Intelligent bid evaluation method and system based on large model

Abstract

The invention discloses an intelligent bid evaluation method and system based on a large model. The method comprises the following steps: carrying out format preprocessing on a bidding and tendering file; performing structure analysis processing on the bidding and tendering file; constructing a bid evaluation rule model in real time based on a large language model; constructing a bid evaluation knowledge base based on the analyzed bid invitation file and historical bid evaluation data; a mixed retrieval method is adopted to retrieve associated knowledge from the bid evaluation knowledge base; based on the large language model, the associated knowledge and the bid evaluation rule model, performing constraint reasoning on the bid file, and generating a bid evaluation result including bid sorting on the bid file; and explaining the bid evaluation process and the bid evaluation result, and outputting structured bid evaluation result information. According to the invention, by introducing a cooperation mechanism of the large language model and the bid evaluation knowledge base, the conversion of the bid evaluation process from artificial subjective review to intelligent evaluation and sorting with semantic driving, rule constraint and explainable results is realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to an intelligent evaluation method and system based on a large model. Background Technology

[0002] In engineering construction and procurement activities, the tendering party typically needs to review, evaluate, and rank the bids submitted by multiple bidders within a limited timeframe. Bids are usually complex, covering multiple dimensions such as commercial terms, technical solutions, qualification certificates, service capabilities, and price responsiveness, and the evaluation criteria and weightings vary significantly between different projects.

[0003] The current evaluation of bid documents mainly relies on manual methods, which generally suffers from the following technical problems: Low review efficiency: The tender documents are long and contain many cross-references. Manual review and scoring of each item is time-consuming and cannot meet the needs of centralized bid evaluation. Insufficient consistency in review: Different reviewers have subjective differences in their understanding of the terms of the tender documents, resulting in poor stability of the review results; Rules and experience are difficult to reuse: Bidding rules, expert experience, and historical bidding cases are scattered in different documents or personnel experiences, making it difficult to systematically accumulate and reuse them; Semantic level assessment difficulties: Bids often respond to tender requirements in natural language, making it difficult for traditional keyword-based or fixed rule-based systems to accurately determine the degree of substantive response. Insufficient interpretability of bid evaluation results: The manual bid evaluation process makes it difficult to fully record the reasoning basis, which is not conducive to result verification and post-event traceability. Summary of the Invention

[0004] In view of the above-mentioned shortcomings of the existing technology, the present invention provides an intelligent bidding evaluation method and system based on a large model to solve the problems of low review efficiency, insufficient consistency and insufficient interpretability of bidding evaluation results in the existing manual bidding evaluation technology.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following solution: A smart evaluation method based on a large model includes the following steps: Receive tender documents and multiple bid documents, and perform format preprocessing on the tender documents and bid documents to generate uniform text data; Based on the preprocessed text data of the bidding documents and tender documents, a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents is constructed. Based on a large language model, multiple evaluation elements from different chapters of the bidding documents are extracted in parallel, and an evaluation rule model is constructed in real time. Based on the parsed tender documents and historical bid evaluation data, a bid evaluation knowledge base with multi-dimensional blocks is constructed to form a bid evaluation knowledge graph; Based on a large language model, a semantic summary is generated from the structured document of the bid document, and a hybrid retrieval method is used to retrieve related knowledge from the bid evaluation knowledge base; Based on the large language model, the associated knowledge retrieved from the bid evaluation knowledge base, and the bid evaluation rule model, the bid documents are subjected to constraint reasoning on compliance, completeness, and substantive responsiveness, and bid evaluation results including bid ranking are generated. The evaluation process and results are interpreted and structured evaluation result information is output.

[0006] Preferably, the method for preprocessing the tender documents and the bid documents includes: text extraction, encoding standardization, and format standardization.

[0007] Preferably, the bidding documents and tender documents undergo structural parsing processing to construct a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents, including: Based on the format type of the tender documents and bid documents, a corresponding differentiated parsing strategy is adopted for processing to construct a structured document model that represents the hierarchical relationship between the chapters of the tender documents and bid documents; When the tender document or bid document is in Word format, the table of contents is parsed based on the outline level and heading style built into the tender document or bid document. When the tender document or bid document is in PDF format, the asynchronous document parsing service is invoked to parse the content of the tender document or bid document into hierarchical markup language content. When the tender document or bid document is in TXT format, the content of the tender document or bid document is directly read and divided into paragraph levels.

[0008] Preferably, based on a large language model, multiple evaluation elements from different chapters of the bidding documents are extracted in parallel, and an evaluation rule model is constructed in real time, including: By invoking multiple independent large language model inference tasks in parallel, the evaluation elements of different chapters in the bidding documents are extracted independently, and intermediate results are returned in a streaming manner for real-time construction of the evaluation rule model.

[0009] Preferably, in the construction of a bid evaluation knowledge base based on the parsed tender documents and historical bid evaluation data, the multi-dimensional blocks include: The evaluation clauses are divided into blocks based on the evaluation dimensions and terms as the smallest units; Directory structure blocks for quick location; Adaptive length segmentation based on a dynamic adjustment of the maximum content length threshold.

[0010] Preferably, based on the semantic summary of the tender documents, a hybrid retrieval method is used to retrieve related knowledge from the bid evaluation knowledge base. The hybrid retrieval method is a hybrid retrieval method that combines vector retrieval and knowledge graph reasoning.

[0011] Preferably, the hybrid retrieval method includes: The processed structured document model is converted into a vector representation; A preliminary search of the bidding evaluation knowledge base is performed based on vector similarity to obtain candidate clauses and historical bidding evaluation cases. Using the candidate terms retrieved in the initial search as seed nodes, a multi-hop subgraph is constructed in the bid evaluation knowledge graph; A personalized webpage ranking algorithm is executed on a multi-hop subgraph to rank candidate clauses and historical bidding cases by importance; The vector similarity retrieval results and the importance ranking results obtained from the bidding evaluation knowledge graph are weighted and fused to output a set of related candidate clauses and historical bidding evaluation cases as related knowledge retrieved from the bidding evaluation knowledge base.

[0012] Preferably, the content stored in the bid evaluation knowledge base includes: bidding specifications, industry bid evaluation rules, historical bid evaluation cases, expert review experience, and typical problem patterns.

[0013] Preferably, the evaluation process and results are interpreted and structured evaluation result information is output, including: The evaluation results are broken down into evaluation dimensions to generate a scoring composition explanation for each bid document; The evaluation reasoning path is generated based on a large language model to explain the basis for judging each bid document under each evaluation clause. The evaluation results are linked and labeled with industry evaluation rules, historical evaluation cases, or expert review experience in the evaluation knowledge base; Identify uncertainties, potential risks, or points of contention during the bid evaluation process and generate corresponding alerts. Output a bid evaluation report in structured data format, including the bid evaluation conclusions, bid ranking results, and corresponding explanations.

[0014] A system for implementing the above-mentioned intelligent evaluation method based on a large model includes: Document preprocessing module: used to receive tender documents and multiple bid documents, and to perform format preprocessing on the tender documents and bid documents to generate uniform text data; Structure parsing module: Based on the preprocessed text data of the bidding documents and tender documents, structure parsing is performed to construct a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents; The bid evaluation rule model component module is used to extract multiple bid evaluation elements from different chapters of the bidding documents in parallel based on a large language model, and to build the bid evaluation rule model in real time. Bid Evaluation Knowledge Base Component Module: Used to build a bid evaluation knowledge base with multi-dimensional blocks based on the parsed bidding documents and historical bid evaluation data, forming a bid evaluation knowledge graph; Hybrid retrieval module: Based on a large language model, a semantic summary is generated from the structured document of the tender documents, and a hybrid retrieval method is used to retrieve related knowledge from the bid evaluation knowledge base; The reasoning and evaluation module is used to perform constraint reasoning on the compliance, completeness, and substantive responsiveness of the bid documents based on the large language model, the related knowledge retrieved from the evaluation knowledge base, and the evaluation rule model, and to generate evaluation results that include a ranking of the bid documents. The interpretation output module is used to interpret the evaluation process and results, and output structured evaluation result information.

[0015] Compared with existing technologies, this invention first receives tender documents and multiple bid documents, performs format preprocessing on the tender documents and bid documents to generate unified text data; then, it performs structural parsing on the tender documents and bid documents to construct a structured document model representing the hierarchical relationship between the chapters of the tender documents and bid documents; next, based on a large language model, it extracts multiple evaluation elements from different chapters of the tender documents in parallel and constructs an evaluation rule model in real time; then, based on the parsed tender documents and historical evaluation data, it constructs an evaluation knowledge base containing multi-dimensional blocks; next, based on the semantic summary of the bid documents, it uses a hybrid retrieval method to retrieve related knowledge from the evaluation knowledge base; then, based on the large language model, the related knowledge retrieved from the evaluation knowledge base, and the evaluation rule model, it performs constraint reasoning on the compliance, completeness, and substantive responsiveness of the bid documents, and generates an evaluation result that includes a bid ranking of the bid documents; finally, it interprets the evaluation process and evaluation results and outputs structured evaluation result information. Therefore, this invention, by introducing a collaborative mechanism between a large language model and a bid evaluation knowledge base, transforms the bid evaluation process from subjective manual review to intelligent evaluation and ranking driven by semantics, constrained by rules, and with interpretable results. This significantly improves the efficiency of bid document review and ranking; reduces subjective differences in manual bid evaluation and enhances result consistency; achieves the systematic accumulation and reuse of bid evaluation rules and expert experience; supports interpretable output of bid evaluation results for easy verification and traceability; and can flexibly adapt to bidding scenarios with different industries and bid evaluation rules. Attached Figure Description

[0016] Appendix Figure 1 This is a flowchart of the intelligent bidding evaluation method based on a large model according to the present invention; Appendix Figure 2 This is a flowchart of the hybrid retrieval method in the intelligent evaluation method based on a large model of the present invention; Appendix Figure 3 This is a system block diagram of the intelligent bidding evaluation system based on a large model according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] Furthermore, in describing representative embodiments, the specification may have presented methods and / or processes as a specific sequence of steps. However, the method or process should not be limited to a specific order of steps to the extent that it does not depend on that specific order. As will be understood by those skilled in the art, other orders of steps are also possible. Therefore, the specific order of steps set forth in the specification should not be construed as a limitation on the embodiments. Moreover, the method and / or process should not be limited to the steps performed in the order written; those skilled in the art will readily understand that these orders can be varied and still remain within the spirit and scope of the embodiments of this application.

[0020] First, this specific embodiment provides an intelligent evaluation method based on a large model, as shown in the attached figure. Figure 1 As shown, it includes the following steps: Step S1) Receive the tender documents and multiple bid documents, and perform format preprocessing on the tender documents and bid documents to generate uniform text data.

[0021] In this step, the bidding party can upload the bidding documents and multiple bids through the system. Both the bidding documents and bids can be in Word, PDF, or TXT format. After receiving the bidding documents and bids, the system will create an independent analysis and evaluation task instance for each batch of bid evaluation tasks.

[0022] Meanwhile, the system employs differentiated preprocessing strategies for different file formats, including text extraction, encoding standardization, and layout standardization. Text extraction includes direct parsing of editable text and character recognition conversion of non-editable text. Encoding standardization involves automatically identifying and converting text with different character encoding formats into a unified encoding format. Layout standardization involves cleaning up layout elements in the text that are irrelevant to the semantic content and standardizing the text structure.

[0023] This preprocessing step enables the automated extraction and structured representation of the content from the tender documents and multiple bid documents, avoiding human interpretation bias and providing a unified standard for subsequent evaluation.

[0024] Step S2) Based on the preprocessed text data of the bidding documents and tender documents, perform structural parsing processing to construct a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents.

[0025] Specifically, based on the format type of the bidding documents and tender documents, corresponding differentiated parsing strategies are adopted to process them in order to construct a structured document model that represents the hierarchical relationship between the chapters of the bidding documents and tender documents; When the tender documents or bid documents are in Word format, the table of contents is parsed based on the outline level and heading style built into the tender documents or bid documents. When the bidding documents or tender documents are in PDF format, the system calls the asynchronous document parsing service through task submission. Based on page layout analysis and text structure analysis, the system parses the document content, reconstructs the logical structure hierarchy of the document, and converts the parsing results into markup language content with hierarchical tags. When the tender documents or bid documents are in TXT format, the system performs structural analysis on the text content based on preset paragraph division and hierarchy recognition rules in order to build a structured document model.

[0026] Using the above methods, the system will construct a structured document model of the hierarchical relationship between the chapters of the bidding documents and tender documents, providing a structural foundation for subsequent chapter-level evaluation and alignment.

[0027] This step of structural analysis of the bidding documents and tender documents can eliminate the impact of differences in tender document format and structure, and achieve comparability modeling of multiple tender documents.

[0028] Step S3) Based on the large language model, extract multiple evaluation elements from different chapters of the bidding documents in parallel and construct the evaluation rule model in real time. The large language model refers to a general natural language understanding and reasoning model trained on a deep neural network.

[0029] Specifically, based on a large language model, the system employs a parallel processing mechanism to extract evaluation elements from the bidding documents, including but not limited to evaluation criteria, scoring dimensions, scoring rules, weighting information, and substantive clause requirements.

[0030] In this step, the system assigns independent reasoning tasks to each chapter and uses multi-threading to concurrently call reasoning instances of the same large language model. Each reasoning task is managed and scheduled by a unified task scheduling module to achieve parallel extraction of evaluation elements from different chapters.

[0031] This step uses a large language model to break through the limitations of traditional rule or keyword matching, and achieves a deep semantic understanding of the content of the bid response.

[0032] Step S4) Based on the parsed tender documents and historical bid evaluation data, construct a bid evaluation knowledge base containing multi-dimensional blocks to form a bid evaluation knowledge graph.

[0033] The bid evaluation knowledge base is a structured or semi-structured collection of knowledge used to store bid evaluation rules, industry standards, historical bid evaluation cases, and expert experience. The content stored in the bid evaluation knowledge base includes: bidding specifications, industry bid evaluation rules, historical bid evaluation cases (historical bid evaluation data comes from historical bidding projects that have completed the bid evaluation process), expert review experience, and typical problem patterns.

[0034] Specifically, multi-dimensional segmentation includes: Clause Chunk for Bid Evaluation: Knowledge blocks are constructed using evaluation dimensions and clauses as the smallest unit. Each knowledge block contains clause paths, clause content, and constraints.

[0035] Catalog Chunk Structure: Extract the complete bid evaluation directory structure from the bidding documents and construct a directory index for quick location of bid evaluation elements.

[0036] Length-based chunking: Based on the context length limit of the large language model inference input, a maximum content length threshold is set, and knowledge content is dynamically split and merged based on this threshold: when the content length exceeds the threshold, it is split into multiple independent knowledge blocks; when the content length of adjacent knowledge blocks is lower than the threshold and they are semantically continuous, they are merged to ensure the stability of the knowledge block content size and semantic integrity.

[0037] A multi-dimensional, segmented knowledge base for bid evaluation can avoid the instability caused by relying solely on model reasoning, and improve the standardization and credibility of bid evaluation judgments.

[0038] Step S5) Based on a large language model, a semantic summary is generated from the structured document of the tender document. A hybrid retrieval method is then used to retrieve relevant knowledge from the bid evaluation knowledge base. The relevant knowledge includes bid evaluation clauses and historical bid evaluation cases. The hybrid retrieval method combines vector retrieval with knowledge graph reasoning.

[0039] For details, see attached. Figure 2 As shown, the hybrid retrieval method includes: Step S51) Using a semantic embedding model in the field of natural language processing, the tender response text of the processed structured document model is mapped into a dense vector representation; the embedding model is a pre-trained deep neural network model used to capture the semantic features of the text and support semantic similarity calculation. Step S52) Perform a preliminary search in the bid evaluation knowledge base based on vector similarity to obtain candidate clauses and historical bid evaluation cases; Step S53) Using the candidate terms retrieved initially as seed nodes, construct a multi-hop subgraph in the bidding evaluation knowledge graph; wherein, the bidding evaluation knowledge graph is used to describe the structured relationships between bidding evaluation-related knowledge, and its schema includes entity types and relation types: Entity types include bid evaluation clause entities, bid evaluation dimension entities, historical bid evaluation case entities, bidding specification entities, and expert review experience entities; The types of relationships include clause subordination, rule constraint, case association, normative reference, and similarity association.

[0040] The system takes the candidate bid evaluation clause entity as the starting node, and expands the associated entity nodes in the bid evaluation knowledge graph along the relationship type according to the preset hop number threshold. It constructs a multi-hop subgraph containing candidate clauses, associated bid evaluation cases and related bid evaluation rules to depict the relationship structure between bid evaluation knowledge.

[0041] Step S54) Execute a personalized page ranking algorithm on the multi-hop subgraph to sort the candidate terms and historical bidding cases by importance; specifically, on the constructed multi-hop subgraph, execute the personalized PageRank algorithm, take the node corresponding to the candidate terms as the personalized seed node set, calculate the personalized access probability of all nodes in the subgraph relative to the seed set, and sort the candidate terms and related cases by importance based on the probability value. Step S55) The vector similarity retrieval results and the importance ranking results obtained by the bidding knowledge graph reasoning are weighted and fused to output the set of related candidate clauses and historical bidding cases as related knowledge retrieved from the bidding knowledge base.

[0042] This step enables a multi-dimensional and quantitative comprehensive evaluation of the bid documents, providing an objective basis for ranking.

[0043] Step S6) Based on the large language model, the associated knowledge retrieved from the bid evaluation knowledge base, and the bid evaluation rule model, perform constraint reasoning on the compliance, completeness, and substantive responsiveness of the bid documents, and generate bid evaluation results that include a ranking of the bid documents. Substantive responsiveness refers to the degree to which the bid documents truly satisfy the key terms of the tender documents.

[0044] Specifically, based on a large language model, the content of the tender documents is semantically understood, and combined with the set of candidate clauses obtained in step S5 as evaluation clauses, the compliance, completeness and substantive responsiveness of the tender response are inferred and judged.

[0045] During the reasoning process, a model constraint mechanism based on evaluation rules is introduced to verify the model's reasoning results and reduce the uncertainty brought about by free reasoning.

[0046] This step automates and standardizes the bid document sorting process, reducing subjective differences caused by manual sorting.

[0047] Step S7) The evaluation process and results are interpreted and structured evaluation result information is output.

[0048] Specifically, the structured evaluation results output includes: The evaluation results are broken down into evaluation dimensions to generate a scoring composition explanation for each bid document; The evaluation reasoning path is generated based on a large language model to explain the basis for judging each bid document under each evaluation clause. The evaluation results are linked and labeled with industry evaluation rules, historical evaluation cases, or expert review experience in the evaluation knowledge base; Identify uncertainties, potential risks, or points of contention during the bid evaluation process and generate corresponding alerts. Output a bid evaluation report in structured data format, including the bid evaluation conclusions, bid ranking results, and corresponding explanations.

[0049] The data format includes at least a key-value pair structure to support system storage, interface calls, and subsequent auditing and tracing.

[0050] Simultaneously, key intermediate results generated during the bid evaluation process (i.e., steps S1 to S7 of this invention are the entire bid evaluation process) are recorded, enabling traceability and verifiability of the bid evaluation process. This achieves interpretable output of bid evaluation results and process traceability, improving the transparency and credibility of the bid evaluation.

[0051] Key intermediate results include: Text data and structured document models generated during the document preprocessing and structure parsing stages; Evaluation elements extracted from the tender documents and the constructed evaluation rule model; Semantic summary and vector representation results of the tender documents; Candidate bid evaluation clauses, historical bid evaluation cases, and their relationships are retrieved from the bid evaluation knowledge base; Intermediate sorting results obtained based on knowledge graph reasoning and sorting algorithms; The reasoning and judgment results under each evaluation dimension and evaluation clause, and their corresponding basis.

[0052] In addition, this specific embodiment also provides a system for implementing the above-mentioned intelligent evaluation method based on a large model, as shown in the attached figure. Figure 3 As shown, it includes: Document preprocessing module: Used to receive tender documents and multiple bid documents, and to preprocess the tender documents and bid documents to generate uniform text data; Structure parsing module: used to perform structure parsing processing on the preprocessed text data of the bidding documents and tender documents, in order to construct a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents; The bid evaluation rule model component module is used to extract multiple bid evaluation elements from different chapters of the bidding documents in parallel based on a large language model, and to build the bid evaluation rule model in real time. Bid Evaluation Knowledge Base Component Module: Used to build a bid evaluation knowledge base with multi-dimensional blocks based on the parsed bidding documents and historical bid evaluation data, forming a bid evaluation knowledge graph; Hybrid retrieval module: Based on a large language model, a semantic summary is generated from the structured document of the tender documents, and a hybrid retrieval method is used to retrieve related knowledge from the bid evaluation knowledge base; The reasoning and evaluation module is used to perform constraint reasoning on the compliance, completeness, and substantive responsiveness of bid documents based on a large language model, related knowledge retrieved from the bid evaluation knowledge base, and bid evaluation rule model, and to generate bid evaluation results that include a ranking of bid documents. The interpretation output module is used to interpret the evaluation process and results, and output structured evaluation result information.

[0053] In summary, the method of the present invention is geared towards the application scenario of the bidding party. Based on the bidding documents and bidding requirements, it automatically reviews and comprehensively evaluates multiple bid documents and generates a bid ranking result.

[0054] This method analyzes the bidding documents and bidding requirements to extract evaluation criteria, substantive clauses, scoring dimensions, and corresponding weights. It then constructs a bidding evaluation knowledge base related to the bidding field, storing bidding specifications, industry evaluation rules, historical evaluation cases, expert review experience, and typical problem patterns. Based on a large language model, it performs semantic understanding of the bid documents, identifying their responses to the various requirements of the bidding documents. Combining this with the evaluation knowledge base, it reviews and judges the bid documents in terms of compliance, completeness, and substantive responsiveness. Following the evaluation dimensions and weights defined in the bidding documents, it comprehensively evaluates the bid documents, generating structured evaluation results. Based on the comprehensive evaluation results of each bid document, it automatically generates a bid ranking to assist the bidding party in making evaluation decisions. Finally, it outputs explanations and related knowledge entries related to the evaluation conclusions and ranking results, making the evaluation process interpretable. Therefore, this invention can automatically complete the review, comprehensive evaluation, and ranking of bid documents; significantly shorten the evaluation cycle while ensuring evaluation quality; improve the standardization, consistency, and transparency of the evaluation process; and provide the bidding party with reliable intelligent evaluation support decision-making capabilities.

[0055] Compared to methods that rely solely on manual or fixed-rule evaluation, this invention achieves intelligent, standardized, and interpretable evaluation processes through the collaboration of a large language model and an evaluation knowledge base. This effectively improves evaluation efficiency and quality, and significantly enhances the reusability and credibility of the evaluation results.

[0056] In addition, the present invention also provides a computer storage medium storing executable instructions, which, when executed by a processor, execute the aforementioned intelligent evaluation method based on a large model.

[0057] In addition, the present invention also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement some or all of the steps in the above-described method. This computer program product can be implemented specifically through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied as a computer storage medium; in another optional embodiment, the computer program product is specifically embodied as a software product, such as a Software Development Kit (SDK), etc.

[0058] Furthermore, in this embodiment, the functional modules can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional module.

[0059] If the integrated unit is implemented as a software functional module and is not 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 this embodiment, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the method of this embodiment. 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.

[0060] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0061] This invention is described with reference to schematic and / or block diagrams illustrating the implementation of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the schematic and / or block diagrams can be implemented by computer program instructions, and combinations of blocks in the schematic and / or block diagrams can be implemented. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the schematic and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 Devices that specify the functions in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction device, which is implemented in the implementation flow diagram. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment, causing a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A smart evaluation method based on a large model, characterized in that, Includes the following steps: Receive tender documents and multiple bid documents, and perform format preprocessing on the tender documents and bid documents to generate uniform text data; Based on the preprocessed text data of the bidding documents and tender documents, a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents is constructed. Based on a large language model, multiple evaluation elements from different chapters of the bidding documents are extracted in parallel, and an evaluation rule model is constructed in real time. Based on the parsed tender documents and historical bid evaluation data, a bid evaluation knowledge base with multi-dimensional blocks is constructed to form a bid evaluation knowledge graph; Based on a large language model, a semantic summary is generated from the structured document of the bid document, and a hybrid retrieval method is used to retrieve related knowledge from the bid evaluation knowledge base; Based on the large language model, the associated knowledge retrieved from the bid evaluation knowledge base, and the bid evaluation rule model, the bid documents are subjected to constraint reasoning on compliance, completeness, and substantive responsiveness, and bid evaluation results including bid ranking are generated. The evaluation process and results are interpreted and structured evaluation result information is output.

2. The intelligent evaluation method based on a large model according to claim 1, characterized in that, The methods for preprocessing the tender documents and the bid documents include: text extraction, encoding standardization, and format standardization.

3. The intelligent evaluation method based on a large model according to claim 1, characterized in that, The bidding documents and tender documents undergo structural parsing processing to construct a structured document model representing the hierarchical relationship between the chapters of the bidding documents and tender documents, including: Based on the format type of the tender documents and bid documents, a corresponding differentiated parsing strategy is adopted for processing to construct a structured document model that represents the hierarchical relationship between the chapters of the tender documents and bid documents; When the tender document or bid document is in Word format, the table of contents is parsed based on the outline level and heading style built into the tender document or bid document. When the tender document or bid document is in PDF format, the asynchronous document parsing service is invoked to parse the content of the tender document or bid document into hierarchical markup language content. When the tender document or bid document is in TXT format, the content of the tender document or bid document is directly read and divided into paragraph levels.

4. The intelligent evaluation method based on a large model according to claim 1, characterized in that, Based on a large language model, multiple evaluation elements from different chapters of the tender document are extracted in parallel, and an evaluation rule model is constructed in real time, including: By invoking multiple independent large language model inference tasks in parallel, the evaluation elements of different chapters in the bidding documents are extracted independently, and intermediate results are returned in a streaming manner for real-time construction of the evaluation rule model.

5. The intelligent evaluation method based on a large model according to claim 1, characterized in that, Based on the parsed tender documents and historical bid evaluation data, a bid evaluation knowledge base is constructed that includes multi-dimensional blocks, namely: The evaluation clauses are divided into blocks based on the evaluation dimensions and terms as the smallest units; Directory structure blocks for quick location; Adaptive length segmentation based on a dynamic adjustment of the maximum content length threshold.

6. The intelligent evaluation method based on a large model according to claim 1, characterized in that, Based on the semantic summary of the tender documents, a hybrid retrieval method is used to retrieve related knowledge from the bid evaluation knowledge base. The hybrid retrieval method is a combination of vector retrieval and knowledge graph reasoning.

7. The intelligent evaluation method based on a large model according to claim 6, characterized in that, The hybrid retrieval method includes: The processed structured document model is converted into a vector representation; A preliminary search of the bidding evaluation knowledge base is performed based on vector similarity to obtain candidate clauses and historical bidding evaluation cases. Using the candidate terms retrieved in the initial search as seed nodes, a multi-hop subgraph is constructed in the bid evaluation knowledge graph; A personalized webpage ranking algorithm is executed on a multi-hop subgraph to rank candidate clauses and historical bidding cases by importance; The vector similarity retrieval results and the importance ranking results obtained from the bidding evaluation knowledge graph are weighted and fused to output the set of related candidate clauses and historical bidding evaluation cases as related knowledge retrieved from the bidding evaluation knowledge base.

8. The intelligent evaluation method based on a large model according to claim 1, characterized in that, The content stored in the bid evaluation knowledge base includes: bidding specifications, industry bid evaluation rules, historical bid evaluation cases, expert review experience, and typical problem patterns.

9. The intelligent evaluation method based on a large model according to claim 1, characterized in that, The evaluation process and results are interpreted and structured evaluation result information is output, including: The evaluation results are broken down into evaluation dimensions to generate a scoring composition explanation for each bid document; The evaluation reasoning path is generated based on a large language model to explain the basis for judging each bid document under each evaluation clause. The evaluation results are linked and labeled with industry evaluation rules, historical evaluation cases, or expert review experience in the evaluation knowledge base; Identify uncertainties, potential risks, or points of contention during the bid evaluation process and generate corresponding alerts. Output a bid evaluation report in structured data format, including the bid evaluation conclusions, bid ranking results, and corresponding explanations.

10. A system for implementing the intelligent evaluation method based on a large model as described in claim 1, characterized in that, include: Document preprocessing module: used to receive tender documents and multiple bid documents, and to perform format preprocessing on the tender documents and bid documents to generate uniform text data; Structure parsing module: used to perform structure parsing processing on the preprocessed text data of the bidding documents and tender documents, so as to construct a structured document model representing the hierarchical relationship of the chapters in the bidding documents and tender documents; Evaluation rule model component module: used to extract multiple evaluation elements from different chapters of the bidding documents in parallel based on a large language model, and to build the evaluation rule model in real time; Bid Evaluation Knowledge Base Component Module: Used to build a bid evaluation knowledge base with multi-dimensional blocks based on the parsed bidding documents and historical bid evaluation data, forming a bid evaluation knowledge graph; Hybrid retrieval module: used to generate semantic summaries of the structured documents of the tender documents based on a large language model, and to retrieve related knowledge from the bid evaluation knowledge base using a hybrid retrieval method; The reasoning and evaluation module is used to perform constraint reasoning on the compliance, completeness, and substantive responsiveness of the bid documents based on the large language model, the related knowledge retrieved from the evaluation knowledge base, and the evaluation rule model, and to generate evaluation results that include a ranking of the bid documents. The interpretation output module is used to interpret the evaluation process and results, and output structured evaluation result information.

Images