CN-121981808-A - Intelligent auxiliary bid evaluation method and system based on AI policy reasoning

CN121981808ACN 121981808 ACN121981808 ACN 121981808ACN-121981808-A

Abstract

The invention provides an intelligent auxiliary bid evaluation method and system based on AI policy reasoning, and relates to the technical field of intelligent bid evaluation and bid information processing. Further extracting text semantic features, unstructured graphic topological features and document DNA features, generating a multi-mode homologous detection fingerprint set, performing clause level analysis on the evaluation rules and generating executable rule intermediate representation and constraint sets on the basis, generating a scoring candidate scheme through AI policy reasoning, and performing policy optimization on scoring results, risk factors and evidence integrity on the premise of meeting rule constraints to determine an optimal scoring scheme. The method and the device can improve the interpretability, consistency and risk identification capability of the bid evaluation process, and are suitable for intelligent auxiliary bid evaluation application in complex bidding scenes.

Inventors

XU XIAODONG
SHU LIYA
LI JILIANG
Jin Zhuoyang
Wang Juhang

Assignees

中航招采科技(北京)有限公司

Dates

Publication Date: 20260505
Application Date: 20260120

Claims (10)

1. The intelligent auxiliary bid evaluation method based on AI policy reasoning is characterized by comprising the following steps: S1, acquiring multiple sets of bidding data corresponding to the same bid evaluation item, performing format analysis and field alignment on the bidding data, generating bidding feature vectors of all bidding parties, and cutting text paragraphs, table units and accessory items which can be positioned and referenced into evidence fragment sets; s2, extracting multi-mode semantic fingerprints from the data fragment set, and extracting document DNA features representing editing sources to form a homologous detection fingerprint set of a bidding party; S3, carrying out distance measurement on homologous detection fingerprint sets of different bidders based on a twin neural network, and calculating the homology probability; s4, acquiring a scoring rule text, extracting scoring items, overruling items and mutual exclusion conditions, compiling the scoring rule into a computable rule intermediate representation, and generating a constraint set, wherein the constraint set is used for carrying out consistency check on scoring results; and S5, performing AI policy reasoning to generate a scoring candidate set based on the bidding feature vector, the evidence segment set and bidder ring risk prompt information, performing policy optimization on the scoring candidate set to optimize a preset objective function on the premise of meeting the constraint set, and outputting an optimal scoring scheme and evidence segment references corresponding to each scoring item.
2. The intelligent auxiliary bid evaluation method based on AI policy reasoning according to claim 1, further comprising S6, executing rule consistency check on the optimal scoring scheme, outputting conflict positioning information and triggering candidate rollback or local reasonement if conflict exists, and outputting a final auxiliary bid evaluation result.
3. The intelligent auxiliary bid evaluation method based on AI policy reasoning of claim 1, wherein S1 is specifically: multiple sets of bidding data corresponding to the same bid evaluation item are obtained, wherein the bidding data at least comprise bidding document text, quotation list data, qualification and performance proving data and bidding submitting metadata, and a bidding data set is established according to a bidding party; performing format analysis, field standardization and missing completion on the bidding data set to obtain structured bidding data, and forming a unified field view for subsequent feature extraction; cutting the bidding document text, the quotation data, the qualification and performance proving data into evidence segment sets according to paragraphs, clauses, table units and accessory items, and generating unique evidence segment identifiers for each evidence segment to support subsequent positioning references; And extracting bidding feature vectors based on the structured bidding data, and storing the bidding feature vectors and the evidence segment sets in a correlated manner according to the bidding party to obtain a consistent input structure of the bidding party, the bidding feature vectors and the evidence segment sets.
4. The intelligent auxiliary bid evaluation method based on AI policy reasoning of claim 1, wherein S2 is specifically: reading a evidence fragment set, determining a processing mode for each evidence fragment according to the type of the evidence fragment, and completing necessary normalization pretreatment to ensure that the subsequent characteristics are comparable; extracting text semantic features from the text type evidence segments, generating text semantic fingerprint components, and binding the text semantic fingerprint components with corresponding evidence segment identifiers; extracting topological structure features from unstructured graphic evidence segments, generating topological structure fingerprint components, and binding the topological structure fingerprint components with corresponding evidence segment identifiers; And fusing the text semantic fingerprint component, the topological structure fingerprint component and the document DNA characteristic to form a homologous detection fingerprint set, and summarizing and storing according to a bidding party.
5. The intelligent auxiliary bid evaluation method based on AI policy reasoning of claim 1, wherein S3 is specifically: Constructing homologous detection pairs of bidders in a pairwise manner based on the homologous detection fingerprint set, and establishing fingerprint pair input to be calculated for each homologous detection pair; inputting each homologous detection pair into the twin neural network to obtain a fingerprint distance measurement result, wherein the fingerprint distance measurement result is used for representing the similarity degree of two bidding parties on a homologous detection fingerprint set; calculating the homology probability based on the fingerprint distance measurement result, and comparing the homology probability with a preset threshold condition to obtain a homology judgment result; When the homology judging result meets the preset threshold condition, generating bidder ring risk prompt information, and outputting an evidence segment identification set corresponding to the bidder ring risk prompt information to form a risk evidence chain capable of being checked and positioned.
6. The intelligent auxiliary bid evaluation method based on AI policy reasoning of claim 1, wherein S4 is specifically: obtaining a bid evaluation rule text corresponding to a bid evaluation item, and performing clause level segmentation on the bid evaluation rule text to form a rule clause set; extracting scoring items, overruling items and mutual exclusion conditions from the rule clause set, and mapping the scoring items, overruling items and mutual exclusion conditions into computable scoring item fields and judgment fields to form a rule element structure; Compiling and generating a rule intermediate representation based on the rule element structure, so that the rule intermediate representation can be directly called by a follow-up consistency verification process; a constraint set is generated from the rule intermediate representation and an integrity check is performed on the constraint set to ensure that each scoring item and each overruling item have an executable decision constraint in the constraint set.
7. The intelligent auxiliary bid evaluation method based on AI policy reasoning of claim 1, wherein S5 specifically comprises: Inputting the bidding feature vector, the evidence segment set, bidder ring risk prompt information and the constraint set into an AI policy reasoning process to generate a scoring candidate set, wherein the scoring candidate set comprises a plurality of candidate scoring schemes; Aiming at each candidate scoring scheme in the scoring candidate set, generating an evidence segment identification quotation set corresponding to each scoring item, so that each scoring item has a locatable evidence support; On the premise of meeting the constraint set, executing strategy optimization on the scoring candidate set to optimize a preset objective function, and taking bidder ring risk prompt information as risk constraint or penalty item to participate in solving in the strategy optimization process so as to inhibit the high risk candidate scoring scheme.
8. The intelligent auxiliary bid evaluation method based on AI policy reasoning of claim 7, wherein S5 further comprises: And determining an optimal scoring scheme from the scoring candidate set according to the strategy optimization result, and outputting the optimal scoring scheme and the evidence segment identification reference set corresponding to the optimal scoring scheme as input of subsequent consistency verification.
9. The intelligent auxiliary bid evaluation method based on AI policy reasoning according to claim 2, wherein S6 specifically comprises: executing rule consistency check on the optimal scoring scheme based on the constraint set to obtain a consistency check result; when the consistency check result shows that the conflict exists, conflict positioning information is generated, wherein the conflict positioning information at least comprises rule clause identifiers for triggering the conflict, corresponding scoring item fields and associated evidence segment identifiers; triggering candidate rollback or local re-reasoning based on conflict positioning information, and re-executing AI strategy reasoning and strategy optimization on scoring items affected by the conflict to generate a new optimal scoring scheme meeting a constraint set; And outputting a final auxiliary evaluation result which at least comprises a grading detail, bidder ring risk prompt information, an evidence segment identification reference set and an reasoning process abstract so as to support examination and rechecking and result tracing.
10. An intelligent auxiliary bid evaluation system based on AI policy reasoning adopts an intelligent auxiliary bid evaluation method based on AI policy reasoning according to any one of claims 1-9, and is characterized by comprising a bid data processing module, an evidence segment generation and management module, a bid feature construction module, a multi-mode homology detection fingerprint construction module, a homology detection and bidder ring risk analysis module, a bid evaluation rule analysis and constraint generation module, an AI policy reasoning and scoring candidate generation module, a policy optimization and optimal scoring scheme determination module, a consistency verification and conflict processing module and a result output and tracing management module.

Description

Intelligent auxiliary bid evaluation method and system based on AI policy reasoning Technical Field The invention relates to the technical field of intelligent bid evaluation and bid information processing, in particular to an intelligent auxiliary bid evaluation method and system based on AI policy reasoning. Background In the scenes of engineering construction, government purchasing, large-scale equipment bidding and the like, the bid evaluation process usually involves a large number of bidding documents with complex structures and heterogeneous sources, and the contents of the bid evaluation process not only comprise text descriptions, but also comprise a quotation list, a qualification certificate, performance materials, unstructured technical documents such as a system architecture diagram, a flow chart and the like. With the continuous increase of the size and complexity of bidding documents, the conventional bid evaluation system has the following defects in actual operation: Most of the existing bid evaluation systems focus on surface field comparison or text matching of bidding documents, unified modeling of multi-mode evidence association relations inside the bidding documents is difficult, stable and locatable mapping relations between scoring results and specific evidence are lacked, and automatic rechecking and consistency verification are difficult to realize from a system level once disputes occur. Secondly, under the complex scene of participation of multiple bidders, hidden homologous relations formed by content rewriting, structure adjustment or template multiplexing may exist between the bid files, and the relations are not recognized by simple text similarity, so that the existing bid evaluation technology is difficult to reliably detect under the condition of not depending on manual experience. Therefore, we propose an intelligent auxiliary bid evaluation method and system based on AI policy reasoning, the above information disclosed in the background section is only for enhancing understanding of the background of the disclosure, so it may include information that does not constitute prior art known to those of ordinary skill in the art. Disclosure of Invention The invention aims at overcoming the defects of the prior art, and provides an intelligent auxiliary bid evaluation method and system based on AI policy reasoning, which solve the technical problems in the background art. In order to achieve the above purpose, the present invention provides the following technical solutions: An intelligent auxiliary bid evaluation method based on AI policy reasoning comprises the following steps: S1, acquiring multiple sets of bidding data corresponding to the same bid evaluation item, performing format analysis and field alignment on the bidding data, generating bidding feature vectors of all bidding parties, and cutting text paragraphs, table units and accessory items which can be positioned and referenced into evidence fragment sets; s2, extracting multi-mode semantic fingerprints from the data fragment set, and extracting document DNA features representing editing sources to form a homologous detection fingerprint set of a bidding party; S3, carrying out distance measurement on homologous detection fingerprint sets of different bidders based on a twin neural network, and calculating the homology probability; s4, acquiring a scoring rule text, extracting scoring items, overruling items and mutual exclusion conditions, compiling the scoring rule into a computable rule intermediate representation, and generating a constraint set, wherein the constraint set is used for carrying out consistency check on scoring results; s5, performing AI policy reasoning based on the bidding feature vector, the evidence segment set and bidder ring risk prompt information to generate a scoring candidate set, performing policy optimization on the scoring candidate set to optimize a preset objective function on the premise of meeting a constraint set, and outputting an optimal scoring scheme and evidence segment references corresponding to each scoring item; s6, executing rule consistency check on the optimal scoring scheme, outputting conflict positioning information and triggering candidate rollback or local reasoner if conflict exists, and outputting a final auxiliary scoring result. The method comprises the steps of S1, obtaining multiple sets of bidding data corresponding to the same evaluation item, wherein the bidding data at least comprise a bidding document text, a quotation list data, qualification and performance proving data and bidding submitting metadata, and establishing a bidding data set according to a bidding party, performing format analysis, field standardization and missing completion on the bidding data set to obtain structured bidding data, forming a unified field view which can be used for subsequent feature extraction, dividing the bidding document text, the quotation list data, the qualifi