A method for generating digital models of architectural intelligent design specification clauses

Abstract

The present invention relates to the technical field of building model construction, and discloses a method for generating a digital model of building intelligent design specification clauses, comprising: extracting an original clause set from a pre-constructed specification clause database according to a generation instruction, wherein each original clause in the original clause set is in text form, obtaining a structural clause of each original clause in text form, wherein the structural clause is in the form of a quadruple and consists of an entity clause and a true confidence value, obtaining a standard clause original model, inputting the original clause set in text form into the standard clause original model to perform confidence determination, obtaining a confidence determination value, adjusting the model parameters according to the difference between the true confidence value and the confidence determination value, until the difference between the true confidence value and the confidence determination value is less than a pre-set threshold value, obtaining a standard clause model. The main purpose of the present invention is to ensure that the standard clause model is efficiently generated under accurate confidence determination.

Description

Technical Field

[0001] The invention relates to a method for generating a digital model of building intelligent design specification clauses, and belongs to the technical field of building model construction. Background Art

[0002] Architectural design involves numerous complex regulations, and traditional manual processing is inefficient and error-prone. Digitizing these regulations enables automated parsing, storage, and application, improving design efficiency and ensuring compliance with regulatory requirements. Digital models convert textual regulations into structured data, facilitating computer processing and analysis, providing reliable technical support for intelligent design.

[0003] However, current technologies for generating digital models of regulatory texts have drawbacks. Existing models often struggle to accurately capture the semantics and structure of texts, leading to inaccurate confidence levels and impacting the reliability of model outputs. Furthermore, there is a lack of systematic methods for adjusting and optimizing model parameters. Therefore, ensuring efficient generation of digital models while ensuring accurate confidence levels is an urgent technical challenge. Summary of the Invention

[0004] The present invention provides a method, device and computer-readable storage medium for generating a digital model of building intelligent design specification clauses, the main purpose of which is to ensure accurate confidence judgment and efficiently generate the specification clause model.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for generating a digital model of building intelligent design specification clauses, comprising:

[0006] receiving a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and pulling an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form;

[0007] Obtain the structure clause of each original clause in text form, wherein the structure clause is in the form of a four-tuple and consists of an entity clause and a true confidence value;

[0008] Obtain the original model of the regulatory provisions, where the original model of the regulatory provisions includes, in order from data input to data output, a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer;

[0009] Inputting the original clause set in text form into the original model of the standard clause to perform confidence determination and obtain a confidence determination value;

[0010] According to the difference between the true confidence value and the confidence judgment value, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted until the difference between the true confidence value and the confidence judgment value is less than a preset threshold. The model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the specification clause database, and the standard model of the specification clause is obtained at the same time.

[0011] Optionally, according to the generation instruction, an original clause set is pulled from a pre-built standard clause database, wherein each original clause in the original clause set is in text form and includes:

[0012] Parsing the generation instruction to obtain the building information of the digital model of the specification clause before generation, wherein the building information includes building type, construction time, building scale, building style and material;

[0013] Using the building information as a first search condition, searching the specification article database to obtain a first article collection;

[0014] Determine whether the size of the first set of articles is larger than the data size generated by the pre-set digital model of regulatory articles;

[0015] If the size of the first set of articles is not greater than the size of the data generated by the digital model of the standard articles, the construction time in the building information is removed to obtain the second search condition, and the second search condition is used to perform another search in the standard article database to obtain the second set of articles;

[0016] Determine whether the size of the second set of articles is larger than the size of data generated by the digital model of regulatory articles;

[0017] If the size of the second set of articles is not greater than the data size generated by the digital model of the standard articles, a generation failure instruction of the digital model of the standard articles is directly obtained, and the generation failure instruction is sent to the port of the generation instruction of the digital model of the standard articles;

[0018] If the scale of the first or second article collection is larger than the data scale generated by the digital model of the standard articles, the first or second article collection shall be the original article collection.

[0019] Optionally, the structural clause is expressed as:

[0020] ,

[0021] in, Indicates the The structural provisions of the original provisions in the form of article texts, Indicates the The substantive provisions of the original article, Indicates the The header entity of the original article, Indicates the The end entity of the original article, Represents the entity relationship between the head entity and the tail entity, Indicates the The true value of the confidence of the entity clause corresponding to the original clause.

[0022] Optionally, the confidence determination layer includes a confidence calculation function and a confidence determination function, wherein the confidence determination function is:

[0023] ,

[0024] in, Represents the difference between the confidence judgment value and the true confidence value, Indicates the original article The original text in the form of the article is input into the BERT layer, LSTM layer, and GNN layer of the standard article original model to obtain the entity article. The textual difference between the substantive clauses of the original clauses, Indicates the The confidence value of the original article.

[0025] Optionally, the step of inputting the original set of clauses in text form into the original model of the standard clauses to perform confidence determination and obtain a confidence determination value includes:

[0026] Each original article in the original article set in text form is input into the standard article original model in sequence, and the confidence judgment value of each original article is calculated as follows:

[0027] Use the BERT layer to perform word encoding on the original text to obtain the original text in vectorized form;

[0028] Import the original clauses in vectorized form into the LSTM layer, and use LSTM to correct the original clauses in vectorized form to obtain the corrected original clauses;

[0029] Use the GNN layer to transform the corrected original clauses into structured original clauses;

[0030] The confidence level of the original text in structured form is calculated using the confidence level determination layer to obtain the confidence level determination value. .

[0031] Optionally, the calculating the confidence of the structured original clause using the confidence determination layer to obtain the confidence determination value includes:

[0032] The original structured clause is transformed into a single-dimensional vector to obtain a single-dimensional clause, wherein the representation of the single-dimensional clause is:

[0033] ,

[0034] in, Indicates the The single-dimensional provisions of the original article, Indicates the The first entity after the original article is transformed by the GNN layer, and the first entity is obtained by performing a single-dimensional vector transformation vector values, Indicates the The entity relationship of the original article after the GNN layer transformation, the first vector values, Indicates the The tail entity of the original article after the GNN layer transformation is performed on the single-dimensional vector vector values;

[0035] Use the confidence calculation function to calculate the confidence of the single-dimensional clause and obtain the confidence judgment value .

[0036] Optionally, the confidence calculation function is used to calculate the confidence of the single-dimensional clause to obtain the confidence judgment value , including the confidence value calculated according to the following formula :

[0037] ,

[0038] in, yes The weight value of yes The weight value of yes The weight value of .

[0039] Optionally, the The weight value of is negative. and The weight value of is a positive number.

[0040] Optionally, the method adjusts the model parameters of the BERT layer, the LSTM layer, and the GNN layer according to the difference between the true confidence value and the confidence judgment value until the difference between the true confidence value and the confidence judgment value is less than a preset threshold, stores the model parameters of the BERT layer, the LSTM layer, and the GNN layer in the regulatory clause database, and simultaneously obtains the regulatory clause standard model, including:

[0041] Using the confidence judgment function, the difference between the true confidence value and the confidence judgment value is calculated;

[0042] Determine the relationship between the difference between the true confidence value and the confidence determination value and a preset threshold;

[0043] When the difference between the true confidence value and the confidence judgment value is greater than a preset threshold, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted. The model parameters include the number of attention heads of the BERT layer, the number of hidden units and sequence length of the LSTM layer, and the number of stacking layers and aggregation function of the GNN layer.

[0044] When the difference between the true confidence value and the confidence judgment value is less than or equal to a preset threshold, the adjusted model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the standard article database, and when the model parameters of the BERT layer, LSTM layer, and GNN layer are successfully stored in the standard article database, the standard article standard model automatic generation program is started, wherein the standard article standard model automatic generation program is embedded with a standard article generation template;

[0045] The standard model of the regulatory clauses is automatically generated by using the program, and the model parameters of the stored BERT layer, LSTM layer, and GNN layer are extracted from the regulatory clause database, and the regulatory clause generation template is combined to obtain the standard model of the regulatory clauses.

[0046] To achieve the above-mentioned object, the present invention further provides a system for generating a digital model of building intelligent design specification clauses, characterized in that the system comprises:

[0047] A clause extraction module is configured to receive a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and extract an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form;

[0048] The structure clause acquisition module is used to obtain the structure clause of each original clause in text form, wherein the structure clause is in the form of a four-tuple and consists of an entity clause and a true confidence value;

[0049] The confidence determination module is used to obtain the original model of the regulatory provisions. The original model of the regulatory provisions includes a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer in the order from input to output of data. The original set of textual provisions is input into the original model of the regulatory provisions to perform confidence determination and obtain a confidence determination value.

[0050] The standard model completion module is used to adjust the model parameters of the BERT layer, LSTM layer, and GNN layer according to the difference between the true confidence value and the confidence judgment value, until the difference between the true confidence value and the confidence judgment value is less than a preset threshold. The model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the standard clause database, and the standard clause model is obtained at the same time.

[0051] In order to solve the above problem, the present invention further provides an electronic device, comprising:

[0052] at least one processor; and,

[0053] a memory communicatively connected to the at least one processor; wherein,

[0054] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to implement the above-mentioned method for generating a digital model of building intelligent design specification clauses.

[0055] In order to solve the above problems, the present invention also provides a computer-readable storage medium, which stores at least one instruction. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned method for generating a digital model of building intelligent design specification clauses.

[0056] Compared with the problems described in the background technology, the present invention clarifies the model structure of each layer. Specifically, the BERT layer, LSTM layer, and GNN layer have clear division of labor, and comprehensively analyze the clauses from the semantic, sequence, and structured information dimensions respectively to ensure the accuracy of confidence judgment. At the same time, the present invention drives parameter optimization through the difference between the true value of the confidence and the judgment value, and combines threshold control to achieve rapid convergence, taking into account both accuracy and efficiency. In addition, through clause database support, structured processing, automatic generation, and templated output, full automation from data input to model output is achieved, significantly improving generation efficiency. Therefore, the digital model generation method, system, electronic device, and computer-readable storage medium for building intelligent design specification clauses proposed in the present invention can ensure the efficient generation of specification clause models under accurate confidence judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of a process for generating a digital model of building intelligent design specification clauses provided in one embodiment of the present invention;

[0058] Figure 2 This is a functional module diagram of a digital model generation system for building intelligent design specification clauses provided by one embodiment of the present invention;

[0059] Figure 3A schematic diagram of the structure of an electronic device for implementing the method for generating a digital model of intelligent building design specification clauses provided in one embodiment of the present invention.

[0060] Description of reference numerals:

[0061] 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0062] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0063] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0064] The embodiments of the present application provide a method for generating a digital model of a building intelligent design specification. The execution entity of the method includes, but is not limited to, at least one of electronic devices such as a server or a terminal that can be configured to execute the method provided by the embodiments of the present application. In other words, the method for generating a digital model of a building intelligent design specification can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0065] Example 1:

[0066] Reference Figure 1 FIG. 1 is a flow chart of a method for generating a digital model of a building intelligent design specification clause provided by an embodiment of the present invention. In this embodiment, the method for generating a digital model of a building intelligent design specification clause includes:

[0067] S1. Receive a generation instruction for a digital model of standard clauses, wherein the digital model of standard clauses is applied to the field of architectural design; according to the generation instruction, pull out an original clause set from a pre-built standard clause database, wherein each original clause in the original clause set is in text form.

[0068] It should be explained that the generation instructions are instructions for implementing a digital model of regulatory provisions, and in the embodiments of the present invention, the digital model of regulatory provisions is the standard model of regulatory provisions. Specifically, the standard model of regulatory provisions is obtained by adjusting the model parameters of the BERT layer, LSTM layer, and GNN layer within the original model of the regulatory provisions. Furthermore, the architectural design field described in the embodiments of the present invention includes architectural planning, design, construction, and management, and covers functions such as functionality, aesthetics, structure, environment, and sustainability.

[0069] Generally speaking, instructions for generating digital models of regulatory provisions are typically issued by individuals or departments within the architectural design field, using a pre-built app or mini-program for generating digital models of regulatory provisions. For example, Xiao Zhang, the head of the architectural planning department, is currently overseeing the construction of a commercial district. To improve construction efficiency, Xiao Zhang wants to quickly generate a specific regulatory provision based on the digital model of the regulatory provisions. Therefore, Xiao Zhang needs to generate the corresponding digital model of the regulatory provisions, and thus initiates the generation instruction.

[0070] Furthermore, according to the generation instruction, an original clause set is pulled from a pre-built standard clause database, wherein each original clause in the original clause set is in text form, including:

[0071] Parsing the generation instruction to obtain the building information of the digital model of the specification clause before generation, wherein the building information includes building type, construction time, building scale, building style and material;

[0072] Using the building information as a first search condition, searching the specification article database to obtain a first article collection;

[0073] Determine whether the size of the first set of articles is larger than the data size generated by the pre-set digital model of regulatory articles;

[0074] If the size of the first set of articles is not greater than the size of the data generated by the digital model of the standard articles, the construction time in the building information is removed to obtain the second search condition, and the second search condition is used to perform another search in the standard article database to obtain the second set of articles;

[0075] Determine whether the size of the second set of articles is larger than the size of data generated by the digital model of regulatory articles;

[0076] If the size of the second set of articles is not greater than the data size generated by the digital model of the standard articles, a generation failure instruction of the digital model of the standard articles is directly obtained, and the generation failure instruction is sent to the port of the generation instruction of the digital model of the standard articles;

[0077] If the scale of the first or second article collection is larger than the data scale generated by the digital model of the standard articles, the first or second article collection shall be the original article collection.

[0078] Specifically, building type refers to its purpose, such as residential or commercial; construction date reflects historical context and technological advancements; and building scale relates to its size and capacity. Together, these factors determine a building's design, function, and value, and also determine the criteria for extracting the original set of clauses from the regulatory database. For example, Xiao Zhang is responsible for a commercial district whose building type is commercial, construction date is 2024, and capacity is at least 5,000 people. Therefore, using this building information as the first search condition, a search is performed in the regulatory database, yielding the first set of clauses.

[0079] It should be emphasized that the regulatory clause database has pre-collected historical regulatory clauses in the field of architectural design in previous time periods. Therefore, when Xiao Zhang uses the above-mentioned building information as the first search condition, he will retrieve the regulatory clauses that meet the requirements in the previous time period, which is the first set of clauses.

[0080] However, it should be noted that the digital models of the regulatory provisions described in the embodiments of the present invention are all deep learning models, and their generation process is obtained by adjusting the model parameters of the internal BERT layer, LSTM layer, and GNN layer of the original model of the regulatory provisions. The parameter adjustment of the BERT layer, LSTM layer, and GNN layer requires a large amount of data scale, and experimental analysis shows that the data scale is at least 500. That is, when the first set of articles contains less than 500 articles, it is necessary to eliminate the retrieval condition with a construction time of 2024 and perform a second search to meet the requirement that the size of the original set of articles retrieved is greater than 500.

[0081] S2. Obtain the structural clauses of each original clause in text form, wherein the structural clauses are all in the form of four-tuples and consist of entity clauses and true confidence values.

[0082] It should be explained that the main purpose of the structured clause is to display the original text in the form of a quad, thereby providing adjustment direction for the model parameters of the BERT layer, LSTM layer, and GNN layer. It should also be emphasized that there are many ways to generate structured clauses, including but not limited to using graph neural networks, large models, and other methods. In detail, the expression form of the structured clause is:

[0083] ,

[0084] in, Indicates the The structural provisions of the original provisions in the form of article texts, Indicates the The substantive provisions of the original article, Indicates the The header entity of the original article, Indicates the The end entity of the original article, Represents the entity relationship between the head entity and the tail entity, Indicates the The true value of the confidence of the entity clause corresponding to the original clause.

[0085] Furthermore, there are two ways to express entity relationships: ← and →. ← indicates that the tail entity depends on the head entity, and → indicates that the head entity depends on the tail entity. For example, the original text of Article 25 in the original set of articles reads: The load-bearing capacity of each floor of a three-story circular shopping mall shall not be less than 500 tons. It is recommended to adopt a reinforced concrete frame structure and use high-strength steel as the main material. The expression of its structural article is:

[0086] ,

[0087] It should be explained that the true value of confidence indicates the strength of the entity relationship between the head entity and the tail entity (also known as the probability value from a mathematical perspective), such as The true confidence value is , which means that when the building type is a three-story circular shopping mall, there is a 92% probability that it will need to rely on reinforced concrete frame structure and high-strength steel.

[0088] S3. Obtain the original model of the regulatory provisions, wherein the original model of the regulatory provisions includes a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer in the order from input to output of data.

[0089] It should be explained that the structure of the original model of the regulatory clauses described in the embodiment of the present invention has been pre-fixed. In the order from input to output of data, it includes a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer. The confidence determination layer includes a confidence calculation function and a confidence determination function. The confidence determination function is:

[0090] ,

[0091] in, Represents the difference between the confidence judgment value and the true confidence value, Indicates the original article The original text in the form of the article is input into the BERT layer, LSTM layer, and GNN layer of the standard article original model to obtain the entity article. The textual difference between the substantive clauses of the original clauses, Indicates the The confidence value of the original article. The entity clauses in the regulatory clause database are , if the original text of Article 25 is input into the original model of the standard article, the entity article obtained through the BERT layer, LSTM layer, and GNN layer is , then by comparing the text difference between the two entity clauses, we can calculate , and then calculate the difference between the confidence judgment value and the true confidence value.

[0092] In addition, the BERT (Bidirectional Encoder Representations from Transformers) layer is based on the Transformer architecture and captures the contextual information of the original text articles through a bidirectional attention mechanism, thereby achieving the purpose of word vector conversion; the LSTM (Long Short-Term Memory) layer is a special recurrent neural network (RNN) that processes sequence data through memory units and is good at capturing the long-term dependencies of the original text articles, thereby further optimizing the accuracy of word vectorization; the GNN (Graph Neural Network) layer is used to process graph-structured data and updates node representations by aggregating neighbor node information, thereby converting the original text articles into entity articles.

[0093] S4. Input the original set of clauses in text form into the original model of the standard clauses to perform confidence judgment and obtain a confidence judgment value.

[0094] It should be explained that the embodiment of the present invention requires that each original article in the original article set be input into the standard article original model to perform confidence determination and obtain a confidence determination value. For example, if the original article set contains 600 articles, then the 600 original articles need to be input into the standard article original model in sequence. In detail, the inputting of the original article set in text form into the standard article original model to perform confidence determination and obtain a confidence determination value includes:

[0095] Each original article in the original article set in text form is input into the standard article original model in sequence, and the confidence judgment value of each original article is calculated as follows:

[0096] Use the BERT layer to perform word encoding on the original text to obtain the original text in vectorized form;

[0097] Import the original clauses in vectorized form into the LSTM layer, and use LSTM to correct the original clauses in vectorized form to obtain the corrected original clauses;

[0098] Use the GNN layer to transform the corrected original clauses into structured original clauses;

[0099] The confidence level of the original text in structured form is calculated using the confidence level determination layer to obtain the confidence level determination value. .

[0100] It should be emphasized that how BERT and LSTM convert original textual articles into original articles in word vectorized form is a public technology. Similarly, using the GNN layer to convert the corrected original articles into original articles in structured form is also a public technology and will not be repeated in detail in the embodiments of the present invention.

[0101] In addition, the original clause of the structured form is expressed as follows: ,in, Indicates the The original text in structured form, Indicates the The head entity of the original article after being transformed by the GNN layer, Indicates the The tail entity of the original article after being transformed by the GNN layer, Indicates the The entity relationships of the original articles after transformation by the GNN layer.

[0102] Furthermore, the confidence determination layer is used to calculate the confidence of the structured original clause to obtain the confidence determination value, including:

[0103] The original structured clause is transformed into a single-dimensional vector to obtain a single-dimensional clause, wherein the representation of the single-dimensional clause is:

[0104] ,

[0105] in, Indicates the The single-dimensional provisions of the original article, Indicates the The first entity after the original article is transformed by the GNN layer, and the first entity is obtained by performing a single-dimensional vector transformation vector values, Indicates the The entity relationship of the original article after the GNN layer transformation, the first vector values, Indicates the The tail entity of the original article after the GNN layer transformation is performed on the single-dimensional vector vector values;

[0106] Use the confidence calculation function to calculate the confidence of the single-dimensional clause and obtain the confidence judgment value .

[0107] It's important to explain that converting the structured original text into a single-dimensional vector simplifies complex text into a unified format, facilitating the subsequent calculation and comparison of confidence values. This is because single-dimensional vectors can quickly identify similarities and differences between texts, improving processing efficiency. In other words, the semantics and features of the single-dimensional vector are compressed into numerical form, allowing for direct calculation of confidence values.

[0108] Specifically, the confidence calculation function is used to calculate the confidence of the single-dimensional clause to obtain the confidence judgment value , including the confidence value calculated according to the following formula :

[0109] ,

[0110] in, yes The weight value of yes The weight value of yes The weight value of .

[0111] As can be seen from the above, when the original text set is input into the original model of the standard, it passes through the BERT layer, LSTM layer and GNN layer in sequence to obtain the original structured form of the standard. , and secondly, the original clauses in structured form Perform single-dimensional vector conversion to obtain single-dimensional clauses Finally, the confidence of the single-dimensional clause is calculated by the confidence calculation function to obtain the confidence judgment value .

[0112] It should be noted that the embodiment of the present invention sets The weight value of is negative. and The weight value of is a positive number.

[0113] S5. According to the difference between the true confidence value and the confidence judgment value, adjust the model parameters of the BERT layer, LSTM layer, and GNN layer until the difference between the true confidence value and the confidence judgment value is less than a preset threshold value, store the model parameters of the BERT layer, LSTM layer, and GNN layer in the specification clause database, and obtain the specification clause standard model at the same time.

[0114] In detail, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted according to the difference between the true confidence value and the confidence judgment value until the difference between the true confidence value and the confidence judgment value is less than a preset threshold, and the model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the regulatory clause database, and the regulatory clause standard model is obtained at the same time, including:

[0115] Using the confidence judgment function, the difference between the true confidence value and the confidence judgment value is calculated;

[0116] Determine the relationship between the difference between the true confidence value and the confidence determination value and a preset threshold;

[0117] When the difference between the true confidence value and the confidence judgment value is greater than a preset threshold, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted. The model parameters include the number of attention heads of the BERT layer, the number of hidden units and sequence length of the LSTM layer, and the number of stacking layers and aggregation function of the GNN layer.

[0118] When the difference between the true confidence value and the confidence judgment value is less than or equal to a preset threshold, the adjusted model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the standard article database, and when the model parameters of the BERT layer, LSTM layer, and GNN layer are successfully stored in the standard article database, the standard article standard model automatic generation program is started, wherein the standard article standard model automatic generation program is embedded with a standard article generation template;

[0119] The standard model of the regulatory clauses is automatically generated by using the program, and the model parameters of the stored BERT layer, LSTM layer, and GNN layer are extracted from the regulatory clause database, and the regulatory clause generation template is combined to obtain the standard model of the regulatory clauses.

[0120] It should be explained that the difference between the true confidence value and the judgment value output by the model is calculated through the confidence judgment function and compared with the pre-set threshold. If the difference is greater than the threshold, the model parameters of the BERT layer, LSTM layer and GNN layer are adjusted, including the number of attention heads of the BERT layer, the number of hidden units and sequence length of the LSTM layer, and the number of stacking layers and aggregation function of the GNN layer; if the difference is less than or equal to the threshold, the adjusted parameters are stored in the standard clause database. When the model parameters are successfully stored, the automatic generation program of the standard clause model is started. The program extracts the stored model parameters of the BERT layer, LSTM layer and GNN layer from the standard clause database, and combines them with the embedded standard clause generation template to automatically generate the standard clause model. This process iteratively optimizes the model parameters to ensure that the model output meets the preset confidence requirements and ultimately generates a standardized standard clause model.

[0121] Compared with the problems described in the background technology, the present invention clarifies the model structure of each layer. Specifically, the BERT layer, LSTM layer, and GNN layer have clear division of labor, and comprehensively analyze the clauses from the semantic, sequence, and structured information dimensions respectively to ensure the accuracy of confidence judgment. At the same time, the present invention drives parameter optimization through the difference between the true value of the confidence and the judgment value, and combines threshold control to achieve rapid convergence, taking into account both accuracy and efficiency. In addition, through clause database support, structured processing, automatic generation, and templated output, full automation from data input to model output is achieved, significantly improving generation efficiency. Therefore, the digital model generation method, system, electronic device, and computer-readable storage medium for building intelligent design specification clauses proposed in the present invention can ensure the efficient generation of specification clause models under accurate confidence judgment.

[0122] Example 2:

[0123] like Figure 2 The figure shows a functional module diagram of a digital model generation system for building intelligent design specification clauses provided by an embodiment of the present invention. The digital model generation system 100 for building intelligent design specification clauses of the present invention can be installed in an electronic device. According to the functions to be implemented, the digital model generation system 100 for building intelligent design specification clauses can include a clause pulling module 101, a structural clause acquisition module 102, a confidence determination module 103 and a standard model completion module 104. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, which are stored in the memory of the electronic device;

[0124] The clause extraction module 101 is configured to receive a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and extract an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form;

[0125] The structure clause acquisition 102 is used to acquire the structure clause of each original clause in text form, wherein the structure clause is in the form of a quadruple and consists of an entity clause and a true confidence value;

[0126] The confidence determination module 103 is used to obtain the original model of the regulatory provisions, wherein the original model of the regulatory provisions includes a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer in the order from input to output of data. The original set of textual provisions is input into the original model of the regulatory provisions to perform confidence determination and obtain a confidence determination value;

[0127] The standard model completion module 104 is used to adjust the model parameters of the BERT layer, LSTM layer, and GNN layer according to the difference between the true confidence value and the confidence judgment value, until the difference between the true confidence value and the confidence judgment value is less than a preset threshold value, and store the model parameters of the BERT layer, LSTM layer, and GNN layer in the standard clause database, and simultaneously obtain the standard clause standard model.

[0128] In detail, the modules in the digital model generation system 100 for building intelligent design specification clauses in the embodiment of the present invention are used in the same manner as above. Figure 1 The same technical means are used as the method for generating digital models of the building intelligent design specifications described in the specification, and can produce the same technical effects, so I will not go into details here.

[0129] Example 3:

[0130] like Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device for implementing a method for generating a digital model of building intelligent design specification clauses provided by an embodiment of the present invention.

[0131] The electronic device 1 may include a processor 10, a memory 11, a bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a program for generating a digital model of intelligent building design specification clauses.

[0132] The memory 11 includes at least one type of readable storage medium, including flash memory, a removable hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the electronic device 1, such as a removable hard disk of the electronic device 1. In other embodiments, the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in removable hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. Furthermore, the memory 11 may include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 can be used not only to store application software installed in the electronic device 1 and various types of data, such as the code for the program that generates digital models of intelligent building design specifications, but also to temporarily store data that has been output or is about to be output.

[0133] In some embodiments, the processor 10 may be comprised of an integrated circuit, such as a single packaged integrated circuit or a combination of multiple packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting the various components of the electronic device using various interfaces and circuits. It executes programs or modules stored in the memory 11 (e.g., a program for generating digital models of intelligent building design specifications) and accesses data stored in the memory 11 to perform various functions and process data.

[0134] The bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0135] Figure 3 Only the electronic device with components is shown, and it can be understood by those skilled in the art that Figure 3 The structure shown does not constitute a limitation on the electronic device 1 , and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0136] For example, although not shown, the electronic device 1 may further include a power supply (e.g., a battery) to power various components. Preferably, the power supply may be logically connected to the at least one processor 10 via a power management device, thereby enabling functions such as charge management, discharge management, and power consumption management via the power management device. The power supply may further include any components such as one or more DC or AC power supplies, a recharging device, a power failure detection circuit, a power converter or inverter, and a power status indicator. The electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which are not further described here.

[0137] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0138] Optionally, the electronic device 1 may further include a user interface, which may be a display or an input unit (such as a keyboard). Optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display may also be appropriately referred to as a display screen or a display unit, and is used to display information processed in the electronic device 1 and to display a visual user interface.

[0139] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0140] The program for generating a digital model of intelligent building design specification clauses stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When executed in the processor 10, the program can achieve the following:

[0141] receiving a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and pulling an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form;

[0142] Obtain the structure clause of each original clause in text form, wherein the structure clause is in the form of a four-tuple and consists of an entity clause and a true confidence value;

[0143] Obtain the original model of the regulatory provisions, where the original model of the regulatory provisions includes, in order from data input to data output, a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer;

[0144] Inputting the original clause set in text form into the original model of the standard clause to perform confidence determination and obtain a confidence determination value;

[0145] According to the difference between the true confidence value and the confidence judgment value, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted until the difference between the true confidence value and the confidence judgment value is less than a preset threshold. The model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the specification clause database, and the standard model of the specification clause is obtained at the same time.

[0146] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 3 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0147] Furthermore, if the modules / units integrated into the electronic device 1 are implemented as software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. The computer-readable storage medium may be volatile or non-volatile. For example, the computer-readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0148] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by a processor of an electronic device, the computer program can implement:

[0149] receiving a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and pulling an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form;

[0150] Obtain the structure clause of each original clause in text form, wherein the structure clause is in the form of a four-tuple and consists of an entity clause and a true confidence value;

[0151] Obtain the original model of the regulatory provisions, where the original model of the regulatory provisions includes, in order from data input to data output, a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer;

[0152] Inputting the original clause set in text form into the original model of the standard clause to perform confidence determination and obtain a confidence determination value;

[0153] According to the difference between the true confidence value and the confidence judgment value, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted until the difference between the true confidence value and the confidence judgment value is less than a preset threshold. The model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the specification clause database, and the standard model of the specification clause is obtained at the same time.

[0154] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of the solution of this embodiment according to actual needs.

[0155] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional modules.

[0156] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for generating a digital model of building intelligent design specification clauses, characterized in that: The method comprises: receiving a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and pulling an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form; Obtain the structure clause of each original clause in text form, wherein the structure clause is in the form of a four-tuple and consists of an entity clause and a true confidence value; Obtain the original model of the regulatory provisions, where the original model of the regulatory provisions includes, in order from data input to data output, a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer; Inputting the original clause set in text form into the original model of the standard clause to perform confidence determination and obtain a confidence determination value; According to the difference between the true confidence value and the confidence judgment value, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted until the difference between the true confidence value and the confidence judgment value is less than a preset threshold. The model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the regulatory provisions database, and the regulatory provisions standard model is obtained at the same time; The confidence determination layer includes a confidence calculation function and a confidence determination function, wherein the confidence determination function is: ， in, Represents the difference between the confidence judgment value and the true confidence value, Indicates the original article The original text in the form of the article is input into the BERT layer, LSTM layer, and GNN layer of the standard article original model to obtain the entity article. The textual difference between the substantive clauses of the original clauses, Indicates the The confidence value of the original article, Indicates the The true value of the confidence of the entity clause corresponding to the original clause.

2. The method for generating a digital model of a building intelligent design specification clause according to claim 1, wherein: According to the generation instruction, an original article set is pulled from a pre-built standard article database, wherein each original article in the original article set is in text form, including: Parsing the generation instruction to obtain the building information of the digital model of the specification clause before generation, wherein the building information includes building type, construction time, building scale, building style and material; Using the building information as a first search condition, searching the specification article database to obtain a first article collection; Determine whether the size of the first set of articles is larger than the data size generated by the pre-set digital model of regulatory articles; If the size of the first set of articles is not greater than the size of the data generated by the digital model of the standard articles, the construction time in the building information is removed to obtain the second search condition, and the second search condition is used to perform another search in the standard article database to obtain the second set of articles; Determine whether the size of the second set of articles is larger than the size of data generated by the digital model of regulatory articles; If the size of the second set of articles is not greater than the data size generated by the digital model of the standard articles, a generation failure instruction of the digital model of the standard articles is directly obtained, and the generation failure instruction is sent to the port of the generation instruction of the digital model of the standard articles; If the scale of the first or second article collection is larger than the data scale generated by the digital model of the standard articles, the first or second article collection shall be the original article collection.

3. The method for generating a digital model of a building intelligent design specification clause according to claim 2, characterized in that: The structural clauses are expressed as follows: ， in, Indicates the The structural provisions of the original provisions in the form of article texts, Indicates the The substantive provisions of the original article, Indicates the The header entity of the original article, Indicates the The end entity of the original article, Represents the entity relationship between the head entity and the tail entity, Indicates the The true value of the confidence of the entity clause corresponding to the original clause.

4. The method for generating a digital model of building intelligent design specification clauses according to claim 1, characterized in that: The step of inputting the original set of clauses in text form into the original model of the standard clauses to perform confidence determination and obtain a confidence determination value includes: Each original article in the original article set in text form is input into the standard article original model in sequence, and the confidence judgment value of each original article is calculated as follows: Use the BERT layer to perform word encoding on the original text to obtain the original text in vectorized form; Import the original clauses in vectorized form into the LSTM layer, and use LSTM to correct the original clauses in vectorized form to obtain the corrected original clauses; Use the GNN layer to transform the corrected original clauses into structured original clauses; The confidence level of the original text in structured form is calculated using the confidence level determination layer to obtain the confidence level determination value. .

5. The method for generating a digital model of building intelligent design specification clauses according to claim 4, characterized in that: The step of calculating the confidence of the original text in a structured form by using the confidence determination layer to obtain the confidence determination value includes: The original structured clause is transformed into a single-dimensional vector to obtain a single-dimensional clause, wherein the representation of the single-dimensional clause is: ， in, Indicates the The single-dimensional provisions of the original article, Indicates the The first entity after the original article is transformed by the GNN layer, and the first entity is obtained by performing a single-dimensional vector transformation vector values, Indicates the The entity relationship of the original article after the GNN layer transformation, the first vector values, Indicates the The tail entity of the original article after the GNN layer transformation is performed on the single-dimensional vector vector values; Use the confidence calculation function to calculate the confidence of the single-dimensional clause and obtain the confidence judgment value .

6. The method for generating a digital model of building intelligent design specification clauses according to claim 5, characterized in that: The confidence calculation function is used to calculate the confidence of the single-dimensional clause to obtain the confidence judgment value , including the confidence value calculated according to the following formula : ， in, yes The weight value of yes The weight value of yes The weight value of .

7. The method for generating a digital model of building intelligent design specification clauses according to claim 6, characterized in that: described The weight value of is negative. and The weight value of is a positive number.

8. The method for generating a digital model of building intelligent design specification clauses according to claim 7, characterized in that: The method adjusts the model parameters of the BERT layer, the LSTM layer, and the GNN layer according to the difference between the true confidence value and the confidence judgment value until the difference between the true confidence value and the confidence judgment value is less than a preset threshold value, stores the model parameters of the BERT layer, the LSTM layer, and the GNN layer in the standard clause database, and simultaneously obtains the standard clause model, including: Using the confidence judgment function, the difference between the true confidence value and the confidence judgment value is calculated; Determine the relationship between the difference between the true confidence value and the confidence determination value and a preset threshold; When the difference between the true confidence value and the confidence judgment value is greater than a preset threshold, the model parameters of the BERT layer, LSTM layer, and GNN layer are adjusted. The model parameters include the number of attention heads of the BERT layer, the number of hidden units and sequence length of the LSTM layer, and the number of stacking layers and aggregation function of the GNN layer. When the difference between the true confidence value and the confidence judgment value is less than or equal to a preset threshold, the adjusted model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the standard article database, and when the model parameters of the BERT layer, LSTM layer, and GNN layer are successfully stored in the standard article database, the standard article standard model automatic generation program is started, wherein the standard article standard model automatic generation program is embedded with a standard article generation template; The standard model of the regulatory clauses is automatically generated by using the program, and the model parameters of the stored BERT layer, LSTM layer, and GNN layer are extracted from the regulatory clause database, and the regulatory clause generation template is combined to obtain the standard model of the regulatory clauses.

9. A digital model generation system for building intelligent design specification clauses, characterized by: The system is used to perform the method according to any one of claims 1 to 8, and the system comprises: A clause extraction module is configured to receive a generation instruction for a digital model of a specification clause, wherein the digital model of the specification clause is applied in the field of architectural design, and extract an original clause set from a pre-built specification clause database according to the generation instruction, wherein each original clause in the original clause set is in text form; The structure clause acquisition module is used to obtain the structure clause of each original clause in text form, wherein the structure clause is in the form of a four-tuple and consists of an entity clause and a true confidence value; The confidence determination module is used to obtain the original model of the regulatory provisions. The original model of the regulatory provisions includes a BERT layer, an LSTM layer, a GNN layer, and a confidence determination layer in the order from input to output of data. The original set of textual provisions is input into the original model of the regulatory provisions to perform confidence determination and obtain a confidence determination value. The standard model completion module is used to adjust the model parameters of the BERT layer, LSTM layer, and GNN layer according to the difference between the true confidence value and the confidence judgment value, until the difference between the true confidence value and the confidence judgment value is less than a preset threshold. The model parameters of the BERT layer, LSTM layer, and GNN layer are stored in the standard clause database, and the standard clause model is obtained at the same time.

Images