A method and system for intelligent compliance checking of a tender

Abstract

The application discloses a kind of intelligent compliance inspection methods and systems of bidding and collecting, it is related to bidding and collecting text processing technical field, the application includes: step 1.data set construction, step 2.feature analysis, step 3.generate compliance inspection report, by extracting the text content of tender document and tender document, paragraph structure and document attribute, constructs three kinds of data sets of tender demand label, tender response content, document attribute record, based on the corresponding relationship of tender demand item and tender response item, analysis response matching degree of each tender demand, text similarity between different tender documents and time regularity characteristics, finally based on the satisfaction determination result of each tender demand and the grade of stringing bid between bidders, generate structured, traceable bidding and collecting compliance inspection report, realize review precision, process standardization and result visualization.

Description

Technical Field

[0001] This invention relates to the field of procurement text processing technology, specifically to a method and system for intelligent compliance inspection of procurement. Background Technology

[0002] With the continuous advancement of digital technology, procurement-related work is gradually moving online, and electronic document processing and information exchange technologies are maturing. Technologies such as natural language processing, text parsing, and data feature analysis are constantly iterating and being applied to procurement document review scenarios. These technologies are evolving from basic rule comparison to semantic understanding, content association analysis, and anomaly feature identification. Multi-dimensional intelligent detection, data linkage analysis, and automatic identification and judgment functions are continuously improving, driving the procurement compliance inspection towards intelligent, standardized, and integrated development.

[0003] Existing technologies, such as the invention patent applications related to procurement compliance checks disclosed in announcement numbers CN121456514A and CN120337873A, have significant shortcomings upon comparison: existing procurement compliance check solutions can only achieve single-dimensional text comparison or basic clause verification, and cannot perform structured decomposition and requirement tagging of bidding documents and tender documents. They are unable to accurately locate the correspondence between bidding requirements and tender responses, cannot differentiate between mandatory and non-mandatory requirements to determine the degree of response matching, and lack quantitative analysis of key indicators such as technical parameters, qualification requirements, and delivery deadlines. In addition to the comparison, most solutions do not combine the creation time, modification time, author, saver and other document attributes of the tender documents to conduct time pattern and operation subject correlation analysis. They cannot comprehensively determine the level of bid rigging from multiple dimensions such as text similarity, abnormal time concentration and operation subject overlap, and it is difficult to form a complete risk classification and identification mechanism. In addition, the existing solutions cannot output the requirements fulfillment details and bid rigging analysis results at the same time, and cannot generate an integrated and traceable procurement compliance inspection report. The review coverage is not comprehensive, the judgment criteria are not precise and the results are not fully presented, which makes it difficult to meet the actual needs of intelligent compliance management of the entire procurement process. Summary of the Invention

[0004] To address the aforementioned technical shortcomings, the present invention aims to provide a method and system for intelligent compliance inspection in procurement.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The first aspect of the present invention provides a method for intelligent compliance inspection of bidding and procurement, including step 1. Data set construction: extracting text content, paragraph structure and document attributes from the bidding documents and all bid documents respectively, and constructing a bidding requirement tag dataset, a bid response content dataset and a document attribute record dataset respectively.

[0006] Step 2. Feature Analysis: Based on the bidding requirement tag dataset and the bid response content dataset, analyze the response matching degree of each bidding requirement. Based on the bid response content dataset and the document attribute record dataset, analyze the text similarity and time pattern characteristics between different bid documents.

[0007] Step 3. Generate a compliance inspection report: Based on the response matching degree of each bidding requirement, determine the satisfaction judgment result of each bidding requirement. Based on the text similarity and time pattern characteristics between different bid documents, comprehensively determine the level of collusion among bidders. Based on the satisfaction judgment result of each bidding requirement and the level of collusion among bidders, generate a procurement compliance inspection report.

[0008] A second aspect of the present invention provides a system for an intelligent compliance inspection method for bidding and procurement, comprising a dataset construction module: used to extract text content, paragraph structure and document attributes from bidding documents and all bid documents respectively, and to construct a bidding requirement tag dataset, a bid response content dataset and a document attribute record dataset respectively.

[0009] Feature Analysis Module: This module analyzes the response matching degree of each bidding requirement based on the bidding requirement tag dataset and the bid response content dataset. It also analyzes the text similarity and temporal pattern characteristics between different bid documents based on the bid response content dataset and the document attribute record dataset.

[0010] The compliance inspection report generation module is used to determine the satisfaction result of each bidding requirement based on the response matching degree of each bidding requirement, comprehensively determine the level of collusion among bidders based on the text similarity and time pattern characteristics between different bid documents, and generate a procurement compliance inspection report based on the satisfaction result of each bidding requirement and the level of collusion among bidders.

[0011] The beneficial effects of the present invention are as follows: (1) The first part of the present invention: by extracting text, parsing paragraph structure and collecting document attributes from bidding documents and tender documents, unstructured procurement electronic documents can be transformed into standardized and comparable datasets, realizing the unified collection and structured storage of bidding requirements, tender responses and document metadata, which can stably support subsequent accurate comparison and anomaly identification, improve the integrity and consistency of document processing, and provide a reliable data foundation for compliance inspection.

[0012] (2) The second part of the present invention: Based on the bidding requirements and bid responses, a matching calculation is carried out item by item to accurately distinguish the degree of response satisfaction of mandatory and non-mandatory clauses. At the same time, multi-dimensional analysis is carried out by combining text similarity, time pattern characteristics and consistency of operating subjects to realize the simultaneous execution of response compliance judgment and bid rigging suspicion identification, improve the precision and credibility of inspection judgment, reduce the situation of missed judgment and misjudgment, and ensure the rigor of the bidding and procurement process review.

[0013] (3) The third part of the present invention: automatically summarizes the demand response results and the information on the level of collusion, integrates them to form a structured and traceable procurement compliance inspection report, clearly presents the status of each demand and the abnormal correlation of the bidder, facilitates the quick location of problems and the review, reduces the workload of manual verification and summary, improves the efficiency of review output, and provides an intuitive and standardized decision-making basis for the compliance management of the entire procurement process. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0016] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0017] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 As shown, the present invention provides a method for intelligent compliance inspection of bidding and procurement, including step 1. Dataset construction: extracting text content, paragraph structure and document attributes from the bidding documents and all bid documents respectively, and constructing a bidding requirement tag dataset, a bid response content dataset and a document attribute record dataset respectively.

[0019] It should be noted that the paragraph structure refers to the hierarchical organization of bidding documents and tender documents, consisting of chapter titles, clause numbers, paragraph breaks, list markers, and table structures. This is used to locate the correspondence between requirement items and response items and to identify the consistency of formatting between different tender documents. The document attributes refer to the metadata information attached to the electronic file, including but not limited to file creation time, last modification time, author name, last saver, print records, and file path. This is used to analyze the correlation characteristics of different tender documents in terms of time patterns and operation traces.

[0020] In a specific embodiment of the present invention, the method for constructing the bidding requirement tag dataset is as follows: based on the chapter titles and clause numbers in the paragraph structure of the bidding document, each independent bidding requirement item is identified and segmented; the requirement description text of each bidding requirement item is extracted; using keyword extraction and semantic role labeling methods, technical parameters, qualification requirements, and delivery deadlines are extracted from the requirement description text; a unique requirement code is assigned to each bidding requirement item in the bidding document; and the mandatory requirement identifier corresponding to each bidding requirement item is determined.

[0021] Each tender requirement item's unique code, technical parameters, qualification requirements, delivery deadline, mandatory requirement identifier, and original text summary are used as requirement tag fields. All requirement tag fields corresponding to all tender requirements items are aggregated and stored to form a tender requirement tag dataset.

[0022] The unique code for each requirement is used to uniquely identify each requirement within the tender document, facilitating subsequent matching of bid responses. For example, the code REQ-03-02-01 represents the first requirement in Section 2 of Chapter 3. Technical parameters are used to describe quantifiable and comparable specific indicators in technical requirements, and are the key basis for determining whether a bid response meets the requirements. Example: The tender requirement is that the server CPU frequency is not less than 2.5GHz, and the technical parameter is the CPU frequency, with an indicator value of greater than or equal to 2.5GHz.

[0023] The qualification requirements are used to record the eligibility conditions that bidders must meet, typically including certifications, licenses, performance records, etc. Example: The tender requirement states that bidders must possess a Class II or higher qualification for general contracting of building construction projects. Delivery deadlines specify the delivery or completion time commitments to be made in the tender documents. Example: The tender requirement states that delivery must be completed within 30 calendar days after contract signing. Mandatory requirements are used to distinguish whether a requirement is an inviolable substantive requirement. Requirements marked as mandatory will result in an invalid bid if the tender documents do not respond or do not meet the requirements. Requirements marked as non-mandatory allow for some deviation or can be used as scoring items. Example: Clauses marked with a five-pointed star in the tender documents usually correspond to mandatory requirements.

[0024] It should be noted that keyword extraction and semantic role labeling are existing technologies in the field of natural language processing. Keyword extraction refers to the technique of automatically identifying and extracting representative and thematic words or phrases from text. Commonly used methods include statistical TF-IDF and TextRank, as well as deep learning-based sequence labeling models. Semantic role labeling refers to the technique of identifying and classifying predicates and arguments in text, such as agent, patient, time, and place, to parse the semantic structure of who did what to whom in a sentence. Keyword extraction and semantic role labeling methods are relatively mature and will not be described in detail here.

[0025] In a specific embodiment of the present invention, the method for constructing the bid response content dataset is as follows: based on the unique code of each requirement and its requirement description text in the bid requirement tag dataset, locate the response content position corresponding to each bid requirement item in the bid document, extract the response description text at the response content position, and parse the commitment indicators and deviation descriptions from the response description text through a semantic matching method.

[0026] Each bid response item is associated with a unique requirement code and a bidder identifier. The unique requirement code, bidder identifier, commitment indicators, deviation description, and original text summary of each bid response item are used as response tag fields. All bid response items are aggregated and stored to form a bid response content dataset.

[0027] It should be noted that semantic matching methods refer to techniques that determine the consistency of content through implication relationships. Commonly used methods include cosine similarity combined with word vectors, Siamese networks, and cross-encoders based on Transformer, which are used to extract structured information from demand text and response text. Those skilled in the art can flexibly configure these methods based on conventional experience and specific scenarios without affecting the implementation and reproduction of the technical solution of this invention.

[0028] The commitment indicators are the specific values ​​or conditions given by the bidder for the technical parameters, qualification requirements, and delivery deadlines in the tender requirements.

[0029] It should be noted that deviation explanation refers to the response in the tender document to a specific tender requirement that is not entirely consistent with the tender requirements, along with an explanation of the reasons. In procurement compliance checks, deviations are categorized into three types: no deviation, positive deviation that is better than the tender requirements, and negative deviation that is lower than or does not meet the tender requirements. Deviation explanations typically include the specific content of the deviation, the degree of deviation, the reasons for the deviation, and whether there is a commitment to correct it during performance. For example, if the tender requires a CPU frequency of no less than 2.5GHz, and the tender document responds with a CPU frequency of 2.3GHz, explaining that it cannot be met temporarily due to supply chain reasons, but promises to upgrade after winning the bid, then this deviation explanation is recorded as a frequency of 2.3GHz, lower than the required 2.5GHz, a negative deviation, and a commitment to upgrade. This field is used in subsequent steps to determine whether the requirement meets the tender requirements and to assess whether the negative deviation constitutes a substantial non-responsiveness.

[0030] In a specific embodiment of the present invention, the method for constructing the document attribute record dataset is as follows: extract the creation time, last modification time, author name, and last saver of each bid document, associate each bid document with a corresponding bidder identifier, and use the bidder identifier, file creation time, last modification time, author name, and last saver of each bid document as document attribute fields, and summarize and store them to form a document attribute record dataset.

[0031] It should be noted that the file creation time and last modification time are used to compare whether the generation times of different bid documents are highly concentrated or have an abnormal order, and the author name and last saver are used to identify the relationship between the operating entities of different bid documents.

[0032] Step 2. Feature Analysis: Based on the bidding requirement tag dataset and the bid response content dataset, analyze the response matching degree of each bidding requirement. Based on the bid response content dataset and the document attribute record dataset, analyze the text similarity and time pattern characteristics between different bid documents.

[0033] In a specific embodiment of the present invention, the method for analyzing the response matching degree of each bidding requirement is as follows: traverse the unique code of each bidding requirement item in the bidding requirement label dataset, select all bidding response items corresponding to the unique code from the bidding response content dataset, compare the committed indicators of each bidding response item with the technical parameters, qualification requirements and delivery period in the bidding requirement item item by item, determine the deviation result in combination with the deviation description, and determine the response matching degree of each bidding requirement based on the mandatory identifier.

[0034] In one specific embodiment, the promised indicators are compared item by item with the technical parameters, qualification requirements, and delivery deadlines in the tender requirements. The deviation result is determined in conjunction with the deviation description. The specific method is as follows: if the promised indicators meet the technical parameters, qualification requirements, and delivery deadlines in the tender requirements, the theoretical result is determined to be no deviation. If all promised indicators are better than the technical parameters, qualification requirements, and delivery deadlines in the tender requirements, the theoretical result is determined to be a positive deviation. If any promised indicator does not meet the technical parameters, qualification requirements, and delivery deadlines in the tender requirements, the theoretical result is determined to be a negative deviation. The self-described deviation content in the deviation description is then read. If the two are consistent, the theoretical judgment result is directly adopted. If the two are inconsistent, the theoretical comparison result is taken as the actual deviation result, and the deviation description is marked as abnormal.

[0035] In one specific embodiment, the response matching degree of each bidding requirement is determined based on the mandatory identifier. The specific method is as follows: if the mandatory identifier of the bidding requirement is mandatory and the deviation result is negative, the response matching degree is 0%; if the mandatory identifier is mandatory and the deviation result is positive or no deviation, the response matching degree is 100%; if the mandatory identifier is non-mandatory and the deviation result is negative, the number of indicators with negative deviation is divided by the total number of all indicators included in the bidding requirement, and the resulting ratio is used as the response matching degree; if the mandatory identifier is non-mandatory and the deviation result is positive or no deviation, the response matching degree is 100%.

[0036] For example, if a mandatory requirement is marked as non-mandatory and the deviation result is negative, the number of indicators with negative deviation is divided by the total number of all indicators included in the tender requirement, and the resulting ratio is used as the response matching degree. For example, if a non-mandatory tender requirement includes 5 technical indicators, and one indicator in the bidder's response has a negative deviation, then the number of negative deviation indicators is 1, the total number of indicators is 5, and the ratio is 1 / 5 = 0.2, or 20%. Therefore, the response matching degree of this bid response item is assigned to 20%. The more negative deviation indicators there are, the lower the matching degree.

[0037] In a specific embodiment of the present invention, the method for analyzing the text similarity and temporal pattern characteristics between different bid documents is as follows: extract the original text summaries of all bid response items from the bid response content dataset, concatenate all response summary texts of each bid document into a single document, and calculate the text similarity between different bid documents using a similarity calculation method.

[0038] Extract the file creation time, last modification time, author name, and last saver of each tender document from the document attribute record dataset. Calculate the difference in creation time and last modification time between any two tender documents, compare the consistency of the author name and last saver between the two tender documents, and count the number of identical items. Use the difference in creation time, the difference in last modification time, and the number of identical items as time pattern features.

[0039] It should be noted that the text similarity is used to measure the degree of similarity in the response content of different bid documents. Jaccard similarity or similarity calculation method based on edit distance can be used. The higher the similarity score, the more similar the response content between the bid documents is. The above calculation methods are all common knowledge in the prior art. Those skilled in the art can flexibly configure them based on conventional experience and specific scenarios without affecting the implementation and reproduction of the technical solution of the present invention.

[0040] It should be noted that the time pattern characteristics include two aspects: time difference and consistency of the operating entity. The time difference is used to measure the proximity of the time in which the bid documents are generated and modified, while the consistency of the operating entity is used to measure whether different bid documents are created or modified by the same person. The combination of the two can effectively identify the common characteristics of centralized production and the same operating source in bid rigging behavior.

[0041] Step 3. Generate a compliance inspection report: Based on the response matching degree of each bidding requirement, determine the satisfaction judgment result of each bidding requirement. Based on the text similarity and time pattern characteristics between different bid documents, comprehensively determine the level of collusion among bidders. Based on the satisfaction judgment result of each bidding requirement and the level of collusion among bidders, generate a procurement compliance inspection report.

[0042] In a specific embodiment of the present invention, the method for determining the satisfaction result of each bidding requirement is as follows: traverse the response matching degree of each bidder under each bidding requirement item. When the response matching degree of a bidder for a certain bidding requirement item is greater than or equal to a preset satisfaction threshold, the bidder is determined to satisfy the bidding requirement item; otherwise, it is determined not to satisfy.

[0043] In a specific embodiment of the present invention, the method for comprehensively determining the level of collusion between bidders is as follows: if the text similarity between two bid documents is greater than or equal to a preset similarity threshold, then it is determined that there is text similarity between the bidders.

[0044] If the time difference between the creation of two bid documents and the time difference between their last modification are both less than the preset time difference threshold, then it is determined that there is an abnormal concentration of time among the bidders.

[0045] If the number of identical entries between the author's name and the last saver in two bid documents is greater than or equal to a preset threshold, then it is determined that there is overlap in the operating entities between the bidders.

[0046] If there is no textual similarity, abnormally concentrated time, or overlapping operating entities between two bid documents, the level of collusion between the bidders is determined to be Level 1.

[0047] If two bid documents exhibit any of the following characteristics: textual similarity, unusually concentrated timing, or overlapping operating entities, then the level of collusion between the bidders is determined to be Level 2.

[0048] If two bid documents are similar in text, have an unusually concentrated time period, or have the same operating entity, then the level of collusion between the bidders is determined to be Level 3.

[0049] When two bid documents exhibit textual similarity, unusually concentrated timing, and overlapping operational entities, the level of collusion between the bidders is determined to be Level 4.

[0050] It should be noted that the level of collusion among bidders is lower than the level of collusion among bidders at level one, lower than the level of collusion among bidders at level two, lower than the level of collusion among bidders at level three, and lower than the level of collusion among bidders at level four.

[0051] It should be noted that situations where the difference between the creation time and the last modification time are both less than a preset threshold are judged as abnormal time concentrations and used as one of the indicators for determining bid rigging. The basis for this is that in a normal bidding process, different bidders independently prepare their bid documents, and the creation time and last modification time of their documents are usually randomly distributed and do not show a high concentration phenomenon. If the creation time and modification time of multiple bid documents are concentrated in the same short period of time, it is highly suspected that related bidders collaborated to produce or batch generate bid documents in the same period of time, which is one of the typical characteristics of bid rigging.

[0052] In a specific embodiment of the present invention, the method for generating the procurement compliance inspection report is as follows: summarizing the response matching degree and satisfaction judgment result of each bidder for each bidding requirement, and generating a requirement satisfaction detail table, wherein the requirement satisfaction detail table includes a unique requirement code, a summary of the original text of the requirement, a mandatory requirement identifier, response matching degree, and satisfaction judgment result.

[0053] The text similarity, time pattern characteristics, and corresponding collusion levels among different bid documents are summarized to generate a collusion analysis table. The collusion analysis table includes bidder identifiers, text similarity between different bid documents, creation time difference, last modification time difference, number of items with the same operating entity, and collusion level among bidders.

[0054] The demand fulfillment details table and the bid rigging analysis table are integrated to form a procurement compliance inspection report.

[0055] For example, a procurement compliance inspection report might show the following: The requirement fulfillment details table indicates that: Requirement REQ-01 is a server CPU with a clock speed greater than or equal to 2.5GHz, marked as a mandatory requirement. Bidder A meets this requirement 100%, Bidder B meets it 0%, and Bidder C meets it 100%. Requirement REQ-02 is ISO9001 certification, marked as a mandatory requirement. Bidder A meets this requirement, Bidder B meets this requirement, and Bidder C meets this requirement. Requirement REQ-03 is a delivery period of 30 days, marked as a non-mandatory requirement. Bidder A meets this requirement 100%, Bidder B meets it 60%, and Bidder C meets it. The 30% similarity requirement is not met. The collusion analysis table shows that Bidder A and Bidder B have a text similarity of 92%, a creation time difference of 5 minutes, a last modification time difference of 3 minutes, and 2 identical operation entities, indicating a collusion level of four. Bidder A and Bidder C have a text similarity of 45%, time differences of 120 minutes and 150 minutes, and 0 identical items, indicating a collusion level of one. Bidder B and Bidder C have a text similarity of 88%, time differences of 10 minutes and 8 minutes, and 1 identical item, indicating a collusion level of three. The report, after integration, recommends that Bidder B does not meet the bidding requirement REQ-01, and that Bidder A and Bidder B are highly suspected of collusion, requiring manual review.

[0056] Reference Figure 2 As shown, the present invention provides a system for intelligent compliance inspection of bidding and procurement, including a dataset construction module: used to extract text content, paragraph structure and document attributes from bidding documents and all bid documents respectively, and construct bidding requirement tag dataset, bid response content dataset and document attribute record dataset respectively.

[0057] Feature Analysis Module: This module analyzes the response matching degree of each bidding requirement based on the bidding requirement tag dataset and the bid response content dataset. It also analyzes the text similarity and temporal pattern characteristics between different bid documents based on the bid response content dataset and the document attribute record dataset.

[0058] The compliance inspection report generation module is used to determine the satisfaction result of each bidding requirement based on the response matching degree of each bidding requirement, comprehensively determine the level of collusion among bidders based on the text similarity and time pattern characteristics between different bid documents, and generate a procurement compliance inspection report based on the satisfaction result of each bidding requirement and the level of collusion among bidders.

[0059] The examples described in this invention are not limited to the specific embodiments listed above. The examples are merely illustrative to facilitate understanding of the invention and do not constitute a limitation on the scope of protection of this invention. Any modifications, equivalent substitutions, etc., made within the spirit and principles of this invention should be included within the scope of protection.

[0060] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. A method for intelligent compliance inspection in procurement, characterized in that, Includes the following steps: Step 1. Dataset Construction: Extract text content, paragraph structure, and document attributes from the tender documents and all bid documents respectively, and construct the tender requirement tag dataset, bid response content dataset, and document attribute record dataset respectively; Step 2. Feature Analysis: Based on the bidding requirement tag dataset and the bid response content dataset, analyze the response matching degree of each bidding requirement. Based on the bid response content dataset and the document attribute record dataset, analyze the text similarity and time pattern characteristics between different bid documents. Step 3. Generate a compliance inspection report: Based on the response matching degree of each bidding requirement, determine the satisfaction judgment result of each bidding requirement. Based on the text similarity and time pattern characteristics between different bid documents, comprehensively determine the level of collusion among bidders. Based on the satisfaction judgment result of each bidding requirement and the level of collusion among bidders, generate a procurement compliance inspection report.

2. The intelligent compliance inspection method for procurement as described in claim 1, characterized in that, The specific method for constructing the bidding demand label dataset is as follows: Based on the chapter titles and clause numbers in the paragraph structure of the tender document, each independent tender requirement item is identified and segmented. The requirement description text of each tender requirement item is extracted. Using keyword extraction and semantic role labeling methods, technical parameters, qualification requirements and delivery deadlines are extracted from the requirement description text. Each tender requirement item is assigned a unique requirement code in the tender document, and the mandatory requirement identifier corresponding to each tender requirement item is determined. Each tender requirement item's unique code, technical parameters, qualification requirements, delivery deadline, mandatory requirement identifier, and original text summary are used as requirement tag fields. All requirement tag fields corresponding to all tender requirements items are aggregated and stored to form a tender requirement tag dataset.

3. The intelligent compliance inspection method for procurement according to claim 2, characterized in that, The specific method for constructing the bid response content dataset is as follows: Based on the unique code and description text of each requirement in the bidding requirement tag dataset, locate the corresponding response content in the bid document, extract the response description text at the response content location, and parse the commitment indicators and deviation descriptions from the response description text using a semantic matching method. Each bid response item is associated with a unique requirement code and a bidder identifier. The unique requirement code, bidder identifier, commitment indicators, deviation description, and original text summary of each bid response item are used as response tag fields. All bid response items are aggregated and stored to form a bid response content dataset.

4. The intelligent compliance inspection method for procurement according to claim 3, characterized in that, The specific method for constructing the document attribute record dataset is as follows: For each bid document, extract its creation time, last modification time, author name, and last saver. Associate each bid document with a corresponding bidder identifier. Use the bidder identifier, document creation time, last modification time, author name, and last saver of each bid document as document attribute fields, and summarize and store them to form a document attribute record dataset.

5. The intelligent compliance inspection method for procurement according to claim 3, characterized in that, The specific method for analyzing the response matching degree of each bidding requirement is as follows: Iterate through the unique code of each tender requirement item in the tender requirement label dataset, filter out all tender response items corresponding to the unique code from the tender response content dataset, compare the promised indicators with the technical parameters, qualification requirements and delivery period in the tender requirement item item by item, determine the deviation result in combination with the deviation description, and determine the response matching degree of each tender requirement based on the mandatory identifier.

6. The intelligent compliance inspection method for procurement according to claim 4, characterized in that, The specific method for analyzing the text similarity and temporal patterns among different bid documents is as follows: Extract the original text summaries of all bid response items from the bid response content dataset, concatenate all response summary texts of each bid document into a single document, and calculate the text similarity between different bid documents using a similarity calculation method. Extract the file creation time, last modification time, author name, and last saver of each tender document from the document attribute record dataset. Calculate the difference in creation time and last modification time between any two tender documents, compare the consistency of the author name and last saver between the two tender documents, and count the number of identical items. Use the difference in creation time, the difference in last modification time, and the number of identical items as time pattern features.

7. The intelligent compliance inspection method for procurement as described in claim 5, characterized in that, The specific method for determining the satisfaction result of each bidding requirement is as follows: Iterate through the response matching degree of each bidder under each bidding requirement. When the response matching degree of a bidder for a certain bidding requirement is greater than or equal to the preset satisfaction threshold, the bidder is determined to meet the bidding requirement; otherwise, it is determined not to meet the requirement.

8. The intelligent compliance inspection method for procurement according to claim 6, characterized in that, The specific method for comprehensively determining the level of collusion among bidders is as follows: If the text similarity between two bid documents is greater than or equal to the preset similarity threshold, then it is determined that there is text similarity between the bidders. If the time difference between the creation of two bid documents and the time difference between the last modification are both less than the preset time difference threshold, it is determined that there is an abnormal concentration of time among the bidders. If the number of identical entries between the author's name and the last saver in two bid documents is greater than or equal to a preset threshold for the number of identical entries, it is determined that there is an overlap in the operating entities between the bidders. If there is no text similarity, abnormally concentrated time, or overlapping operating entities between two bid documents, the level of collusion between the bidders is determined to be Level 1. If two bid documents are similar in text, have an unusually concentrated time period, or have the same operating entity, the level of collusion between the bidders is determined to be Level 2. If two bid documents are similar in text, have an unusually concentrated time period, or have the same operating entity, then the level of collusion between the bidders is determined to be Level 3. When two bid documents exhibit textual similarity, unusually concentrated timing, and overlapping operational entities, the level of collusion between the bidders is determined to be Level 4.

9. The intelligent compliance inspection method for procurement according to claim 1, characterized in that, The specific method for generating the procurement compliance inspection report is as follows: Summarize the response matching degree and satisfaction judgment result of each bidder for each bidding requirement, and generate a requirement satisfaction detail table. The requirement satisfaction detail table includes a unique requirement code, a summary of the original text of the requirement, a mandatory requirement identifier, response matching degree and satisfaction judgment result. The text similarity, time pattern characteristics, and corresponding collusion levels among different bid documents are summarized to generate a collusion analysis table. The collusion analysis table includes bidder identifiers, text similarity between different bid documents, creation time difference, last modification time difference, number of items with the same operating entity, and collusion level among bidders. The demand fulfillment details table and the bid rigging analysis table are integrated to form a procurement compliance inspection report.

10. A system for implementing the intelligent compliance inspection method for procurement as described in any one of claims 1-9, characterized in that, Includes the following modules: Dataset construction module: used to extract text content, paragraph structure and document attributes from the tender documents and all bid documents respectively, and to build a tender requirement label dataset, a bid response content dataset and a document attribute record dataset respectively; Feature Analysis Module: Used to analyze the response matching degree of each bidding requirement based on the bidding requirement tag dataset and the bid response content dataset, and to analyze the text similarity and time pattern features between different bid documents based on the bid response content dataset and the document attribute record dataset; The compliance inspection report generation module is used to determine the satisfaction result of each bidding requirement based on the response matching degree of each bidding requirement, comprehensively determine the level of collusion among bidders based on the text similarity and time pattern characteristics between different bid documents, and generate a procurement compliance inspection report based on the satisfaction result of each bidding requirement and the level of collusion among bidders.

Images