponding scoring criteria from the bidding documents using the bidding document disassembly model. The bidding document disassembly model is a model trained based on labeled data and belongs to the information extraction model. Its main function is to extract bidding terms and scoring criteria from the documents. Then, the retrieval enhancement method is used to find the bidding response content corresponding to these bidding terms in the bidding documents. Finally, the bidding terms and their corresponding bidding response content constitute a bidding data pair.

[0027] Furthermore, based on the retrieval enhancement method, the bid response content corresponding to the tender terms is searched in the bid document; including: dividing the bid document into multiple small text segments according to semantics; converting the small text segments into vector form, and storing them in a database that supports vector retrieval and full-text keyword retrieval; obtaining the tender terms to be queried, and searching the database for candidate bid response contents based on vector retrieval and full-text keyword retrieval; based on the text relevance model, the bid response content with the highest relevance to the tender terms is screened out from the candidate bid response contents as the bid response content corresponding to the tender terms.

[0028] Specifically, the retrieval enhancement in step S2. First, the bidding documents to be retrieved are divided into small segments according to semantics, and then these small segments are converted into vector form and stored in a database that supports vector retrieval and full-text keyword retrieval. In this way, when a bidding clause is input, several candidate bidding response contents can be retrieved from the database. Subsequently, the relevance between these candidate contents and the bidding clauses is further verified using the text relevance model, and finally the bidding response content with the highest relevance is selected and paired with the corresponding bidding clause to form a bidding data pair.

[0029] It should be noted that (Chinese) semantic segmentation is a text processing technology that segments text according to a preset length standard based on a comprehensive consideration of the text's own format, punctuation, and deep semantic information. This process ensures that the text segments obtained from each segmentation not only do not exceed the specified length limit, but also maintain a certain length of overlap with adjacent text segments. This segmentation method not only maintains control over the physical length of the text segment, but more importantly, ensures the semantic integrity and coherence of each segment. This semantic integrity is extremely beneficial to the subsequent information retrieval steps, as it helps to improve the accuracy and efficiency of the retrieval and ensures that the retrieval system can better understand the key information and contextual relationships in the text.

[0030] In addition, in the disassembly of bidding documents, text vectorization can use a special Chinese vectorization model, which can convert the input text into a dense vector of a specific length. The disassembly of bidding documents can be completed fully automatically or combined with manual operations to ensure the accuracy and efficiency of the disassembly.

[0031] In a possible implementation, S3, constructing instructions for bidding data pairs based on a historical bidding database to obtain prompt instructions for the bidding data pairs; including: based on the bidding data pairs, querying historical bidding data pairs with a similarity greater than a threshold from the historical bidding database as candidate data pairs; inputting the candidate data pairs into a text similarity evaluation model, and selecting the candidate data pairs with the greatest similarity to the bidding data pairs as examples; generating prompt instructions including role definitions, task descriptions, specific requirements and example content according to the bidding data pairs, the scoring criteria for the bidding data pairs, the examples, and the scoring criteria for the examples.

[0032] Specifically, step S3 is the construction of prompt instructions. Prompt instructions play a key role in the big model. They can guide the big model to complete specific tasks and generate expected outputs. Prompt instructions consist of several key parts: role setting, task description, specific requirements, and related examples. In order to ensure that the big model can accurately understand and perform tasks, it is necessary to select closely related examples based on different bidding data pairs. These examples can be obtained from the historical bidding database through a hybrid retrieval method, and the higher the similarity between the example and the input bidding data, the more significant the effect will usually be.

[0033] Specifically, after inputting a set of bidding data pairs, similar historical bidding data pairs are found in the historical bidding database as references. This retrieval process can use neighbor-based spatial vector retrieval or keyword-based retrieval methods. After a series of candidate data pairs are retrieved, their similarity needs to be further verified. These candidate data pairs are input into a text similarity evaluation model with semantic interaction capabilities to evaluate their similarity with the input bidding data pairs, and finally the most similar data pairs are selected as important examples for subsequent large model scoring. If the retrieval result is empty, that is, no suitable historical data pairs are found, then in this case, the example will not be used to construct the prompt instruction.

[0034] The following are examples of prompt commands: “ You are a mathematician and intelligent bid evaluation expert who is good at mathematical calculations and logical reasoning; Please rate the review content according to the following review criteria; It requires thinking step by step, thinking rigorously, and ensuring that the calculations are accurate; Please refer to the example I gave: <Example> <Evaluation Criteria> **** <Evaluation Criteria> <Review content> **** <Review content> Your answer: ”.

[0035] Understandably, prompts that include role definitions, task descriptions, specific requirements, and example content can ensure that the large model can refer to the most relevant historical examples and scoring criteria when scoring.

[0036] In a possible implementation, S4, based on the trained and fine-tuned big model, the prompt instructions of the bidding data pairs are processed to obtain the scores of the bidding data pairs, and the scores are accumulated to obtain the scores of the bidding documents; including: inputting the prompt instructions of the bidding data pairs into the trained and fine-tuned big model to obtain the big model score description text; inputting the big model score description texts of all bidding data pairs into the trained and fine-tuned big model to obtain the scores of all bidding data pairs; accumulating the scores of all bidding data pairs to obtain the scores of the bidding documents.

[0037] Specifically, step S4 is the large model scoring. After receiving the prompt instruction, the large model will think and calculate according to the established logic, and finally output its analysis results, that is, the large model scoring description text. However, since the output of the large model may not follow the preset format, the specific score is extracted from the text by calling the model for the second time, and finally the scores are comprehensively accumulated to ensure the accuracy and consistency of the scoring results.

[0038] Specifically, the output of the big model usually includes its detailed analysis process of the bidding data pair and the final scoring results. In order to ensure the accuracy of the scoring, it is necessary to extract and standardize the scoring information from these output contents. This process also uses the big model to extract the results. The prompt instructions at this time are as follows: “ You are an AI assistant that excels at text summarization and information extraction; Please extract the score from the following text and output it in json format; Text: {$Large model rating description text} ”.

[0039] After extracting the scores of all bidding data pairs, the scores of each bidding data pair are accumulated, and the total score is the final score of the bidding document. This final score comprehensively reflects the performance of the bidding document in various bidding terms and can serve as an important basis for evaluating the quality of the bidding document.

[0040] It should be noted that before carrying out the above process, it is necessary to establish a historical bidding database and fine-tune the large model on the bidding data in advance. Extract historical bidding data pairs from historical bidding documents and add corresponding scores to save them in the database. Based on the disassembled historical bidding data pairs and scores as data sets, LoRA technology is used to fine-tune the large model to adapt it to the bidding data pair scoring task.

[0041] In a possible implementation, the historical bidding database is obtained through the following steps: obtaining historical bidding documents; performing disassembly processing based on the historical bidding documents to obtain bidding terms, bidding response contents and corresponding scores; constructing historical bidding data pairs based on the bidding terms and corresponding bidding response contents; storing the historical bidding data pairs and corresponding scores in a standardized format in a database that supports vector retrieval and full-text keyword retrieval to form a historical bidding database.

[0042] Specifically, the historical bidding documents are disassembled using the bidding document disassembly model to extract independent bidding terms, bidding response content and corresponding scores. Then, these bidding terms are paired with their corresponding bidding response content to form historical bidding data pairs. Finally, these data pairs and their scores are stored in a database in a standardized format to support efficient semantic retrieval and keyword retrieval functions.

[0043] In a possible implementation, the trained and fine-tuned big model is obtained through the following steps: selecting a basic big model; based on the historical bidding database as the data set for the big model thinking chain reasoning, using the LoRA fine-tuning technology to fine-tune the basic big model to obtain the trained and fine-tuned big model.

[0044] Specifically, open source models such as Qwen2-72b can be selected as the basic large model. Based on the historical bidding data pairs and scoring information organized in the historical bidding database, a data set suitable for large model thinking chain reasoning is constructed. Subsequently, the low-rank adaptation (LoRA) fine-tuning technology is used to fine-tune the basic large model, and multiple (10 rounds) of fine-tuning training are performed to achieve the best effect.

[0045] It should be noted that the big model can continuously learn new bidding data during the actual bid evaluation process, and fine-tune the big model by dynamically updating the historical bidding database to improve the accuracy of the scoring and ensure that it can adapt to the ever-changing bidding environment and evaluation standards.

[0046] In addition, in the subsequent design, considering the sensitive data in the bidding process, advanced data encryption and access control technologies can be adopted to ensure the security and privacy of data and prevent data leakage and unauthorized access.

[0047] It is understandable that the present invention aims to use artificial intelligence, especially large model technology, to achieve intelligent bid evaluation, automatically or semi-automatically complete the bid evaluation work, thereby improving the efficiency, accuracy and fairness of the overall bidding and procurement process. The specific effects are as follows: 1. Improve efficiency: The large model can quickly process and analyze a large number of bidding documents, greatly improving the efficiency of scoring; 2. Enhanced accuracy: The large model learns massive Chinese data during training and learns data in the bidding field during fine-tuning, so it can accurately grasp industry norms, terminology and evaluation standards to ensure the accuracy of the scoring results; 3. Enhanced fairness: The large model can make fair judgments and calculations, avoiding bias and injustice caused by human factors, and effectively reducing the occurrence of human errors and biases; 4. Improve transparency: Intelligent bid evaluation improves the transparency of scoring, making the entire bid evaluation process more open and fair, thereby enhancing the trust in the bid evaluation results; 5. Reduce costs: By reducing manual intervention, the labor cost and time cost in the bid evaluation process are reduced.

[0048] See also Figure 2 As shown, the embodiment of the present application also provides an intelligent bid evaluation system based on a large model. It is used to implement the above method. The system includes: a file acquisition module, which is used to acquire bidding documents; a file disassembly module, which is used to perform disassembly processing based on the bidding documents to obtain multiple bidding data pairs; a prompt instruction generation module, which is used to construct instructions based on the bidding data pairs to obtain prompt instructions for the bidding data pairs; a large model scoring module, which is used to process the prompt instructions of the bidding data pairs based on the large model after training and fine-tuning, obtain the scores of the bidding data pairs, and accumulate the scores to obtain the scores of the bidding documents.

[0049] The intelligent bid evaluation system provided in this application organically integrates the above-mentioned modules (file acquisition module, file disassembly module, prompt instruction generation module, and large model scoring module) to form an efficient, transparent, and fair intelligent bid evaluation system, and improves the adaptability and scalability of the system through continuous optimization.

[0050] An embodiment of the present application provides a computer storage medium, including computer instructions. When the computer instructions are executed on a computer device, the computer device executes the above-mentioned large model-based intelligent bid evaluation method.

[0051] An embodiment of the present application provides a computer program product. When the computer program product is executed on a computer device, the computer device executes the above-mentioned large model-based intelligent bid evaluation method.

[0052] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent bid evaluation method based on a large model, characterized in that: include: S1. Obtain bidding documents; S2. Disassemble and process the bidding documents to obtain multiple bidding data pairs; S3, constructing instructions for the bidding data pair based on the historical bidding database to obtain prompt instructions for the bidding data pair; S4. Based on the trained and fine-tuned large model, the prompt instructions of the bidding data pair are processed to obtain the score of the bidding data pair, and the scores are accumulated to obtain the score of the bidding document.

2. According to the large model-based intelligent bidding evaluation method of claim 1, it is characterized in that: S2. Decomposing the bidding documents to obtain multiple bidding data pairs; including: S21. Extract all bidding clauses and corresponding scoring criteria from the bidding documents based on the bidding document disassembly model; S22, searching the bidding documents for the bidding response content corresponding to the bidding terms based on the search enhancement method; S23. Based on the bidding terms and the corresponding bid response content, a bidding data pair is formed.

3. According to the large model-based intelligent bidding evaluation method of claim 2, it is characterized in that: Based on the search enhancement method, the bidding documents are searched for the bidding response content corresponding to the bidding terms; including: Divide the bidding documents into multiple small sections according to semantics; Convert small text segments into vector form and store them in a database that supports vector retrieval and full-text keyword retrieval; Obtain the bidding terms to be queried, and query candidate bidding response contents from the database based on vector retrieval and full-text keyword retrieval; Based on the text relevance model, the bid response content with the highest relevance to the tender terms is screened out from the candidate bid response contents as the bid response content corresponding to the tender terms.

4. According to the large model-based intelligent bidding evaluation method of claim 1, it is characterized in that: S3, constructing instructions for the bidding data pair based on the historical bidding database to obtain prompt instructions for the bidding data pair; including: Based on the bidding data pair, querying historical bidding data pairs whose similarity is greater than a threshold from a historical bidding database as candidate data pairs; Input the candidate data pair into the text similarity evaluation model, and select the candidate data pair with the greatest similarity to the bidding data pair as an example; According to the bidding data pair, the scoring criteria of the bidding data pair, the examples, and the scoring criteria of the examples, prompt instructions including role definition, task description, specific requirements and example content are generated.

5. According to the large model-based intelligent bidding evaluation method of claim 1, it is characterized in that: S4. Processing the prompt instructions of the bidding data pair based on the trained and fine-tuned large model to obtain the score of the bidding data pair, and accumulating the scores to obtain the score of the bidding document; including: Inputting the prompt instructions of the bidding data pair into the trained and fine-tuned big model to obtain a big model scoring description text; Input the big model scoring description texts of all bidding data pairs into the trained and fine-tuned big model to obtain the scores of all bidding data pairs; The scores of all bidding data pairs are added up to obtain the score of the bidding document.

6. According to the large model-based intelligent bidding evaluation method of claim 1, it is characterized in that: The historical bidding database is obtained through the following steps: Obtain historical bidding documents; Based on the historical bidding documents, the bidding terms, bid response content and corresponding scores are obtained; Based on the bidding terms and the corresponding bid response content, construct a historical bidding data pair; The historical bidding data pairs and the corresponding scores are stored in a database supporting vector retrieval and full-text keyword retrieval in a standardized format to form a historical bidding database.

7. The intelligent bid evaluation method based on a large model according to claim 1 is characterized in that: The large model after training and fine-tuning is obtained through the following steps: Select the basic large model; Based on the historical bidding database as the data set for the big model thinking chain reasoning, the LoRA fine-tuning technology is used to fine-tune the basic big model to obtain the big model after training and fine-tuning.

8. An intelligent bidding evaluation system based on a large model, characterized in that: Used to implement an intelligent bid evaluation method based on a large model as described in any one of claims 1 to 7; comprising: Document acquisition module, used to obtain bidding documents; A file disassembly module, used to perform disassembly processing based on the bidding document to obtain multiple bidding data pairs; A prompt instruction generating module, used for constructing instructions for the bidding data pair based on the historical bidding database to obtain prompt instructions for the bidding data pair; The large model scoring module is used to process the prompt instructions of the bidding data pair based on the trained and fine-tuned large model, obtain the score of the bidding data pair, and accumulate the scores to obtain the score of the bidding document.

9. A computer storage medium, characterized in that It includes computer instructions, which, when executed on a computer device, enable the computer device to execute an intelligent bid evaluation method based on a large model as described in any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product runs on a computer device, the computer device executes an intelligent bid evaluation method based on a large model as described in any one of claims 1 to 7.

Images