Intelligent bid invitation document review method, system and equipment and medium

Abstract

The invention relates to an intelligent bid invitation document review method, system and device and a medium, and belongs to the technical field of bid invitation and bidding management.The intelligent bid invitation document review method comprises the steps that structured analysis is conducted on an obtained to-be-reviewed bid invitation document based on a pre-trained language model, and a structured vector is obtained; constructing a vector library based on the obtained historical bid invitation files, performing similarity matching on the structured vector and the vector library, and determining a first historical bid invitation file corresponding to the to-be-examined bid invitation file in the historical bid invitation files based on a similarity matching result; obtaining an examination rule of the first historical bid invitation file, constructing a rule base, and performing compliance verification on the to-be-examined bid invitation file based on the rule base to obtain a rule verification result; and performing risk analysis on the structured vector based on the rule verification result and the constructed tendering and bidding knowledge graph to obtain a risk analysis result, and generating an examination report based on the risk analysis result and the rule verification result, thereby improving the examination efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bidding management, and in particular to a bidding document intelligent review method, system, device and medium. BACKGROUND

[0002] There are many problems in the bidding document review process: first, most of the reviewers are part-time, with high job mobility and insufficient experience, making it difficult to meet the new standards and requirements of procurement specifications; second, procurement specifications are complicated, and review requires checking a large number of regulatory documents and enterprise policies, which takes a long time to manually review the data and is prone to lax control and omission of problems; third, review and determination rely on human experience, lack objective basis, risk control ability is weak, and the multi-department review process is long, affecting procurement efficiency.

[0003] The traditional manual review method is inefficient and has a high error rate, and cannot adapt to large-scale and high-frequency bidding requirements; some simple review tools can only check based on fixed keywords and cannot achieve multi-dimensional compliance judgment and dynamic risk identification, therefore, an intelligent review method that can automatically analyze bidding documents, dynamically adapt to regulatory standards and comprehensively identify risks is needed to improve review efficiency and accuracy and reduce bidding compliance risks. SUMMARY

[0004] Therefore, it is necessary to provide a bidding document intelligent review method, system, device and medium to solve the technical problems of low review efficiency and low accuracy of bidding documents.

[0005] To solve the above problems, in a first aspect, the present application provides a bidding document intelligent review method, comprising: structurally analyzing a to-be-reviewed bidding document obtained based on a pre-trained language model to obtain a structured vector; constructing a vector library based on historical bidding documents obtained, performing similarity matching between the structured vector and the vector library, and determining a first historical bidding document corresponding to the to-be-reviewed bidding document in the historical bidding documents based on the similarity matching result; obtaining review rules of the first historical bidding document, constructing a rule library based on the review rules, regulatory rules and enterprise rules, and performing compliance verification on the to-be-reviewed bidding document based on the rule library to obtain a rule verification result; performing risk analysis on the structured vector based on the rule verification result and a constructed bidding and tender knowledge graph to obtain a risk analysis result, and generating a review report based on the risk analysis result and the rule verification result.

[0006] In a possible implementation manner, the structured vector is obtained by structurally analyzing the to-be-reviewed bidding document based on the pre-trained language model, comprising: The tender documents to be reviewed are broken down according to the legal structure of the tender documents to obtain module texts; The module text is segmented into sentences to obtain a text sequence; Based on a pre-trained language model, entity recognition, relation extraction, and text classification are performed on the text sequence to obtain structured information; The structured information is normalized to obtain a structured vector.

[0007] In one possible implementation, the structured information includes at least the basic bidding information, core project information, participant information, process rule information, and compliance information.

[0008] In one possible implementation, the step of performing similarity matching between the structured vector and the vector library, and determining the first historical bidding document corresponding to the bidding document to be reviewed based on the similarity matching result, includes: The similarity value between the structured vector and the vector library is calculated using cosine similarity. The tender documents to be reviewed and historical tender documents are matched based on a preset similarity threshold and the similarity value to obtain similarity matching results; Based on the similarity matching results, the first historical bidding document corresponding to the bidding document to be reviewed is determined from the historical bidding documents.

[0009] In one possible implementation, the cosine similarity is: , in, For structured vectors, Vectors of historical tender documents in the vector library. For vector dimensions, For the structured vector, the first A vector, For the first in the vector library A vector.

[0010] In one possible implementation, the compliance verification of the tender document to be reviewed based on the rule base to obtain the rule verification result includes: Based on the rule base, the general compliance baseline of the tender documents to be reviewed is verified to obtain the general compliance verification result; Based on the rule base, the legally mandatory requirements of the tender documents to be reviewed are verified to obtain legal compliance verification results; Based on the rule base, the enterprise-specific standards of the tender documents to be reviewed are verified to obtain the enterprise-customized compliance verification results. Obtain a rule verification result based on the general compliance verification result, the legal compliance verification result, and the enterprise customized compliance verification result.

[0011] In a possible implementation, the risk analysis result is obtained by performing risk analysis on the structured vector based on the rule verification result and the constructed bidding knowledge graph. Determine a violation clause of the structured vector based on the rule verification result, and determine a core node of the violation clause based on the constructed bidding knowledge graph; Determine a violation type node based on the bidding knowledge graph, and perform similarity calculation on the core node of the violation clause and the violation type node to determine a violation type; Quantify a risk level of the violation clause based on historical case data to determine the risk level of the violation clause; Generate the risk analysis result based on the violation clause, the violation type, and the risk level.

[0012] In a second aspect, the present application further provides a bidding document intelligent review system, comprising: A structured analysis module configured to perform structured analysis on the obtained to-be-reviewed bidding document based on a pre-trained language model to obtain a structured vector; A similarity matching module configured to construct a vector library based on the obtained historical bidding documents, perform similarity matching on the structured vector and the vector library, and determine a first historical bidding document corresponding to the to-be-reviewed bidding document based on a similarity matching result; A rule verification module configured to obtain review rules of the first historical bidding document, construct a rule library based on the review rules, legal rules, and enterprise rules, perform compliance verification on the to-be-reviewed bidding document based on the rule library, and obtain a rule verification result; A review report generation module configured to perform risk analysis on the structured vector based on the rule verification result and the constructed bidding knowledge graph to obtain a risk analysis result, and generate a review report based on the risk analysis result and the rule verification result.

[0013] In a third aspect, the present application further provides an electronic device, comprising a processor and a memory. The memory stores a computer readable program that can be executed by the processor; The processor executes the computer readable program to implement the steps in the bidding document intelligent review method described above.

[0014] In a fourth aspect, the present application further provides a computer readable storage medium for storing computer readable programs or instructions, which can realize the steps of the intelligent review method of the bidding document according to any one of the above methods when executed by a processor.

[0015] The beneficial effects of the present application are: based on the pre-trained language model, the obtained to-be-reviewed bidding document is structurally parsed to obtain a structured vector; based on the obtained historical bidding documents, a vector library is constructed, the structured vector is matched with the vector library in similarity, and based on the similarity matching result, a first historical bidding document corresponding to the to-be-reviewed bidding document in the historical bidding documents is determined; the review rules of the first historical bidding document are obtained, a rule library is constructed based on the review rules, the regulations and the enterprise rules, the to-be-reviewed bidding document is checked for compliance based on the rule library, and a rule checking result is obtained; the review experience of the historical bidding documents, the laws and regulations and the internal policies of the enterprise are integrated through the rule library to ensure that there is no omission in the review, the structured vector is analyzed for risks based on the rule checking result and the constructed bidding and tender knowledge graph to obtain a risk analysis result, and a review report is generated based on the risk analysis result and the rule checking result; the key information of the bidding document is extracted through the pre-trained language model, the review efficiency and the error rate are improved, the most similar case to the to-be-reviewed document is found from the historical data through the similarity matching, the manual search cost is reduced, the historical review experience, the laws and regulations and the internal policies of the enterprise are integrated through the rule library to ensure that there is no omission in the review, the potential problems are quickly identified by the reviewer, the risk analysis is performed through the knowledge graph to ensure the accuracy of the analysis result, and the efficiency and the accuracy of the review are improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 An embodiment flow chart of the intelligent review method of the bidding document provided by the present application; Figure 2 An embodiment structure diagram of the intelligent review system of the bidding document provided by the present application; Figure 3 An embodiment structure diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0018] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0019] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0020] This invention discloses a method, system, device, and medium for intelligent review of tender documents, which can be used in a computer. The method, device, or computer-readable storage medium involved in this invention can be integrated with the aforementioned device or be relatively independent.

[0021] One specific embodiment of the present invention discloses an intelligent review method for tender documents, which can be executed by a computer, specifically by one or more processors of the computer. For example... Figure 1 As shown, the intelligent review method for tender documents includes: S101. Based on the pre-trained language model, the obtained tender documents to be reviewed are parsed in a structured manner to obtain structured vectors; It should be noted that by using a pre-trained language model to parse the tender documents to be reviewed and extract key information, manual word-by-word analysis is avoided, parsing time is shortened, and review efficiency is improved.

[0022] S102. Construct a vector library based on the obtained historical bidding documents, perform similarity matching between the structured vectors and the vector library, and determine the first historical bidding document corresponding to the bidding document to be reviewed based on the similarity matching results. It should be noted that similarity matching identifies historical tender documents corresponding to the tender documents, reducing manual retrieval costs and providing a basis for tender document review, thus helping reviewers quickly identify potential problems.

[0023] S103. Obtain the review rules of the first historical bidding documents, construct a rule base based on the review rules, regulatory rules, and enterprise rules, perform compliance verification on the bidding documents to be reviewed based on the rule base, and obtain the rule verification results.

[0024] It should be noted that the rule base integrates historical review experience, laws and regulations, and corporate internal policies to ensure that no review is missed and to reduce the burden of manual judgment.

[0025] S104, based on the rule checking result and the constructed bidding knowledge graph, performing risk analysis on the structured vector to obtain a risk analysis result, and generating an examination report based on the risk analysis result and the rule checking result; It should be noted that through automatic analysis, intelligent matching, rule checking and risk analysis, the efficiency of bidding document examination is improved, the accuracy is guaranteed, and the standardized management is realized.

[0026] In some embodiments, in step S101, the pre-trained language model is used to perform structured analysis on the obtained to-be-examined bidding document to obtain a structured vector, the to-be-examined bidding document is split according to the statutory structure of the bidding document to obtain module texts, the module texts are processed by sentence to obtain text sequences, the pre-trained language model is used to perform entity recognition, relationship extraction and text classification on the text sequences to obtain structured information, the structured information is normalized to obtain the structured vector. The statutory structure of the bidding document includes bidding announcement, bidder's guide, bid evaluation method, contract terms and formats, bill of quantities, drawings, technical standards and requirements, etc. The starting and ending positions of each module in the to-be-examined bidding document are located by regular matching or keyword retrieval, the to-be-examined bidding document is split into independent module texts according to the starting and ending positions of each module, each module text is processed by sentence using punctuation symbols combined with semantic judgment, and the processed texts are numbered in sequence to generate text sequences. The pre-trained language model can be BERT. The BERT is used to perform entity recognition, relationship extraction and text classification on the text sequences. A CRF (Conditional Random Field) layer is added to the output layer of the BERT to identify the entities of the text sequences. The classification layer of the BERT outputs the relationship types. A Softmax classifier is added to the output layer of the BERT to predict the text categories. The extracted entities, relationships and classification results are combined into structured data, i.e., structured information. The structured information at least includes bidding basic information, project core information, participant information, process rule information and compliance information. The structured data is normalized to obtain the structured vector.

[0027] In some embodiments, in step S102, based on the obtained historical bidding documents, a vector library is constructed, the structured vector is matched with the vector library, the first historical bidding document corresponding to the to-be-reviewed bidding document is determined based on the similarity matching result, the historical compliance bidding documents, bidding-related regulatory documents and review cases are parsed after being structured, and are split into document slices, vector data is generated through an Embedding service, and is stored in a vector database to construct the vector library, each document slice of which is associated with a document ID, a project type (engineering / material / service) and a review result label, a cosine similarity is used to calculate the similarity value of the structured vector and the vector library, the to-be-reviewed bidding document and the historical bidding document are matched based on a preset similarity threshold and the similarity value, a similarity matching result is obtained, the first historical bidding document corresponding to the to-be-reviewed bidding document in the historical bidding document is determined based on the similarity matching result, and the cosine similarity is: , wherein, is the structured vector, is the vector of the historical bidding document in the vector library, is the vector dimension, is the i-th vector in the structured vector, is the i-th vector in the vector library; Through similarity calculation, the similar case in the historical bidding document, i.e., the first historical bidding document, is found out, the first historical bidding document is taken as a review reference, and the similar case contains historical review problem points, rectification suggestions and corresponding regulatory basis.

[0028] ​​In some embodiments, in step S103, the review rules of the first historical bidding document are obtained, a rule library is constructed based on the review rules, the regulations rules and the enterprise rules, the compliance of the bidding document under review is verified based on the rule library to obtain a rule verification result, the general compliance bottom line of the bidding document under review is verified based on the rule library to obtain a general compliance verification result, the mandatory requirements of the bidding document under review are verified based on the rule library to obtain a legal compliance verification result, the enterprise individualized standards of the bidding document under review are verified based on the rule library to obtain an enterprise customized compliance verification result, the rule verification result is obtained based on the general compliance verification result, the legal compliance verification result and the enterprise customized compliance verification result, the structured information of the bidding document under review is compared with the review rules in the rule library through the general compliance verification to check whether the bidding document has a fundamental and principle violation, the structured information of the bidding document under review is compared with the regulations rules in the rule library through the legal compliance verification to check whether the bidding document complies with the mandatory provisions of laws and regulations, the structured information of the bidding document under review is compared with the enterprise rules in the rule library through the enterprise customized compliance verification to check whether the bidding document under review complies with the management specifications and business preferences of the bidding unit, the general compliance verification result, the legal compliance verification result and the enterprise customized compliance verification result are summarized, combined and prioritized to form a clear and usable rule verification result, the combined verification result is sorted according to the severity of the problem, the priority is that the problem of violating the mandatory requirements is high risk, the problem of violating the general compliance bottom line is medium risk, and the problem of not meeting the enterprise individualized standards is low risk, the general compliance bottom line is filtered, the laws and regulations are checked, and the enterprise standards are refined to ensure the compliance, legality and applicability of the bidding document, and the final output of the rule verification result provides clear modification guidance for the review personnel, greatly improving the efficiency and accuracy of the review work.

[0029] In some embodiments, in step S104, risk analysis is performed on the structured vector based on the rule verification result and the constructed bidding knowledge graph, a risk analysis result is obtained, a violation clause of the structured vector is determined based on the rule verification result, and a core node of the violation clause is determined based on the constructed bidding knowledge graph; a violation type node is determined based on the bidding knowledge graph, similarity calculation is performed between the core node of the violation clause and the violation type node to determine the violation type; risk level quantification is performed on the violation clause based on historical case data to determine the risk level of the violation clause, and the risk analysis result is generated based on the violation clause, the violation type, and the risk level of the violation clause in the structured vector; based on the rule verification result (violation list of general compliance, statutory compliance, and enterprise customized compliance), the risk location and the specific violation clause are located in combination with the structured vector, the complete text of the violation clause is extracted from the structured vector, the core entity and the attribute type (such as time attribute, qualification attribute, and amount attribute) and the associated entity of the core entity involved in the violation clause are quickly located through the knowledge graph according to the complete text of the violation clause, a structured list is generated, the structured list includes the violation clause, the core entity of the violation clause, the attribute type of the core entity, and the associated entity, a clustering algorithm of natural language processing is adopted, the scattered violation clauses are classified into a unified violation type based on the core entity of the violation clause, that is, the core node extracted from each violation clause is combined with the violation type node determined in the bidding knowledge graph, similarity calculation (based on cosine similarity) is performed between the core node of the violation clause and the violation type node by using a K-Means or hierarchical clustering algorithm, and the violation clauses with similar similarity are classified into the same type; violation case data of historical bidding documents is collected, a risk quantification model is constructed based on the historical case data, the risk level of each violation clause is quantified based on the type of the violation clause and the priority of the violated rule, and the risk level of the violation clause is determined; risk positioning, type classification, and level quantification are realized based on the knowledge graph and the historical case data, human subjective judgment errors are reduced, scattered violation clauses are classified into a unified type through semantic clustering, risk fragmentation is avoided, core risks are easily grasped by the reviewers, the risk level quantification clearly indicates the rectification priority, the structured presentation of the risk analysis result directly guides the modification of the bidding documents, and the review and rectification efficiency is improved.

[0030] To sum up, the intelligent review method for bidding documents provided by the application is based on a pre-trained language model to perform structured analysis on the obtained bidding documents to be reviewed to obtain a structured vector; a vector library is constructed based on the obtained historical bidding documents, the structured vector is matched with the vector library in terms of similarity, a first historical bidding document corresponding to the bidding document to be reviewed is determined based on the similarity matching result; the review rules of the first historical bidding document are obtained, a rule library is constructed based on the review rules, the rules and the enterprise rules, the bidding document to be reviewed is checked for compliance based on the rule library to obtain a rule checking result; the structured vector is analyzed for risks based on the rule checking result and the constructed bidding and tender knowledge graph to obtain a risk analysis result, and a review report is generated based on the risk analysis result and the rule checking result, thereby improving the efficiency and accuracy of the review.

[0031] In order to better implement the intelligent review method for bidding documents in the embodiments of the application, on the basis of the intelligent review method for bidding documents, as shown in Figure 2 The application also provides an intelligent review system for bidding documents, which comprises: A structured analysis module 201 is configured to perform structured analysis on the obtained bidding documents to be reviewed based on a pre-trained language model to obtain a structured vector. A similarity matching module 202 is configured to construct a vector library based on the obtained historical bidding documents, match the structured vector with the vector library in terms of similarity, and determine a first historical bidding document corresponding to the bidding document to be reviewed based on the similarity matching result. A rule checking module 203 is configured to obtain the review rules of the first historical bidding document, construct a rule library based on the review rules, the rules and the enterprise rules, check the bidding document to be reviewed for compliance based on the rule library, and obtain a rule checking result. A review report generation module 204 is configured to analyze the structured vector for risks based on the rule checking result and the constructed bidding and tender knowledge graph to obtain a risk analysis result, and generate a review report based on the risk analysis result and the rule checking result.

[0032] As shown in Figure 3 The application also provides an electronic device 300, which can be a mobile terminal, a desktop computer, a notebook, a palm computer or a server. The electronic device 300 comprises a processor 301, a memory 302 and a display 303. Figure 3 Only some components of the electronic device 300 are shown, but it should be understood that all the shown components are not required to be implemented, and more or fewer components can be alternatively implemented.

[0033] The memory 302 can be an internal storage unit of the electronic device 300, such as a hard disk or a memory of the electronic device 300 in some embodiments. The memory 302 can also be an external storage device of the electronic device 300, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 300 in other embodiments. Further, the memory 302 can include both an internal storage unit and an external storage device of the electronic device 300. The memory 302 is used to store application software and various data installed on the electronic device 300, such as program codes installed on the electronic device 300. The memory 302 can also be used to temporarily store data that has been output or will be output. In an embodiment, the memory 302 stores a tender file intelligent review program, which can be executed by the processor 301 to implement the tender file intelligent review method of various embodiments of the present application.

[0034] The processor 301 can be a Central Processing Unit (CPU), a microprocessor, or other data processing chip in some embodiments, and is used to run program codes or process data stored in the memory 302, such as the tender file intelligent review method.

[0035] The display 303 can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. in some embodiments. The display 303 is used to display identification information of the tender file intelligent review program and to display a visualized user interface. The components 301-303 of the electronic device 300 communicate with each other through a system bus.

[0036] In some embodiments, the processor 301 implements each step of the tender file intelligent review method as described in the above embodiments when executing the tender file intelligent review program in the memory 302. Since the tender file intelligent review method has been described in detail above, no further description is given here.

[0037] Accordingly, the present application also provides a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps or functions of the tender file intelligent review method provided by the above method embodiments.

[0038] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiment methods can be completed by instructing the relevant hardware by a computer program, and the program can be stored in a computer readable storage medium. The computer readable storage medium is a disk, an optical disk, a read-only memory, a random access memory, etc.

[0039] The above description is merely preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for intelligent review of a tender document, characterized by, The method comprises the following steps: Based on the pre-trained language model, the obtained to-be-reviewed bidding document is structurally parsed to obtain a structured vector; Based on the obtained historical bidding documents, a vector library is constructed, the structured vector is matched with the vector library in terms of similarity, and based on the similarity matching result, a first historical bidding document corresponding to the to-be-reviewed bidding document is determined in the historical bidding documents; The review rules of the first historical bidding document are obtained, a rule library is constructed based on the review rules, legal rules and enterprise rules, the to-be-reviewed bidding document is checked for compliance based on the rule library, and a rule checking result is obtained; Based on the rule checking result and the constructed bidding and tendering knowledge graph, the structured vector is analyzed for risks to obtain a risk analysis result, and an review report is generated based on the risk analysis result and the rule checking result.

2. The method of claim 1, wherein, The method comprises the following steps: The to-be-reviewed bidding document is split according to the legal structure of the bidding document to obtain module texts; The module texts are processed for sentences to obtain text sequences; Based on the pre-trained language model, the text sequences are subjected to entity recognition, relationship extraction and text classification to obtain structured information; The structured information is normalized to obtain a structured vector.

3. The method of claim 2, wherein, The structured information at least includes bidding basic information, project core information, participant information, process rule information and compliance information.

4. The method of claim 2, wherein, The method comprises the following steps: The cosine similarity is used to calculate the similarity value of the structured vector and the vector library; Based on the preset similarity threshold and the similarity value, the to-be-reviewed bidding document and the historical bidding documents are matched to obtain a similarity matching result; Based on the similarity matching result, a first historical bidding document corresponding to the to-be-reviewed bidding document is determined in the historical bidding documents.

5. The method of claim 4, wherein, The cosine similarity is: ， in, For structured vectors, Vectors of historical tender documents in the vector library. For vector dimensions, For the structured vector, the first A vector, For the first in the vector library A vector.

6. The method of claim 4, wherein, The method comprises the following steps: Based on the rule library, the general compliance bottom line of the to-be-reviewed bidding document is checked to obtain a general compliance checking result; Based on the rule library, the mandatory requirements of the to-be-reviewed bidding document are checked to obtain a legal compliance checking result; Based on the rule library, the enterprise individualized standards of the to-be-reviewed bidding document are checked to obtain an enterprise customized compliance checking result; Based on the general compliance checking result, the legal compliance checking result and the enterprise customized compliance checking result, a rule checking result is obtained.

7. The method of claim 6, wherein, The method comprises the following steps: Based on the rule checking result, the violation clauses of the structured vector are determined, and based on the constructed bidding and tendering knowledge graph, the core nodes of the violation clauses are determined; determine a violation type based on the bidding knowledge graph, and perform similarity calculation on the core node of the violation clause and the violation type node to determine a violation type; quantify a risk level of the violation clause based on historical case data to determine a risk level of the violation clause; generate a risk analysis result based on the violation clause, violation type, and risk level.

8. An intelligent review system for a tender document, characterized in that, The method comprises the following steps: a structured analysis module configured to perform structured analysis on the obtained bidding document to be reviewed based on a pre-trained language model to obtain a structured vector; a similarity matching module configured to construct a vector library based on obtained historical bidding documents, perform similarity matching on the structured vector and the vector library, and determine a first historical bidding document corresponding to the bidding document to be reviewed based on a similarity matching result; a rule verification module configured to obtain review rules of the first historical bidding document, construct a rule library based on the review rules, legal rules, and enterprise rules, perform compliance verification on the bidding document to be reviewed based on the rule library, and obtain a rule verification result; an audit report generation module configured to perform risk analysis on the structured vector based on the rule verification result and the constructed bidding knowledge graph to obtain a risk analysis result, and generate an audit report based on the risk analysis result and the rule verification result.

9. An electronic device, comprising: comprises a memory and a processor; the memory stores a computer readable program that can be executed by the processor; the processor executes the computer readable program to implement the steps of the bidding document intelligent review method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, a program or instruction readable by a computer, which can implement the steps of the bidding document intelligent review method according to any one of claims 1-7 when executed by a processor.

Images