CN-121996768-A - Intelligent generation method, system and medium for building decision

CN121996768ACN 121996768 ACN121996768 ACN 121996768ACN-121996768-A

Abstract

The embodiment of the invention provides an intelligent generation method, system and medium for building decisions, and relates to the field of building engineering. The method comprises the steps of generating corresponding construction context vectors based on problem description information input by a user, driving at least two intelligent agents to execute countermeasure search in parallel to correspondingly obtain at least two evidence sets, carrying out credibility quantification on the at least two evidence sets, obtaining credibility evaluation values of each evidence, controlling the at least two intelligent agents to carry out multi-round interactive dialectical analysis to dynamically update effectiveness states and credibility evaluation values of each evidence in a multi-round mode, generating a judging result through a preset judging model, and generating decision guiding information suitable for a user angle based on the judging result and outputting the decision guiding information. The system is configured to perform the method. The method may also be run after being read by program instructions stored on a computer readable storage medium. The invention can more intelligently solve the technical dispute problem and complete the intelligent building decision.

Inventors

LIU JUNJIE
HE JIAN
ZHANG YINLIN

Assignees

一智科技(成都)有限公司

Dates

Publication Date: 20260508
Application Date: 20260331

Claims (10)

1. An intelligent building decision generation method is characterized by comprising the following steps: Generating a corresponding construction context vector based on the problem description information input by the user; Driving at least two agents in parallel to perform a countercheck based on the construction context vector to correspondingly obtain at least two evidence sets, the at least two agents configured to have dissimilar search preferences; Carrying out credibility quantification on the at least two evidence sets, and acquiring credibility evaluation values of each evidence in the at least two evidence sets; controlling the at least two agents to conduct multi-round interactive dialects based on the credibility evaluation value, and dynamically updating the efficacy state of each evidence and the credibility evaluation value in a plurality of rounds; generating a decision result by a preset decision model based on each evidence and the corresponding updated efficacy state and updated credibility evaluation value, and And generating and outputting decision guiding information suitable for the role of the user based on the judging result.
2. The method of claim 1, wherein generating the corresponding construction context vector based on the user-entered problem description information comprises: based on the problem description information, carrying out structural analysis on the construction scene to obtain scene elements, and The construction context vector is generated based on the scene element.
3. The method of claim 1, wherein the driving at least two agents in parallel to perform an antagonistic search comprises: configuring a corresponding retrieval bias strategy for each of the at least two agents, such that each agent tends to retrieve evidence conforming to respective retrieval preferences; Wherein the dissimilar retrieval preferences include at least two of evidence that tends to retrieve the most stringent compliance requirements, evidence that tends to retrieve the most convenient to construct, and evidence that tends to retrieve the most cost effective.
4. The method of claim 1, wherein the performing the confidence quantization on the at least two evidence sets to obtain a confidence score for each evidence in the at least two evidence sets comprises: performing metadata structured extraction on the at least two evidence sets to obtain structured metadata of each evidence, and And acquiring the credibility evaluation value based on the structured metadata.
5. The method of claim 1, wherein the controlling the at least two agents to perform multiple rounds of interactive dialects based on the confidence score to dynamically update the efficacy state of each evidence and the confidence score in multiple rounds comprises: Initializing the efficacy state of each of the evidence to a valid state, and In response to the at least two agents asking in the multi-round interactive dialect, updating the efficacy state and the confidence rating value for each evidence based on state transition rules, the efficacy state including the valid state, pending state, and invalid state.
6. The method of claim 1, wherein generating the arbitration result by the preset arbitration model based on the each evidence and the corresponding updated efficacy status and updated confidence score value comprises: Responsive to entering the arbitration phase, determining a washed evidence pair based on the each evidence and the corresponding updated efficacy state; Identifying semantic force coefficients of conflicting ones of the pair of cleaned evidence based on a semantic disambiguation engine, and And calculating the semantic decision value of the conflict evidence based on the updated credibility evaluation value of the conflict evidence and the semantic compulsory coefficient.
7. The method of claim 6, wherein generating a arbitration result by a preset arbitration model based on the each evidence and the corresponding updated efficacy status and updated confidence score value, further comprises: Acquiring efficacy coverage relation and graph path coefficient of the conflict evidence based on efficacy knowledge graph, and A weighted arbitration value for the conflict evidence is determined based on the efficacy coverage relationship, the graph path coefficients, and the semantic arbitration value.
8. The method of claim 1, wherein generating decision guide information that is compatible with a user role based on the arbitration result comprises: Identifying the user role; selecting a corresponding speaking template from a preset speaking template library based on the user role, and And filling the judging result into the conversation template to generate the decision guiding information.
9. An intelligent building decision generation system, comprising: the problem analysis module is configured to generate a corresponding construction context vector based on the problem description information input by the user; A countercheck retrieval module configured to drive at least two agents in parallel to perform a countercheck retrieval based on the construction context vector to correspondingly obtain at least two evidence sets, the at least two agents configured to have dissimilar retrieval preferences; the credibility quantifying module is configured to conduct credibility quantification on the at least two evidence sets and acquire a credibility evaluation value of each evidence in the at least two evidence sets; an interactive debate module configured to control the at least two agents to conduct multiple rounds of interactive debate based on the confidence score to dynamically update the efficacy status of each evidence and the confidence score in multiple rounds; a arbitration module configured to generate an arbitration result by a preset arbitration model based on the each evidence and the corresponding updated efficacy status and updated credibility evaluation value, and And the decision module is configured to generate and output decision guiding information suitable for the role of the user based on the decision result.
10. A computer readable storage medium, characterized in that program instructions are stored, which program instructions, when executed by a processor, implement the intelligent generation method of a building decision according to claim 1.

Description

Intelligent generation method, system and medium for building decision Technical Field The invention relates to the field of constructional engineering, in particular to an intelligent generation method, system and medium for a building decision. Background In the field of building engineering, conditions of content intersection and even clause conflict often exist among construction specifications, design drawings, industry standards and local standards. Meanwhile, the applicability of the same clause in different specific scenes is also often controversial. At present, solving the technical disputes mainly relies on manual turning over of a large number of paper or electronic version specifications, and negotiation decision is carried out empirically. And due to the ubiquitous existence of technical disputes, negotiation decisions can lead to delays in the construction period for a period of time. In addition, the question-answering system based on the search enhancement generation (RETRIEVAL-Augmented Generation, RAG) is a main intelligent means, and the working mode of the question-answering system is that a user asks questions, the system searches the most relevant document, and a large model generates a unique answer. However, the question-answering system has the problems of single view angle deviation, lack of efficacy judgment, isolated evidence chain, lack of scene suitability and the like. Therefore, it is necessary to provide an intelligent generation method, system and medium for building decisions, which can more intelligently solve the technical dispute problem and complete the intelligent building decisions. Disclosure of Invention In order to solve the problem of how to intelligently generate a role-adaptive building decision, the invention provides an intelligent building decision generation method, system and medium. The invention comprises an intelligent generation method of building decisions. The method comprises the steps of generating corresponding construction context vectors based on problem description information input by a user, driving at least two agents to execute countermeasure search in parallel based on the construction context vectors so as to correspondingly obtain at least two evidence sets, wherein the at least two agents are configured to have different search preferences, carrying out credibility quantification on the at least two evidence sets, obtaining credibility evaluation values of each evidence in the at least two evidence sets, controlling the at least two agents to carry out multi-round interactive dialectical based on the credibility evaluation values, dynamically updating efficacy states of each evidence and the credibility evaluation values in multiple rounds, generating a decision result through a preset decision model based on each evidence and the corresponding updated efficacy states and the updated credibility evaluation values, and generating decision guide information adaptive to a user angle based on the decision result and outputting the decision guide information. The invention comprises an intelligent generation system for building decisions. The system comprises a problem analysis module, an countermeasure retrieval module, a decision module, a reliability quantification module and an interactive decision module, wherein the problem analysis module is configured to generate corresponding construction context vectors based on problem description information input by a user, the countermeasure retrieval module is configured to drive at least two agents to execute countermeasure retrieval based on the construction context vectors in parallel to correspondingly acquire at least two evidence sets, the at least two agents are configured to have different retrieval preference, the reliability quantification module is configured to conduct reliability quantification on the at least two evidence sets to acquire a reliability evaluation value of each evidence in the at least two evidence sets, the interactive decision module is configured to conduct multi-round interactive dialections based on the reliability evaluation value, the effectiveness state of each evidence and the reliability evaluation value are dynamically updated in a plurality of rounds, the decision module is configured to generate a decision result based on the each evidence and the corresponding updated effectiveness state and the updated reliability evaluation value through a preset decision angle model, and the interactive dialections module is configured to generate and output adaptive hue information based on the decision result and the user. The inventive content comprises a computer readable storage medium storing program instructions which, when executed by a processor, implement the intelligent generation method of building decisions as described above. The beneficial effects brought by the invention include, but are not limited to, (1) the shortcomings of single-sided and sing