Construction scheme comparison and examination method and system based on retrieval enhancement generation technology

Abstract

The invention provides a construction scheme comparison and examination method based on a retrieval enhancement generation technology. The construction scheme comparison and examination method comprises the steps that S1, a standard specification named entity database and a standard specification vector database are created and enriched; s2, according to the construction scheme of the to-be-examined document format, analyzing, preprocessing data, segmenting into construction scheme short sentence texts, and outputting a construction scheme short sentence text set; s3, processing the construction scheme short sentence text set, and finally generating lt; constructing a scheme short sentence text, a named entity, a semantic vector, a newest standard specification and a named entity gt corresponding to the newest standard specification; a quintuple array; s4, judging lt; whether the short sentence text of the construction scheme, the named entity, the semantic vector, the latest standard specification and the named entity corresponding to the latest standard specification conform to gt or not is judged; each record in the six-element array set is converted into a corresponding lt; whether the short sentence text of the construction scheme, the named entities, the newest standard specification and the named entities corresponding to the newest standard specification conform to gt or not is judged; and carrying out five-element array and summarizing and outputting.

Description

Technical Field

[0001] The present disclosure relates to the fields of construction engineering informatization and intelligence, natural language processing, retrieval enhancement generation, etc. , and more particularly, to a construction plan comparison and review method and system based on retrieval enhanced generation technology. Background Art

[0002] The construction plan plays a very important role in the construction process of the building project. The review of the construction plan is a very important link, which is related to the quality, safety and efficiency of the entire construction project. The review of the construction plan helps to ensure that each link in the construction process meets the design requirements and construction standards, thereby ensuring the quality of construction. In the process of reviewing the construction plan, problems in the construction plan can be discovered and solved, errors in the construction plan can be reduced, and quality risks can be eliminated in the bud. Through the review of the construction plan, the safety measures and preventive measures in the construction plan can be evaluated, potential risks and dangers can be identified, and improvement suggestions and measures can be put forward to ensure the construction safety of the project. Through the review of the construction plan, it can ensure that the design intent of the project is effectively guaranteed, thereby ensuring the smooth progress of the entire project, shortening the construction time, and improving the construction efficiency.

[0003] However, current construction plan reviews rely on a large number of management documents such as construction industry standards and specifications. Traditional construction plan reviews are conducted manually, which not only consumes a lot of time and human resources, but also may result in omissions due to human factors, thus affecting the efficiency of the review work.

[0004] Retrieval-augmented Generation (RAG) is a technology that combines information retrieval and text generation. It retrieves relevant information from large data sets and uses this information to assist large language models (LLMs) in generating more accurate and relevant text content. The core of RAG technology is that it can provide LLMs with external knowledge sources, thereby making up for the model's lack of knowledge updating and deep understanding in specific fields.

[0005] The advantages of RAG technology are mainly reflected in the following aspects: (A) External knowledge utilization: The RAG model can effectively utilize external knowledge bases and reference a large amount of information to provide more in-depth, accurate, and valuable answers, thereby improving the reliability of generated text.

[0006] (B) Timely data update: The RAG model has a mechanism for updating the retrieval library, which enables instant updating of knowledge without retraining the model, thereby providing answers related to the latest information.

[0007] (C) Responses are explainable: Since the answers of the RAG model come directly from the retrieval library, its responses are highly explainable, reducing the illusion of large models, and users can verify the accuracy of the answers.

[0008] (D) Highly customizable: The RAG model can be customized based on the knowledge base and prompts in a specific field, enabling it to quickly acquire the capabilities of that field and be widely applicable to different fields and applications.

[0009] (E) Security and privacy management: The RAG model can implement security control by limiting the permissions of the knowledge base, ensuring that sensitive information is not leaked and improving data security.

[0010] (F) Improved response quality: RAG technology improves the comprehensiveness and accuracy of answers by integrating retrieved information during the generation process, and the generated responses are more accurate and rich.

[0011] Currently, there is no technical solution to apply RAG technology to the review of building construction plans. Summary of the invention

[0012] In response to the above-mentioned technical problems in the related technology, the present disclosure proposes a construction plan comparison and review method based on retrieval enhancement generation technology, which can overcome the above-mentioned shortcomings of the prior art.

[0013] To achieve the above technical objectives, the technical solution of the present disclosure is implemented as follows: The first object of the present disclosure is to provide a construction plan comparison and review method based on retrieval enhancement generation technology, which comprises the following steps: S1: Create and enrich the standard specification named entity database and standard specification vector database; S1.1: Collect text data of standard documents in the field of construction, and create a standard specification named entity database and a standard specification vector database; S1.2: Perform data preprocessing on the text data of the specification file; S1.3: Using the sentence segmentation function provided by the natural language processing library, the text data of the standard file is segmented into a number of standard standard short sentence texts, all the standard standard short sentence texts are merged into a standard standard short sentence text set, and then the standard standard short sentence text set is output; S1.4: First, for each standard normative short sentence text in the standard normative short sentence text set, a named entity recognition algorithm is used to identify the named entities therein, and a corresponding <standard normative short sentence text, named entity> binary array is generated respectively; as for the standard normative short sentence text without the named entity identified, it is indicated that it has no value and is discarded, ensuring that each generated <standard normative short sentence text, named entity> binary array contains the named entity; Then, all <standard specification short sentence text, named entity> binary arrays are stored in the standard specification named entity database; S1.5: First, the language model is used to convert all standard normative short sentence texts in the binary array of <standard normative short sentence text, named entity> into semantic vectors, and the corresponding binary arrays of <standard normative short sentence text, semantic vector> are generated respectively; Then all the <standard specification short sentence text, semantic vector> binary arrays are stored in the standard specification vector database, and each <standard specification short sentence text, semantic vector> is used as a standard specification for comparative review; S2: According to the construction plan in the document format to be reviewed, the data is parsed, pre-processed, and cut into a number of construction plan short sentence texts, and then a set of construction plan short sentence texts is output; S2.1: Select the construction plan to be reviewed. The construction plan is in document format and parse the text content of the construction plan; S2.2: Perform data preprocessing on the text content of the construction plan; S2.3: Using the sentence segmentation function provided by the natural language processing library, the text content of the construction plan is segmented into a number of construction plan short sentence texts, all the construction plan short sentence texts are merged into a construction plan short sentence text set, and then the construction plan short sentence text set is output; S3: Process the short sentence text set of the construction plan, and finally generate a five-element array of <short sentence text of the construction plan, named entity, semantic vector, latest standard specification, and named entity corresponding to the latest standard specification>; S3.1: First, for each construction plan short sentence text in the construction plan short sentence text set, a named entity recognition algorithm is used to identify the named entities therein, and a corresponding <construction plan short sentence text, named entity> binary array is generated respectively; as for the construction plan short sentence text without named entity recognition, it is indicated that it has no value and is discarded, ensuring that each generated <construction plan short sentence text, named entity> binary array contains a named entity; Then, it is summarized and output as a binary array set of <construction plan short sentence text, named entity>; S3.2: First, the language model is used to convert each construction plan short sentence text in the binary array set of <construction plan short sentence text, named entity> into a semantic vector, and a corresponding <construction plan short sentence text, named entity, semantic vector> ternary array is generated respectively; Then, the data is summarized and output as a set of three-element arrays of <construction plan short sentence text, named entity, semantic vector>; S3.3: First, for the <construction plan short sentence text, named entity, semantic vector> ternary array, based on the semantic vector of each construction plan short sentence text, retrieve the K standard specifications with the closest semantics from the standard specification vector database to generate a <construction plan short sentence text, named entity, semantic vector, K standard specifications> quaternary array; if the semantic vector of a construction plan short sentence text cannot retrieve a standard specification, it means that it has no similar standard specification, and the ternary array data is discarded; where K>=1; Then, the data is summarized and output as a four-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications>; S3.4: for all four-element arrays of <construction plan short sentence text, named entity, semantic vector, K standard specifications>, based on the respective K standard specifications, the corresponding named entity is queried from the standard specification named entity database, thereby generating the corresponding five-element array of <construction plan short sentence text, named entity, semantic vector, K standard specifications, named entity corresponding to the K standard specifications>, and summarizing and outputting it as a five-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications, named entity corresponding to the K standard specifications>; S3.5: First, input a five-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications, and named entities corresponding to the K standard specifications>, select the latest published standard specifications from the K standard specifications of each record, and select the named entity of the latest standard specification from the named entities corresponding to the K standard specifications of each record, thereby generating a corresponding five-element array of <construction plan short sentence text, named entity, semantic vector, latest standard specification, and named entity corresponding to the latest standard specification>; Then, for all the five-element arrays of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification>, compare the named entity and named entity of the latest standard specification in each record respectively. If the two are the same, record "yes" - indicating compliance with the standard specification, otherwise record "no" - indicating non-compliance with the standard specification, and generate the corresponding <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> respectively; Finally, output the six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>; S4: convert each record in the six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> into a corresponding five-element array of <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>, and finally summarize and output; S4.1: Input a six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>, and for each record therein, convert it into a corresponding five-element array of <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>; S4.2: All the <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> five-element arrays generated by S4.1 are aggregated into the final <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> five-element array set and output. The final <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> five-element array set can not only judge whether the construction plan complies with the standard specification, but also gives the content of the latest standard specification, which has a certain degree of explanatory power.

[0014] Preferably, in S1.1, the regulatory documents include laws, regulations, standards, specifications and technical manuals in the field of construction.

[0015] Preferably, in S1.3, the text data of the specification document is segmented according to punctuation marks.

[0016] Preferably, the punctuation mark is a period, but it may also be other specific symbols.

[0017] Preferably, data preprocessing includes eliminating noise, unifying professional terminology, performing word segmentation and screening out common stop words, so as to improve the quality and practicality of the processed data.

[0018] Preferably, the value range of K is 3--10, and can also be set to other values ​​according to actual needs.

[0019] Preferably, the language model is based on the Transformer architecture.

[0020] The second object of the present disclosure is to provide a system for a construction plan comparison and review method, including a user interface, a construction plan comparison and review module, a standard specification named entity database, a standard specification vector database, and a language model; The user interface is connected to the construction plan comparison and review module, the standard specification named entity database, the standard specification vector database, and the language model are respectively connected to the construction plan comparison and review module, and the language model is connected to the standard specification vector database; The user interface is used for the user to input operations into the construction plan comparison and review module; The construction plan comparison and review module uniformly performs various operations, including data preprocessing, data segmentation, merging, identifying named entities, discarding worthless data, generating arrays, converting semantic vectors, and searching; The standard specification named entity database, standard specification vector database, and language model are respectively used for calling by the construction plan comparison and review module, so that the construction plan comparison and review module can perform specific operations.

[0021] Preferably, an instruction to create a standard specification named entity database, a standard specification vector database, and a language model can be issued to the construction plan comparison and review module through a user interface.

[0022] Preferably, the user interface can be used to scan and input or directly input the specification file, so that it can be processed by the standard specification named entity database and then stored in the standard specification named entity database and the standard specification vector database.

[0023] Compared with the prior art, the present disclosure has the following beneficial effects: The present disclosure pioneered the application of RAG technology in the review of construction plans in the construction field, providing an efficient and advanced technical means. The application of RAG technology also helps to reduce negligence and omissions caused by human factors and improve the automation and intelligence level of the review process. In this way, the review process is optimized based on RAG technology, bringing innovative technical solutions to the construction field, thus having the following advantages: (I) It has pioneered the proposal of an intelligent construction plan comparison and review method and system, using cutting-edge artificial intelligence technology to automatically review the grouping and sub-item plans of construction projects, significantly promoting the automation and intelligent development of the construction industry.

[0024] (II) It innovatively integrates advanced search enhancement generation technology to create a complete technical architecture and intelligent processing system, which greatly reduces the time and labor intensity required for manual review of construction plans and significantly improves work efficiency and review quality.

[0025] (III) A comprehensive processing flow has been carefully designed, covering the core steps of building a standard specification vector database, intelligent processing of construction plan content, and comparative review of construction plans. The implementation of this process ensures high efficiency and high standards for intelligent review of construction plans, significantly improves the efficiency and quality of review work, and brings revolutionary improvements to the construction industry.

[0026] (IV) A detailed comparative review of the construction plan was conducted in accordance with the building standards and specifications, ensuring the professionalism, standardization and interpretability of the intelligent review process. This innovative method significantly improves the accuracy and reliability of the construction plan review and provides strong technical support for the implementation of construction industry standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 It is a flow chart of the construction plan comparison and review method based on retrieval enhancement generation technology described in the present invention.

[0029] Figure 2 It is a logic diagram of the construction plan comparison and review method based on retrieval enhancement generation technology described in the present invention.

[0030] Figure 3 It is a structural logic diagram of the construction plan comparison and review system based on retrieval enhancement generation technology described in the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present disclosure.

[0032] like Figure 1-3 As shown, in order to facilitate the understanding of the above technical solutions of the present disclosure, the above technical solutions of the present disclosure are described in detail below through specific usage methods.

[0033] The first object of the present disclosure is to provide a construction plan comparison and review method based on retrieval enhancement generation technology, which comprises the following steps: S1: Create and enrich the standard specification named entity database and standard specification vector database; S1.1: Collect text data of standard documents in the field of construction, and create a standard specification named entity database and a standard specification vector database; S1.2: Perform data preprocessing on the text data of the specification file; S1.3: Using the sentence segmentation function provided by the natural language processing library, the text data of the standard file is segmented into a number of standard standard short sentence texts, all the standard standard short sentence texts are merged into a standard standard short sentence text set, and then the standard standard short sentence text set is output; S1.4: First, for each standard normative short sentence text in the standard normative short sentence text set, a named entity recognition algorithm is used to identify the named entities therein, and a corresponding <standard normative short sentence text, named entity> binary array is generated respectively; as for the standard normative short sentence text without the named entity identified, it is indicated that it has no value and is discarded, ensuring that each generated <standard normative short sentence text, named entity> binary array contains the named entity; Then, all <standard specification short sentence text, named entity> binary arrays are stored in the standard specification named entity database; S1.5: First, the language model is used to convert all standard normative short sentence texts in the binary array of <standard normative short sentence text, named entity> into semantic vectors, and the corresponding binary arrays of <standard normative short sentence text, semantic vector> are generated respectively; Then all the <standard specification short sentence text, semantic vector> binary arrays are stored in the standard specification vector database, and each <standard specification short sentence text, semantic vector> is used as a standard specification for comparative review; S2: According to the construction plan in the document format to be reviewed, the data is parsed, pre-processed, and cut into a number of construction plan short sentence texts, and then a set of construction plan short sentence texts is output; S2.1: Select the construction plan to be reviewed. The construction plan is in document format and parse the text content of the construction plan; S2.2: Perform data preprocessing on the text content of the construction plan; S2.3: Using the sentence segmentation function provided by the natural language processing library, the text content of the construction plan is segmented into a number of construction plan short sentence texts, all the construction plan short sentence texts are merged into a construction plan short sentence text set, and then the construction plan short sentence text set is output; S3: Process the short sentence text set of the construction plan, and finally generate a five-element array of <short sentence text of the construction plan, named entity, semantic vector, latest standard specification, and named entity corresponding to the latest standard specification>; S3.1: First, for each construction plan short sentence text in the construction plan short sentence text set, a named entity recognition algorithm is used to identify the named entities therein, and a corresponding <construction plan short sentence text, named entity> binary array is generated respectively; as for the construction plan short sentence text without named entity recognition, it is indicated that it has no value and is discarded, ensuring that each generated <construction plan short sentence text, named entity> binary array contains a named entity; Then, it is summarized and output as a binary array set of <construction plan short sentence text, named entity>; S3.2: First, the language model is used to convert each construction plan short sentence text in the binary array set of <construction plan short sentence text, named entity> into a semantic vector, and a corresponding <construction plan short sentence text, named entity, semantic vector> ternary array is generated respectively; Then, the data is summarized and output as a set of three-element arrays of <construction plan short sentence text, named entity, semantic vector>; S3.3: First, for the <construction plan short sentence text, named entity, semantic vector> ternary array, based on the semantic vector of each construction plan short sentence text, retrieve the K standard specifications with the closest semantics from the standard specification vector database to generate a <construction plan short sentence text, named entity, semantic vector, K standard specifications> quaternary array; if the semantic vector of a construction plan short sentence text cannot retrieve a standard specification, it means that it has no similar standard specification, and the ternary array data is discarded; where K>=1; Then, the data is summarized and output as a four-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications>; S3.4: for all four-element arrays of <construction plan short sentence text, named entity, semantic vector, K standard specifications>, based on the respective K standard specifications, the corresponding named entity is queried from the standard specification named entity database, thereby generating the corresponding five-element array of <construction plan short sentence text, named entity, semantic vector, K standard specifications, named entity corresponding to the K standard specifications>, and summarizing and outputting it as a five-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications, named entity corresponding to the K standard specifications>; S3.5: First, input a five-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications, and named entities corresponding to the K standard specifications>, select the latest published standard specifications from the K standard specifications of each record, and select the named entity of the latest standard specification from the named entities corresponding to the K standard specifications of each record, thereby generating a corresponding five-element array of <construction plan short sentence text, named entity, semantic vector, latest standard specification, and named entity corresponding to the latest standard specification>; Then, for all the five-element arrays of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification>, compare the named entity and named entity of the latest standard specification in each record respectively. If the two are the same, record "yes" - indicating compliance with the standard specification, otherwise record "no" - indicating non-compliance with the standard specification, and generate the corresponding <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> respectively; Finally, output the six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>; S4: convert each record in the six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> into a corresponding five-element array of <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>, and finally summarize and output; S4.1: Input a six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>, and for each record therein, convert it into a corresponding five-element array of <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>; S4.2: All the <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> five-element arrays generated by S4.1 are aggregated into the final <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> five-element array set and output. The final <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> five-element array set can not only judge whether the construction plan complies with the standard specification, but also gives the content of the latest standard specification, which has a certain degree of explanatory power.

[0034] In one embodiment, in S1.1, the regulatory documents include laws, regulations, standards, specifications, and technical manuals in the field of construction.

[0035] In one embodiment, in S1.3, the text data of the specification document is segmented according to punctuation marks.

[0036] In one embodiment, the punctuation mark is preferably a period, but it may also be other specific symbols.

[0037] In one embodiment, data preprocessing includes eliminating noise, unifying professional terminology, performing word segmentation, and filtering out common stop words, thereby improving the quality and practicality of the processed data.

[0038] In a certain embodiment, the value range of K is 3--10, and can also be set to other values ​​according to actual needs.

[0039] In one embodiment, the language model is based on the Transformer architecture.

[0040] The second object of the present disclosure is to provide a system for a construction plan comparison and review method, including a user interface, a construction plan comparison and review module, a standard specification named entity database, a standard specification vector database, and a language model; The user interface is connected to the construction plan comparison and review module, the standard specification named entity database, the standard specification vector database, and the language model are respectively connected to the construction plan comparison and review module, and the language model is connected to the standard specification vector database; The user interface is used for the user to input operations into the construction plan comparison and review module; The construction plan comparison and review module uniformly performs various operations, including data preprocessing, data segmentation, merging, identifying named entities, discarding worthless data, generating arrays, converting semantic vectors, and searching; The standard specification named entity database, standard specification vector database, and language model are respectively used for calling by the construction plan comparison and review module, so that the construction plan comparison and review module can perform specific operations.

[0041] In one embodiment, an instruction to create a standard specification named entity database, a standard specification vector database, and a language model can be issued to the construction plan comparison and review module through a user interface.

[0042] In one embodiment, the user interface can be used to scan or directly input the specification file, so that the standard specification named entity database can process it and store it in the standard specification named entity database and the standard specification vector database.

[0043] In summary, through the unique design of the present disclosure, the present disclosure pioneered the application of RAG technology in the review of construction plans in the construction field, providing an efficient and advanced technical means. The application of RAG technology also helps to reduce negligence and omissions caused by human factors and improve the automation and intelligence level of the review process. In this way, the review process is optimized based on RAG technology, bringing innovative technical solutions to the construction field.

[0044] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A construction plan comparison and review method based on retrieval enhancement generation technology, characterized in that: The following steps are involved: S1: Create and enrich the standard specification named entity database and standard specification vector database; S1.1: Collect text data of standard documents in the field of construction, and create the standard specification named entity database and standard specification vector database; S1.2: performing data preprocessing on the text data of the specification file; S1.3: using the sentence segmentation function provided by the natural language processing library, the text data of the standard file is segmented into a plurality of standard standard short sentence texts, all the standard standard short sentence texts are merged into a standard standard short sentence text set, and then the standard standard short sentence text set is output; S1.4: First, for each of the standard canonical short sentence texts in the standard canonical short sentence text set, a named entity recognition algorithm is used to identify the named entities therein, and a corresponding <standard canonical short sentence text, named entity> binary array is generated respectively; as for the standard canonical short sentence text for which the named entity is not identified, it is indicated that it has no value and is discarded, ensuring that each generated <standard canonical short sentence text, named entity> binary array contains a named entity; Then, all <standard specification short sentence text, named entity> binary arrays are stored in the standard specification named entity database; S1.5: First, a language model is used to convert all the standard normative short sentence texts in the binary array of <standard normative short sentence text, named entity> into semantic vectors, and a corresponding binary array of <standard normative short sentence text, semantic vector> is generated respectively; Then all the <standard specification short sentence text, semantic vector> binary arrays are stored in the standard specification vector database, and each <standard specification short sentence text, semantic vector> is used as a standard specification for comparison and review; S2: According to the construction plan in the document format to be reviewed, the data is parsed, pre-processed, and cut into a number of construction plan short sentence texts, and then a set of construction plan short sentence texts is output; S2.1: Select a construction plan to be reviewed, the construction plan is in a document format, and parse the text content of the construction plan; S2.2: performing data preprocessing on the text content of the construction plan; S2.3: using the sentence segmentation function provided by the natural language processing library, the text content of the construction plan is segmented into a plurality of construction plan short sentence texts, all the construction plan short sentence texts are merged into the construction plan short sentence text set, and then the construction plan short sentence text set is output; S3: Process the construction plan short sentence text set, and finally generate a five-element array of <construction plan short sentence text, named entity, semantic vector, latest standard specification, and named entity corresponding to the latest standard specification>; S3.1: First, for each of the construction plan short sentence texts in the construction plan short sentence text set, a named entity recognition algorithm is used to identify the named entities therein, and a corresponding <construction plan short sentence text, named entity> binary array is generated respectively; as for the construction plan short sentence text for which the named entity is not identified, it is indicated that it has no value and is discarded, ensuring that each generated <construction plan short sentence text, named entity> binary array contains a named entity; Then, it is summarized and output as a binary array set of <construction plan short sentence text, named entity>; S3.2: First, the language model is used to convert each construction plan short sentence text in the binary array set of <construction plan short sentence text, named entity> into a semantic vector, and a corresponding <construction plan short sentence text, named entity, semantic vector> ternary array is generated respectively; Then, the data is summarized and output as a set of three-element arrays of <construction plan short sentence text, named entity, semantic vector>; S3.3: First, for the <construction plan short sentence text, named entity, semantic vector> ternary array, based on the semantic vector of each construction plan short sentence text, the K standard specifications with the closest semantics are retrieved from the standard specification vector database to generate a <construction plan short sentence text, named entity, semantic vector, K standard specifications> quaternary array; if the semantic vector of a construction plan short sentence text cannot retrieve a standard specification, it means that it has no similar standard specification, and the ternary array data is discarded; wherein, K>=1; Then, the data is summarized and output as a four-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications>; S3.4: for all four-element arrays of <construction plan short sentence text, named entity, semantic vector, K standard specifications>, based on the respective K standard specifications, the corresponding named entity is queried from the standard specification named entity database, thereby generating the corresponding five-element array of <construction plan short sentence text, named entity, semantic vector, K standard specifications, named entity corresponding to the K standard specifications>, and summarizing and outputting it as a five-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications, named entity corresponding to the K standard specifications>; S3.5: First, input a five-element array set of <construction plan short sentence text, named entity, semantic vector, K standard specifications, and named entities corresponding to the K standard specifications>, select the latest published standard specifications from the K standard specifications of each record, and select the named entity of the latest standard specification from the named entities corresponding to the K standard specifications of each record, thereby generating a corresponding five-element array of <construction plan short sentence text, named entity, semantic vector, latest standard specification, and named entity corresponding to the latest standard specification>; Then, for all the five-element arrays of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification>, compare the named entity and named entity of the latest standard specification in each record respectively. If the two are the same, record "yes" - indicating compliance with the standard specification, otherwise record "no" - indicating non-compliance with the standard specification, and generate the corresponding <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> respectively; Finally, output the six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>; S4: convert each record in the six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies> into a corresponding five-element array of <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>, and finally summarize and output; S4.1: Input a six-element array set of <construction plan short sentence text, named entity, semantic vector, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>, and for each record therein, convert it into a corresponding five-element array of <construction plan short sentence text, named entity, latest standard specification, named entity corresponding to the latest standard specification, whether it complies>; S4.2: All the five-element arrays of <construction plan short sentence text, named entities, latest standards and specifications, named entities corresponding to the latest standards and specifications, and whether they comply> generated by S4.1 are aggregated into the final five-element array set of <construction plan short sentence text, named entities, latest standards and specifications, named entities corresponding to the latest standards and specifications, and whether they comply> and output them.

2. The construction plan comparison and review method according to claim 1, characterized in that: In S1.1, the regulatory documents include laws, regulations, standards, specifications, and technical manuals in the field of construction.

3. The construction plan comparison and review method according to claim 1, characterized in that: In S1.3, the text data of the specification document is segmented according to punctuation marks.

4. The construction plan comparison and review method according to claim 3 is characterized in that: The punctuation mark is a period.

5. The construction plan comparison and review method according to claim 1, characterized in that: The data preprocessing includes eliminating noise, unifying professional terminology, performing word segmentation and screening out common stop words, thereby improving the quality and practicality of the processed data.

6. The construction plan comparison and review method according to claim 1 is characterized in that: The value range of K is 3--10.

7. The construction plan comparison and review method according to claim 1, characterized in that: The language model is based on the Transformer architecture.

8. A system for the construction plan comparison and review method according to any one of claims 1 to 7, characterized in that: It includes user interface, construction plan comparison and review module, standard specification named entity database, standard specification vector database, and language model; The user interface is connected to the construction plan comparison and review module, the standard specification named entity database, the standard specification vector database, and the language model are respectively connected to the construction plan comparison and review module, and the language model is connected to the standard specification vector database; The user interface is used for the user to input operations to the construction plan comparison and review module; The construction plan comparison and review module uniformly performs various operations, including data preprocessing, data segmentation, merging, identifying various named entities, discarding worthless data, generating various arrays, converting semantic vectors, and searching; The standard specification named entity database, standard specification vector database, and language model are respectively used for calling by the construction plan comparison and review module, so that the construction plan comparison and review module can perform specific operations.

9. The system according to claim 8, characterized in that Through the user interface, an instruction to create a standard specification named entity database, a standard specification vector database, and a language model is sent to the construction plan comparison and review module.

10. The system according to claim 8, characterized in that The specification file is scanned and inputted or directly inputted through the user interface so that the specification file is processed by the standard specification named entity database and stored in the standard specification named entity database and the standard specification vector database.

Images