BIM model examination method based on text classification and named entity recognition

Abstract

The invention relates to the technical field of model review, in particular to a BIM model review method based on text classification and named entity recognition, and provides the following scheme: building a building specification knowledge graph, obtaining BIM review data through a review data extraction strategy, obtaining a matching rule set by utilizing knowledge graph matching, and adjusting the priority for review. Outputting the report; the knowledge graph construction adopts an ontology model and a knowledge extraction network, component attributes, spatial relationships and time dimensions are considered in the examination data extraction, the matching process is combined with multi-dimensional semantic matching and reasoning, and the examination accuracy and efficiency are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of model review, and in particular to a BIM model review method based on text classification and named entity recognition. Background Art

[0002] The construction industry is an indispensable basic industry in people's daily life. Due to the large population of my country, various types of buildings are needed, so the development of the construction industry is of great significance to my country. Building Information Modeling (BIM) is a revolutionary technology in the field of construction and an advanced development of CAD technology. It is based on the three-dimensional data attributes of the building, takes the three-dimensional geometric model as the carrier, and integrates all the information files and material properties in the construction process of the building project.

[0003] Traditional building compliance review is based on professional reviewers to compare and check the two-dimensional CAD drawings of the building design. This method is time-consuming, labor-intensive and prone to errors. BIM compliance review uses BIM to automatically check the building design. This method greatly liberates manpower and material resources, can find more problems in the design stage, and can express the review results more intuitively. At present, BIM compliance review is implemented by hard-coding the specifications, that is, manually interpreting the specification provisions and translating them into a computer-executable language to review the construction project. This method is highly feasible and the review results have a relatively high accuracy, but it also has its limitations. First, hard coding requires interpreting and programming the provisions one by one. Faced with the current huge building compliance review system, translating and encoding all the provisions will consume a lot of manpower, and most of the provisions have similar review processes but different review objects. This method wastes human resources to a certain extent. Secondly, the hard coding method is not flexible enough. Provisions with similar provisions but inconsistent requirements in different regions need to be rewritten, and the provisions will be constantly revised. Each revision requires upgrading the source code, which has poor maintainability. In actual projects, the specifications and standards issued by the state are only the minimum requirements for construction projects, and the different requirements of different projects may be more flexible. It would be very cumbersome to hard-code such variable construction specifications.

[0004] For example, the patent with publication number CN116992302A discloses a BIM model review method and device, wherein the method includes: obtaining the coding and multiple attribute information of all components in the BIM model, wherein the coding of the components is generated according to the digital coding rules when the BIM model is established, wherein the digital coding rules include multiple nodes and node description information, wherein the key of each node includes the identifier of the node, the values ​​of the root node and the child node include the identifier of the node at the next level, and the value of the leaf node is empty; according to the coding of each component, obtaining the node description information corresponding to each component; sequentially comparing the node description information corresponding to each component with the multiple attribute information of the same component, and outputting the matching degree of each component; according to the matching degree of each component, outputting the review result of each component of the BIM model. This invention can improve the accuracy and efficiency of BIM model review.

[0005] The above patents all have the problems raised by this background technology: the traditional review system relies on manual work, has low processing efficiency, and is difficult to keep up with rapidly changing industry standards and specifications; the weight setting of similar clauses in different contexts is unreasonable, resulting in low review efficiency. To solve the above problems, this application designs a BIM model review method based on text classification and named entity recognition. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a BIM model review method based on text classification and named entity recognition in view of the deficiencies of the prior art, to construct a knowledge graph of building specifications, to obtain BIM review data through review data extraction strategies, to obtain matching rule sets through knowledge graph matching, to adjust the priority for review, and to output a report. The knowledge graph is constructed using an ontology model and a knowledge extraction network, and the review data extraction considers component attributes, spatial relationships, and time dimensions. The matching process combines multi-dimensional semantic matching and reasoning to ensure review accuracy and efficiency.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A BIM model review method based on text classification and named entity recognition, the method comprising:

[0009] Acquire a specification text in the field of construction, construct a construction specification ontology according to the specification text in the field of construction, and construct a construction specification knowledge graph according to the construction specification ontology;

[0010] According to the BIM model to be reviewed, the BIM model is processed by a review data extraction strategy to obtain a review data file and context relevance;

[0011] Matching the review data file in the building specification knowledge graph to obtain a matching rule set;

[0012] The priority of the matching rule set is adjusted according to the context relevance, and the BIM model to be reviewed is reviewed according to the adjusted matching rule set, and a review report is output.

[0013] The construction of the building specification knowledge graph includes:

[0014] Determine the building specification ontology according to the standard text in the building field;

[0015] According to the keywords and key terms of the domain determined by the building specification ontology, a building specification ontology knowledge model is established from top to bottom, wherein the building specification ontology knowledge model includes a top-level ontology, a concept subtree and a sub-layer instance;

[0016] Add the construction field specification text as a knowledge feature to the concept subtree, perform a logical check on the construction specification ontology knowledge model based on expert knowledge, and determine whether it meets the construction field ontology construction principle. If not, continue to add concept subtrees;

[0017] Constructing a knowledge extraction network, processing the construction field specification text corresponding to the concept subtree through the knowledge extraction network, extracting the construction specification sequence, and adding the construction specification sequence to the sublayer instance;

[0018] Construct a building specification knowledge graph, instantiate the building specification ontology knowledge model, and add it to the building specification knowledge graph according to the entry strategy.

[0019] The knowledge extraction network comprises:

[0020] The input layer maps the architectural field specification text corresponding to the concept subtree through a pre-trained word vector model, converts it into a word vector matrix, and calculates the knowledge feature vector according to the word vector matrix;

[0021] A feature extraction layer extracts semantic features from the knowledge feature vector through a bidirectional long short-term memory neural network;

[0022] The output layer calculates the label score matrix by scoring the semantic features, extracts the optimal sequence according to the label score matrix, decodes the optimal sequence, and calculates the building specification sequence.

[0023] The entry strategy includes:

[0024] Obtain the number of concept branches, keywords and keyword attributes of the input instantiated building specification ontology knowledge model;

[0025] Determine the semantic tree where the instantiated building specification ontology knowledge model is located in the building specification knowledge graph, traverse all nodes of the semantic tree, extract the keyword pairs with the smallest semantic distance between the nodes and the instantiated building specification ontology knowledge model as similar node pairs, and calculate the semantic similarity of the similar node pairs;

[0026] The highest semantic similarity is compared with the similarity threshold. If it is less than the similarity threshold, the instantiated building specification ontology knowledge model is used as the sibling node of the node with the highest semantic similarity. If it is greater than or equal to the similarity threshold, the instantiated building specification ontology knowledge model is merged with the node with the highest semantic similarity.

[0027] The data extraction strategy for the review includes:

[0028] The BIM model to be reviewed is processed in layers according to the component type;

[0029] According to the layered BIM model, component attribute information, spatial relationship data and time dimension data are extracted;

[0030] Integrate the component attribute information, spatial relationship data and time dimension data into a review data file, wherein the review data file includes component geometry information, attribute information, spatial relationship table and time node data;

[0031] The context information of the component and its surrounding associated components is modeled through the graph neural network model. According to the constructed context dependency model, the correlation between each component and its associated components is calculated. The correlation is dynamically updated through the message passing mechanism between nodes, and the correlation is weighted and summed to output the context correlation.

[0032] The matching in the building specification knowledge graph according to the review data file includes:

[0033] According to the review data file, feature extraction is performed through a pre-trained model to generate a multi-dimensional feature vector;

[0034] The multi-dimensional feature vector is compared with the specification clause vector in the building specification knowledge graph through a dynamic semantic matching algorithm, and a preliminary match is performed according to the multi-dimensional semantic matching strategy to obtain a candidate answer set;

[0035] Performing semantic reasoning on the candidate answer set according to the relationship between nodes and edges in the building specification knowledge graph, and obtaining a matching rule set according to matching parameters, wherein the matching parameters include semantic similarity, spatial similarity, and temporal similarity;

[0036] If the matching rule set does not satisfy the matching threshold, the matching parameters are adjusted through a feedback optimization mechanism until the matching rule set satisfies the matching threshold.

[0037] The multi-dimensional semantic matching strategy includes:

[0038] Match the geometric properties of components in the BIM model with the geometric specifications of components in the building specification knowledge graph through geometric information features;

[0039] Matching time node data with construction timing requirements in the building specification knowledge graph through time information features;

[0040] The component attribute information is matched with the specification terms in the building specification knowledge graph through semantic information features.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention transforms complex building specification text into a structured knowledge model by constructing a building specification knowledge graph, so that the review process of the BIM model can automatically and quickly find matching rules in the knowledge graph, thereby improving the efficiency and accuracy of the review;

[0043] 2. The present invention calculates the context relevance and adjusts the priority of the matching rule set, so that the review process not only relies on the direct matching of the rules, but also takes into account the complex relationship between the components in the BIM model, thereby making more reasonable and comprehensive review decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0045] Figure 1 This is a flow chart of a BIM model review method based on text classification and named entity recognition according to Embodiment 1 of the present invention;

[0046] Figure 2 This is a knowledge extraction network structure diagram of Example 1 of the present invention;

[0047] Figure 3 This is a flow chart of the instantiation strategy of Example 1 of the present invention;

[0048] Figure 4 This is a module structure diagram of a BIM model review system based on text classification and named entity recognition according to Embodiment 2 of the present invention;

[0049] Figure 5 This is a schematic diagram of the construction clause text classification process of Example 2 of the present invention;

[0050] Figure 6 This is a schematic diagram of the named entity extraction process of building clause text according to Example 2 of the present invention. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0052] Example 1

[0053] See also Figure 1 The present invention provides an embodiment: a BIM model review method based on text classification and named entity recognition, wherein the specific steps of the method are as follows:

[0054] S1: Obtaining a text of a building specification, constructing a building specification ontology according to the text of a building specification, and constructing a building specification knowledge graph according to the building specification ontology;

[0055] In this step, text data is first obtained from multiple construction field specification texts. The specification texts may include national construction standards, local construction regulations, and specification requirements for specific engineering projects. These specification texts are parsed and processed through natural language processing (NLP) technology to identify key terms and rules;

[0056] S2: According to the BIM model to be reviewed, the BIM model is processed by a review data extraction strategy to obtain a review data file and context relevance;

[0057] In this step, data is extracted based on the component type of the BIM model, the spatial position of the component in the building structure, the construction stage, etc., specifically including the following aspects:

[0058] Geometric information: Extract geometric information such as three-dimensional size, shape, position, etc. of each component.

[0059] Property information: including physical properties of components such as material, durability, and load capacity.

[0060] Spatial relationship data: Analyze the relative position of a component with other surrounding components, and calculate the distance, angle, and geometric overlap between components.

[0061] Time dimension data: Extract the construction time nodes and installation sequence of each component according to the construction plan;

[0062] Based on the component location, functional dependency, physical connection and other information in the BIM model, the context dependency model is constructed using graph neural network (GNN) technology. Through the message passing mechanism, the context relevance of each component with its surrounding components is calculated. Components with high context relevance mean that they are more important in the entire building structure;

[0063] S3: matching the review data file in the building specification knowledge graph to obtain a matching rule set;

[0064] S4: adjusting the priority of the matching rule set according to the context relevance, reviewing the BIM model to be reviewed according to the adjusted matching rule set, and outputting a review report;

[0065] Dynamically adjust the priority of clause analysis according to the specific context of the BIM model, automatically identify specific building types, component types and structural characteristics, assign different weights to clause analysis according to their importance in the building, and adjust the priority of the review standards;

[0066] The construction of the building specification knowledge graph relies on the ontology model and natural language processing technology in the construction field. Through the analysis, keyword extraction and serialization of the specification text, the text information is converted into a visual knowledge graph. This process combines the knowledge of domain experts with automated information extraction technology to ensure that the graph not only meets the specific standards of the construction industry, but can also be dynamically updated and expanded.

[0067] The specific steps of S1 are as follows:

[0068] S1.1: Based on the text of the construction field specification, the predefined construction field ontology framework is first used to parse the text and identify the important entities and relationships in the text. For example, keywords such as "structural components", "safety standards", and "building materials" involved in the specification text will be marked as important terms;

[0069] In this process, ontology determination uses the entity recognition model (NER) in natural language processing. The trained model is used to identify specific terms in the field of architecture, and a preliminary ontology structure is established based on the logical relationship between the terms.

[0070] S1.2: After determining the domain keywords and key terms, the building code ontology knowledge model is constructed in a top-down manner. The hierarchical structure of the model includes the top-level ontology, the concept subtree and the sub-level instance;

[0071] The top-level ontology represents the core knowledge in the field of construction, such as the main functional modules of buildings, types of building components, and standard specification clauses; the concept subtree is a refinement of these core modules, including more specific building materials, component details, and related safety standards; the sub-layer instance is an instantiation application in each specific building project or scenario, such as the actual application case of a specific building material or a specific safety specification;

[0072] S1.3: After constructing the preliminary architectural ontology knowledge model, the logic of the model is verified through expert knowledge. The expert system will check the position of each node (keyword or term) in the concept subtree to ensure the hierarchical relationship and logical consistency between terms. If some concepts do not meet the standards of the architectural field, the ontology structure will be automatically adjusted or new subtrees will continue to be added until the model meets the ontology construction principles of the architectural field;

[0073] The construction principles of the ontology in the construction field are derived from the five basic principles proposed by Gruber, including: accuracy, consistency, scalability, minimum ontology commitment and coding bias; the accuracy means that the ontology of the field should accurately convey the meaning of the terms it contains, the consistency means that the definition of the ontology is consistent, and the initial definition of the ontology must be logically consistent with the results derived from it, and there must be no contradictions, the scalability means that in the construction of the ontology, not only the vocabulary that reaches industry consensus should be used, but also the scope of use of the concept should be clarified to ensure that the ontology can be easily expanded without external interference. With the development and improvement of the knowledge of the transfer equipment, the constructed equipment ontology should maintain scalability, the minimum ontology commitment means reducing the constraints on the target object of the construction ontology, and in the process of constructing the equipment ontology, the constraints on the terminology should be reduced as much as possible to ensure the diversity of the constructed ontology, and the coding bias means that the concept needs to be introduced at the knowledge level, rather than relying on the coding of a specific symbol level;

[0074] S1.4: Construct a knowledge extraction network to process the architectural specification text in the concept subtree. The knowledge extraction network is a model based on the combination of bidirectional long short-term memory network (BiLSTM) and conditional random field (CRF), which can effectively identify the semantic relationship in the text and extract the key architectural specification sequence;

[0075] The network first performs word embedding on the input building specification text, and maps each word into a high-dimensional vector through a pre-trained word embedding model to represent the semantic information of the word.

[0076] Then, a bidirectional long short-term memory network (BiLSTM) captures the dependencies between words in the text through forward and backward time steps, ensuring that long-distance dependencies across sentences can be identified, which is particularly suitable for complex semantic structures in building codes.

[0077] Finally, the conditional random field (CRF) is used to optimize and decode the output label sequence to ensure that the extracted building code sequence meets expectations. For example, key clauses such as "firewall thickness requirement ≥ 0.3m" can be accurately extracted from the code text;

[0078] Through the knowledge extraction network, the core provisions can be automatically and efficiently extracted from the massive amount of building specification texts to construct a building specification sequence. This not only reduces the workload of manual screening, but also greatly improves the efficiency and accuracy of building specification review. For different construction projects, the parsing strategy of the knowledge extraction network can be quickly adjusted according to project requirements to ensure the flexibility of the review;

[0079] S1.5: After the extraction of the specification sequence is completed, the building specification ontology knowledge model is further converted into a building specification knowledge graph through an instantiation strategy, and in this process, the structure and node content of the knowledge graph are dynamically adjusted according to different building scenarios;

[0080] The instantiation of the knowledge graph maps the concepts and specification sequences of the building code ontology knowledge model into nodes and edges in the knowledge graph. For example, concepts such as "wall" and "fire resistance level" will be instantiated as nodes in the knowledge graph, and the "dependency between wall and fire resistance level" will be represented as an edge in the graph.

[0081] The instantiated knowledge graph can add new nodes and relationships according to the needs of specific construction projects, ensuring the dynamic scalability of the graph. For example, for a high-rise building project, nodes for specific fire protection regulations and structural stability requirements required for high-rise buildings can be automatically added;

[0082] See also Figure 2 , the knowledge extraction network structure diagram of the embodiment of the present invention, the network uses a multi-layer neural network structure, including an input layer, a feature extraction layer and an output layer, to perform knowledge mapping, feature extraction, scoring and optimal sequence derivation, and finally generate a building specification sequence. The knowledge extraction network is specifically as follows:

[0083] The input layer is mainly responsible for mapping the normative text in the architectural field corresponding to the concept subtree to a high-dimensional vector space, providing a basis for subsequent feature extraction. The normative text in the architectural field corresponding to the concept subtree is mapped through a pre-trained word vector model, and converted into a word vector matrix, which is converted into the query vector H, key vector K and value vector V of the input parameters. They are represented by the word embedding of the text, and the attention matrix is ​​calculated through the self-attention mechanism. The calculation formula of the self-attention mechanism is as follows:

[0084]

[0085] Among them, M A represents the attention matrix, Softmax(·) represents the activation function, T represents the matrix transpose, and d represents the linear dimension;

[0086] The word vector matrix is ​​adjusted according to the attention matrix and continuously updated through the network training process to optimize the knowledge explicit features of the word vector matrix. The knowledge feature vector is calculated according to the weight coefficient matrix. The calculation formula of the knowledge feature vector is:

[0087]

[0088] Among them, V c represents the knowledge feature vector, ω q ,ω k and ω v represents the weight coefficient matrix;

[0089] The main function of the feature extraction layer is to extract deep semantic features from the knowledge feature vector and obtain contextual information in the time series. The layer includes a bidirectional long short-term memory neural network, which is composed of a forward LSTM and a backward LSTM, respectively obtaining forward and backward information in the knowledge feature vector word sequence. The forward LSTM processes the order of the input sequence from left to right to generate forward semantic information for each time step, and the backward LSTM processes the order of the input sequence from right to left to generate backward semantic information for each time step, and merges the forward and backward information into the output at the same time to obtain semantic features;

[0090] The output of the bidirectional long short-term memory network contains the contextual information of each position in the input sequence, so that the network can understand the complex relationship between words in the sequence. Through the feature extraction layer, the time dependency in the knowledge feature vector is extracted to provide semantic context support for subsequent sequence scoring and optimal sequence derivation.

[0091] The output layer first scores the semantic features to obtain a score sequence of the semantic feature sequence. The scoring process is implemented through a linear transformation and a Softmax function to generate a score for the label in each semantic feature and calculate a label score matrix.

[0092] Then, the scoring sequence of the semantic feature sequence is globally evaluated through the conditional random field, and the relationship between adjacent labels in the sequence is considered to determine whether the transfer score matrix meets the global optimality. If it does not meet the global optimality, the transfer score matrix is ​​recalculated. If it meets the global optimality, the label score matrix is ​​adjusted through the transfer score matrix to make the label score matrix more consistent with the law of sequence dependence.

[0093] The Viterbi algorithm is used to obtain the optimal sequence according to adjacent tags. The Viterbi algorithm calculates the scores of all possible paths from the start of the sequence to the current time step, records the best path at each step, that is, the tag sequence with the highest score. After calculating the best path score of the last step, backtracking is performed to gradually determine the optimal tag at each time step, generate the optimal sequence, and decode the optimal sequence to obtain the maintenance knowledge sequence. This step mainly maps the optimal tag sequence back to the original building code knowledge. The calculation formula for the optimal sequence is:

[0094]

[0095] Among them, B q represents the optimal sequence, Score{·} represents the Viterbi algorithm, j represents the serial number of a single tag in the semantic feature sequence, M represents the total number of tags in the semantic feature sequence, D represents the transition score matrix, y j represents the j-th tag in the semantic feature sequence, y j+1 represents the (j + 1)-th tag in the semantic feature sequence, Q represents the tag score matrix, represents matrix multiplication, p represents the conditional probability, p(y j 丨y j+1 ) represents the transition conditional probability between the (j + 1)-th tag and the j-th tag in the semantic feature sequence;

[0096] Through the global evaluation of the random conditional field and the Viterbi algorithm, the output layer not only selects the tag with the highest score at each time step, but also considers the dependencies of the entire sequence to ensure that the generated tag sequence is globally optimal;

[0097] Please refer to Figure 3 , the flowchart of the instantiation strategy disclosed in the present invention. First, according to the instantiated building code ontology knowledge model, the instantiated building code ontology knowledge model is converted into concept branch numbers, keywords, and keyword attribute knowledge for addition, and then similarity detection is performed. The mechanism of similarity detection is to detect whether there is knowledge in the knowledge graph whose similarity with the knowledge concept to be entered is greater than the set threshold. The detection process requires calculating the semantic relevance between the keywords of the knowledge to be entered and the keywords of each record in the database. Experts set the entry threshold. Knowledge entries with a similarity higher than this threshold will be regarded as records with high similarity, indicating that such records already exist in the knowledge base, and this knowledge will not be allowed to be directly entered into the knowledge base. If it is lower than this threshold, it indicates that such a knowledge entry does not exist in the knowledge base, and this knowledge can be entered into the knowledge graph through review;

[0098] The instantiation strategy includes:

[0099] S1.5.1: Obtain the input instantiated building code ontology knowledge model, including building code provisions, items, keywords, key terms and their attributes (such as applicable building types, spatial layouts, component relationships, etc.);

[0100] Through natural language processing (NLP) technology, semantic analysis is performed on the standard text to extract core keywords and their attributes, and to construct concept branches and hierarchical structures. Each standard article is regarded as a concept node, and the keywords and attributes under it are sub-layer instances of the node;

[0101] S1.5.2: The instantiated building code ontology knowledge model needs to find the best semantic position in the building code knowledge graph. In this process, the semantic tree structure position of the instantiated building code ontology knowledge model in the graph is first determined. The semantic tree structure forms a semantic tree with a hierarchical structure by modeling the semantics and logical relationships of all building code provisions;

[0102] S1.5.3: Once the similar node pair with the smallest semantic distance is found, decide whether to merge the instantiated building code ontology knowledge model with the target node or make it a sibling node of the target node based on the similarity threshold;

[0103] If the calculated semantic similarity is less than the similarity threshold, it is considered that there are significant differences between the two clauses, and the instantiated building code ontology knowledge model is inserted into the semantic tree as a new node as a "brother node" of the existing similar nodes.

[0104] If the calculated semantic similarity is greater than or equal to the similarity threshold, the two nodes are considered to be highly similar or consistent, and the two nodes are merged and fused into a common node. The fusion mechanism maintains an efficient and compact knowledge graph structure to avoid redundant data. Through reasonable node fusion and parallel mechanisms, it can ensure that the semantic relationship between articles is clear while maintaining an efficient graph structure;

[0105] Assume that there are two specification clauses:

[0106] Article 1: The thickness of the load-bearing walls of a building shall not be less than 30cm.

[0107] Article 2: The thickness of load-bearing walls in residential buildings shall not be less than 25cm.

[0108] When entering these two clauses:

[0109] The core keywords of the clause “load-bearing wall”, “thickness”, “30cm” and “25cm” are parsed through NLP.

[0110] Use the BERT model to embed keywords, vectorize the articles, and calculate similarity. If the similarity between the new entry in Article 2 and Article 1 is greater than the similarity threshold, the two articles are semantically fused and marked as the same node;

[0111] If the semantic similarity is less than the threshold, clause 2 is inserted as a sibling node of clause 1, and the differences of the clauses (such as special requirements for different building types) can be recorded in the graph;

[0112] The calculation formula for semantic similarity is:

[0113] Sem s =dist α *des β *deg γ ;

[0114] Among them, Sem s Represents the semantic similarity between the node and the instantiated disease ontology, dist represents the semantic distance similarity of similar nodes, α represents the influence weight of semantic distance similarity, des represents the depth similarity of similar nodes, β represents the influence weight of depth similarity, deg represents the degree similarity of similar nodes, and γ represents the influence weight of degree similarity;

[0115] The calculation formula of the semantic distance similarity of similar nodes is:

[0116]

[0117] Among them, di(·) represents the semantic distance calculation function, A and B represent similar node pairs, i represents a single keyword of the instantiated disease ontology, n represents the total number of keywords of the instantiated disease ontology, j represents a single keyword of the node, and m represents the total number of keywords of the node;

[0118] The calculation formula for the deep similarity of similar nodes is:

[0119]

[0120] Where dep(·) represents the depth finding function, d 1 represents the average depth of the node, d 2 represents the average depth of instantiated disease ontology;

[0121] The calculation formula of degree similarity of similar nodes is:

[0122]

[0123] Where E{·} represents the Euclidean distance function, degree(·) represents the degree measurement function, and A an represents the parent node of keyword A, O represents the ancestor node of the semantic tree, and Ban Indicates the parent node of keyword B;

[0124] The BIM model is hierarchically processed according to the component type, and attribute information, spatial relationship data and time dimension data are extracted. The components and their associated components are contextually modeled through the graph neural network (GNN) model, and finally the context relevance is calculated and applied to generate the review data file. The specific steps of the review data extraction strategy are as follows:

[0125] S2.1: Based on the BIM model to be reviewed, the model is first layered according to the type of components (such as walls, beams, columns, doors and windows, etc.). The layering strategy is based on the logical structure of the building, such as spatial location, functional category, and physical connection relationship. The classification of components is not limited to geometric components, but can also be further subdivided according to properties such as function (load-bearing components, non-load-bearing components), material (concrete, steel, wood, etc.);

[0126] The hierarchical processing of components is done by parsing the spatial data and structural information of the BIM model and using spatial indexing technology for rapid classification. This process extracts and stores the spatial coordinates, dimensions and type information of each component in a hierarchical structure, facilitating subsequent multi-dimensional data extraction and review;

[0127] S2.2: The geometric information (such as length, width, height), material properties, and physical properties (such as durability, fire resistance, etc.) of each component are extracted from the BIM model and stored in a structured manner. The extraction of this information is completed through geometric analysis and attribute analysis algorithms, supporting matching with the requirements of the building code;

[0128] Use spatial index and geometric calculation algorithm to extract the spatial relationship between components (such as relative position, distance, overlap, etc.). The spatial relationship data will form a spatial relationship table to describe the position relationship of components in three-dimensional space;

[0129] By synchronizing with the construction plan data, the time dimension information such as the construction phase, installation time, and construction period of the components are extracted from the BIM model. This data is matched with the construction schedule through time series analysis technology to evaluate the compliance of the components in construction;

[0130] The extraction of spatial relationships is based on geometric analysis, using vector operations and spatial query algorithms (such as KD trees) to determine the spatial characteristics between components. The time dimension data is captured by analyzing the construction logs and planning data of the components and building a time series model to capture the construction progress;

[0131] S2.3: Use graph neural networks (GNNs) to model the contextual information of components and their associated components in the BIM model. Components are considered as nodes in the graph, and the edges between nodes are constructed based on the physical connection relationship, spatial distance, functional dependency, etc. of the components. The node features in the graph network include geometric information, functional attributes, and material quality;

[0132] S2.4: Based on the message passing mechanism in the graph neural network, the node (component) will obtain information from its associated neighbor nodes, propagate and aggregate information through the multi-layer neural network, and calculate the contextual relevance between each node (component) and its neighboring nodes. This relevance reflects the importance of a component in its surrounding structure;

[0133] S2.5: Integrate the component's geometric information, attribute information, spatial relationship table, and time node data with context relevance to form a complete review data file;

[0134] The pre-trained model is used to extract features from the review data in the BIM model, generate a multi-dimensional feature vector, and perform semantic matching, reasoning, and optimization adjustment with the provisions in the building specification knowledge graph. The core principle is to achieve accurate matching and compliance review between BIM model components and building specification provisions based on multi-dimensional parameters such as semantic, spatial, and temporal similarity through dynamic semantic matching and graph network reasoning mechanisms. The specific steps of S3 are as follows:

[0135] S3.1: Extract the geometric information, attribute information, spatial relationship data and time dimension data of the components from the BIM model, and perform feature extraction through pre-trained natural language processing models and image processing models. The core of feature extraction is to uniformly convert different behavioral data types in the BIM model into vectorized multi-dimensional representations, and normalize the data from different sources and formats (such as geometric data, attribute data, and time series data) into a unified multi-dimensional vector, while retaining its semantic and contextual information. This feature extraction step greatly improves the efficiency and accuracy of the subsequent matching and reasoning process;

[0136] S3.2: According to the dynamic semantic matching algorithm, the generated multi-dimensional feature vector is compared with the article vector in the building code knowledge graph. The matching process is based on three similarity parameters: semantic similarity, spatial similarity, and temporal similarity. The cosine similarity between the feature vector and the code article vector is calculated to obtain a preliminary set of candidate answers;

[0137] Semantic similarity: Based on the matching of component attributes with the terminology of the specification, such as matching the material and type of the component with the standard requirements defined in the specification;

[0138] Spatial similarity: based on the matching of the geometric information of the component with the spatial layout requirements in the specification, such as the distance and arrangement of the load-bearing wall and the supporting beam;

[0139] Temporal similarity: based on the matching of construction phase and component time information with the specification timing, such as the use requirements of building materials in specific construction phases;

[0140] S3.3: Based on the candidate answer set generated by the preliminary matching, the relationship between nodes and edges in the building code knowledge graph is further used for reasoning. In this stage, the contextual information of components and code provisions (such as related components, construction conditions, spatial relationships) is inferred through the graph neural network (GNN) to generate the optimal matching rule set;

[0141] The nodes of components in the building code knowledge graph represent clause entities, and the edges represent the relationships between them (such as construction stages, spatial dependencies, etc.). GNN derives a deeper relationship between clauses and components through the propagation and update of information between nodes and edges.

[0142] The introduction of semantic reasoning greatly enhances intelligence and flexibility. By combining the relationships between different nodes in the graph, it is not only possible to match a single clause, but also to match compound rules involving multiple clauses, thereby generating a more comprehensive set of matching rules;

[0143] S3.4: If the matching rule set fails to reach the preset matching threshold during the semantic reasoning process, the feedback optimization mechanism will be activated to adjust the matching parameters. The specific steps include:

[0144] By analyzing the calculation results of semantic, spatial and temporal similarity, the weights are dynamically adjusted and the matching calculation is re-performed;

[0145] By using contextual relevance information, we can identify the reasons for matching failures and further optimize matching weights to ensure more accurate rule set generation.

[0146] The feedback optimization mechanism ensures that the system can dynamically respond to the review needs of complex scenarios. By gradually adjusting the matching parameters, the system can adapt to different building types and component environments, thereby generating more accurate review results.

[0147] Example 2

[0148] See also Figure 4 The present invention provides an embodiment: a BIM model review system based on text classification and named entity recognition, the system comprising:

[0149] Data acquisition module: used to obtain building code provisions and BIM models to be reviewed, pre-process the code provisions according to the review standards of the code provisions, and obtain the pre-processed code provisions;

[0150] Split and polish the regulatory articles. For regulatory articles that lack subjects after splitting, add corresponding subjects and adjust the word order to facilitate subsequent article classification and named entity extraction, and obtain preprocessed regulatory articles.

[0151] Clause classification module: used to classify the pre-processed specification clauses according to the review process of the specification clauses, and divide the pre-processed specification clauses into three types: clauses that stipulate the attribute values ​​of BIM model components, clauses that stipulate whether the BIM model contains a certain component, and clauses that stipulate the distance between two components, including:

[0152] The classification of articles is implemented by combining the Robust optimize BERT approach-Whole Word Masking (RoBERTa-wwm) with the Text Convolutional Neural Networks (TextCNN) to form an article classification dataset. The dataset is then used as the input of the RoBERTa-wwm+TextCNN model to establish a mapping relationship between articles and classification labels, thus realizing the input of the model to the standard articles and the output of the model to the corresponding classification labels.

[0153] Article keyword extraction module: used to extract keywords from the preprocessed specification articles. The extraction is based on whether it describes the building type, whether it describes the component type, whether it describes the attribute type, whether it describes the building space, whether it describes the comparative relationship, and whether it is a number or a unit.

[0154] The extraction of keywords in regulatory articles adopts named entity extraction of natural language processing, which is implemented by using a named entity extraction model (RoBERTa-wwm+BiLSTM+CRF) that combines a pre-trained model RoBERTa-wwm with a bidirectional long short-term memory network (Bi-directional Long Short-Term Memory, BiLSTM) and a conditional random field (CRF);

[0155] Review module: used to extract the properties of the BIM model components to be reviewed based on the clause classification results and the extracted keywords, and perform the review;

[0156] According to the classification results of the articles, different types of articles correspond to different review processes, and the review process is a set of pre-set review procedures;

[0157] Prepare the BIM model to be reviewed and pre-process the clauses to be reviewed into the form required by the system.

[0158] Input the clause to be reviewed "The width of residential windows shall not be less than 0.9m" into the model.

[0159] The text classification model (RoBERTa-wwm+TextCNN) is used to classify the incoming articles to be reviewed (such as the text of the building articles). The process and results are as follows: Figure 5 As shown, the classification result of the clause "not less than" is "attribute comparison type". The information required to perform the review of "attribute comparison type" clauses is: components to be checked, attributes and inspection rules. The classification result of the clause "residential windows" is "element existence type", and the classification result of the clause "0.9m" is "distance comparison type".

[0160] Use the entity extraction model (RoBERTa-wwm+BiLSTM+CRF) to extract named entities from the incoming articles to be reviewed. The named entity extraction process is as follows: Figure 6 As shown. The clause "The width of residential windows shall not be less than 0.9m" can be passed to the model through the API for named entity extraction, and the building type entity "residential", the building component entity "window", the attribute entity "width", the comparison relationship entity "not less than", the value entity "0.9", and the unit entity "m" are extracted.

[0161] The system executes the clause review process based on the extracted entities:

[0162] 1) If the building type entity is extracted, it is necessary to extract the building type information in the BIM model to be reviewed to check whether it is the extracted building type entity. If not, jump directly to step 2);

[0163] 2) Extracting the extracted component entities from the BIM model;

[0164] 3) Extract component attributes based on the identified attribute entities;

[0165] 4) Convert the extracted attribute values ​​according to the unit entity;

[0166] 5) Compare the numerical entity with the extracted attributes based on the comparison relationship.

[0167] 6) Record the results.

[0168] The system forms a final review list based on the above review results. The list includes review components, review attribute values, and review results.

[0169] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A BIM model review method based on text classification and named entity recognition, characterized in that: The method comprises: Acquire a specification text in the field of construction, construct a construction specification ontology according to the specification text in the field of construction, and construct a construction specification knowledge graph according to the construction specification ontology; According to the BIM model to be reviewed, the BIM model is processed by a review data extraction strategy to obtain a review data file and context relevance; Matching the review data file in the building specification knowledge graph to obtain a matching rule set; The priority of the matching rule set is adjusted according to the context relevance, and the BIM model to be reviewed is reviewed according to the adjusted matching rule set, and a review report is output.

2. A BIM model review method based on text classification and named entity recognition according to claim 1, characterized in that: The construction of the building specification knowledge graph includes: Determine the building specification ontology according to the standard text in the building field; According to the keywords and key terms of the domain determined by the building specification ontology, a building specification ontology knowledge model is established from top to bottom, wherein the building specification ontology knowledge model includes a top-level ontology, a concept subtree and a sub-layer instance; Add the construction field specification text as a knowledge feature to the concept subtree, perform a logical check on the construction specification ontology knowledge model, and determine whether it meets the construction field ontology construction principle. If not, continue to add concept subtrees; Constructing a knowledge extraction network, processing the construction field specification text corresponding to the concept subtree through the knowledge extraction network, extracting the construction specification sequence, and adding the construction specification sequence to the sublayer instance; Construct a building specification knowledge graph, instantiate the building specification ontology knowledge model, and add it to the building specification knowledge graph according to the entry strategy.

3. A BIM model review method based on text classification and named entity recognition according to claim 2, characterized in that: The knowledge extraction network comprises: The input layer maps the architectural field specification text corresponding to the concept subtree through a pre-trained word vector model, converts it into a word vector matrix, and calculates the knowledge feature vector according to the word vector matrix; A feature extraction layer extracts semantic features from the knowledge feature vector through a bidirectional long short-term memory neural network; The output layer calculates the label score matrix by scoring the semantic features, extracts the optimal sequence according to the label score matrix, decodes the optimal sequence, and calculates the building specification sequence.

4. A BIM model review method based on text classification and named entity recognition according to claim 3, characterized in that: The entry strategy includes: Obtain the number of concept branches, keywords and keyword attributes of the input instantiated building specification ontology knowledge model; Determine the semantic tree where the instantiated building specification ontology knowledge model is located in the building specification knowledge graph, traverse all nodes of the semantic tree, extract the keyword pairs with the smallest semantic distance between the nodes and the instantiated building specification ontology knowledge model as similar node pairs, and calculate the semantic similarity of the similar node pairs; The highest semantic similarity is compared with the similarity threshold. If it is less than the similarity threshold, the instantiated building specification ontology knowledge model is used as a sibling node of the node with the highest semantic similarity. If it is greater than or equal to the similarity threshold, the instantiated building specification ontology knowledge model is merged with the node with the highest semantic similarity.

5. The BIM model review method based on text classification and named entity recognition according to claim 1 is characterized in that: The data extraction strategy for the review includes: The BIM model to be reviewed is processed in layers according to the component type; According to the layered BIM model, component attribute information, spatial relationship data and time dimension data are extracted; Integrate the component attribute information, spatial relationship data and time dimension data into a review data file, wherein the review data file includes component geometry information, attribute information, spatial relationship table and time node data; The context information of the component and its surrounding associated components is modeled through the graph neural network model. According to the constructed context dependency model, the correlation between each component and its associated components is calculated. The correlation is dynamically updated through the message passing mechanism between nodes, and the correlation is weighted and summed to output the context correlation.

6. A BIM model review method based on text classification and named entity recognition according to claim 5, characterized in that: The matching in the building specification knowledge graph according to the review data file includes: According to the review data file, feature extraction is performed through a pre-trained model to generate a multi-dimensional feature vector; The multi-dimensional feature vector is compared with the specification clause vector in the building specification knowledge graph through a dynamic semantic matching algorithm, and a preliminary match is performed according to the multi-dimensional semantic matching strategy to obtain a candidate answer set; Performing semantic reasoning on the candidate answer set according to the relationship between nodes and edges in the building specification knowledge graph, and obtaining a matching rule set according to matching parameters, wherein the matching parameters include semantic similarity, spatial similarity, and temporal similarity; If the matching rule set does not satisfy the matching threshold, the matching parameters are adjusted through a feedback optimization mechanism until the matching rule set satisfies the matching threshold.

7. A BIM model review method based on text classification and named entity recognition according to claim 6, characterized in that: The multi-dimensional semantic matching strategy includes: Match the geometric properties of components in the BIM model with the geometric specifications of components in the building specification knowledge graph through geometric information features; Matching time node data with construction timing requirements in the building specification knowledge graph through time information features; The component attribute information is matched with the specification terms in the building specification knowledge graph through semantic information features.

Images