A method and system for semantic construction and automatic generation of a multi-industry information model

Abstract

This invention belongs to the field of cross-industry information model construction technology, and relates to a method and system for semantic construction and automatic generation of multi-industry information models. The method includes: generating standard format requirement parameters for cross-industry models based on cross-industry model construction requests; decomposing the standard format requirement parameters into industry requirement parameter sets; generating customized meta-models by combining the industry requirement parameter sets, a meta-model base library, and a cross-industry feature library; semantically aligning the customized meta-models based on an industry terminology list and a cross-industry semantic rule library to obtain a cross-industry model element semantic alignment table; constructing lightweight model structures based on the cross-industry model element semantic alignment table; performing syntax validation on each lightweight model structure; and generating standardized model files corresponding to each target industry. This invention achieves automated construction of cross-industry models from scratch, improves model adaptation success rate and semantic accuracy, and avoids the problem of cross-industry models being unable to communicate with each other.

Description

Technical Field

[0001] This invention relates to the field of cross-industry information model construction technology, and in particular to a semantic construction and automatic generation method and system for multi-industry information models. Background Technology

[0002] In the context of the current deep integration of digital transformation and industrial collaboration, building information models that can support cross-industry data interoperability and business integration has become a key technological foundation for realizing industrial system applications such as intelligent manufacturing and smart energy. However, there are three major pain points in the current construction of cross-industry information models within the same industrial project system.

[0003] First, the reusability of cross-industry meta-models is poor: each industry model within the system relies on its own meta-model, lacking a general extension framework. When adding a new domain model, the meta-model structure needs to be redesigned, resulting in a long development cycle within the system. Second, semantic alignment of model elements relies on manual intervention: the semantic mapping of cross-industry model elements within the system requires manually defined rules, leading to a large error rate and an inability to dynamically adapt to updates in industry terminology. Third, the efficiency of model structure assembly and verification is low: the model structure within the system requires manual drag-and-drop construction, and syntax verification relies on manual checks, making automated modeling impossible. Single model construction is time-consuming, and redundant attributes account for a large proportion, affecting the efficiency of subsequent data modeling. Summary of the Invention

[0004] Based on the above analysis, the embodiments of the present invention aim to provide a semantic construction and automatic generation method and system for multi-industry information models, in order to solve the problems of poor reusability of cross-industry meta-models, lack of automated modeling systems, reliance on manual semantic alignment, and low modeling efficiency in the existing technology.

[0005] On one hand, embodiments of the present invention provide a method for semantic construction and automatic generation of multi-industry information models, including:

[0006] Based on the cross-industry model building request, generate cross-industry standard format requirement parameters; decompose the standard format requirement parameters into industry requirement parameter sets corresponding to each target industry;

[0007] By combining the aforementioned set of industry demand parameters, a pre-set meta-model base library, and a cross-industry feature library, customized meta-models corresponding to each target industry are generated.

[0008] Based on a pre-set list of industry terms and a cross-industry semantic rule library, semantic alignment is performed on each of the customized meta-models to obtain a cross-industry model element semantic alignment table.

[0009] Based on the cross-industry model element semantic alignment table, construct a lightweight model structure corresponding to each target industry;

[0010] The lightweight model structures described are syntax-validated, and standardized model files corresponding to each target industry are generated.

[0011] Furthermore, the step of generating cross-industry standard format requirement parameters based on the cross-industry model construction request includes:

[0012] Based on the target industry, model coverage, core model elements, and model file output format specified by the user in the cross-industry model building request, standard format requirement parameters in JSON format are generated.

[0013] Furthermore, the meta-model base library includes at least: object identification rules, basic attribute sets, and general association templates;

[0014] The object identification rules are used to define a unified naming convention for model objects; the basic attribute set is used to define the common basic attributes of the model; and the general association template is used to define the general logical relationships between the attributes.

[0015] Furthermore, the step of combining the industry demand parameter set, the pre-set meta-model base library, and the cross-industry feature library to generate customized meta-models corresponding to each target industry includes:

[0016] Load the object identification rules, basic attribute sets, and general association templates corresponding to each target industry from the metamodel base library;

[0017] Based on the object identification rules, basic attribute set, and general association template, a general framework for the metamodel is generated.

[0018] Based on the set of industry demand parameters, a hash mapping algorithm is used to match the corresponding industry-specific feature set from the cross-industry feature library;

[0019] The industry-specific feature sets are dynamically added to the extended structure of the meta-model general framework to obtain the specific feature meta-models corresponding to each target industry.

[0020] Cross-industry consistency verification is performed on the exclusive feature meta-models corresponding to each target industry, and customized meta-models corresponding to each target industry are generated based on the results of the consistency verification.

[0021] Furthermore, the semantic alignment of each customized meta-model based on a pre-set list of industry terms and a cross-industry semantic rule base includes:

[0022] The list of industry terms is broken down into industry keywords and core terms corresponding to each target industry;

[0023] Based on the industry keywords, the statistical feature values ​​of the corresponding core words in the target industry are calculated, and based on the statistical feature values, the core word set corresponding to each target industry is selected.

[0024] Based on the cross-industry semantic rule base and the user-specified model core elements, the semantic similarity between core words expressing the same concept in the core word set of different target industries is calculated; the semantic similarity calculation results are corrected by preset weight coefficients, and the cross-industry model element semantic alignment table is obtained based on the corrected semantic similarity calculation results.

[0025] Furthermore, the semantic alignment table of the cross-industry model elements, obtained based on the corrected semantic similarity calculation results, includes:

[0026] Based on the corrected semantic similarity calculation results, the core words of each target industry with high similarity are mapped to the same unified standardized element in the customized meta-model, and the core words of each target industry with low similarity are mapped to different industry-specific elements in the customized meta-model.

[0027] Based on the unified standardized elements and the industry-specific elements, the semantic alignment table of the cross-industry model elements is obtained.

[0028] Furthermore, the step of constructing a lightweight model structure corresponding to each target industry based on the cross-industry model element semantic alignment table includes:

[0029] Calculate the correlation between each alignment element in the cross-industry model element semantic alignment table and each target scene in the model coverage specified by the user;

[0030] Based on the correlation calculation results and the user-specified model core elements, a set of core elements is obtained by filtering from the semantic alignment table;

[0031] Based on the core element set and the preset rule base, a lightweight model structure corresponding to each target industry is constructed.

[0032] Furthermore, the step of performing syntax validation on each of the lightweight model structures and generating standardized model files corresponding to each target industry includes:

[0033] The lightweight model structure is syntax-validated based on the LSTM (Long Short-Term Memory) network model, and standardized model files corresponding to each target industry are generated according to the validation results and the model file output format specified by the user.

[0034] On the other hand, embodiments of the present invention provide a semantic construction and automatic generation system for multi-industry information models, including: a front-end request construction module and a back-end model construction module;

[0035] The front-end request building module is used to build requests based on cross-industry models and generate cross-industry standard format requirement parameters;

[0036] The backend model building module specifically includes:

[0037] The requirement parsing unit is used to decompose the standard format requirement parameters into a set of industry requirement parameters corresponding to each target industry.

[0038] The dynamic adaptation unit is used to combine the industry requirement parameter set, the pre-set meta-model base library and the cross-industry feature library to generate customized meta-models corresponding to each target industry.

[0039] The semantic alignment unit is used to perform semantic alignment on each of the customized meta-models based on a pre-set list of industry terms and a cross-industry semantic rule library, so as to obtain a cross-industry model element semantic alignment table.

[0040] The lightweight assembly unit is used to construct a lightweight model structure corresponding to each target industry based on the cross-industry model element semantic alignment table.

[0041] The verification and generation unit is used to perform syntax verification on each of the lightweight model structures and generate standardized model files corresponding to each target industry.

[0042] Furthermore, the dynamic adaptation unit is also used for:

[0043] Load the object identification rules, basic attribute sets, and general association templates corresponding to each target industry from the metamodel base library;

[0044] Based on the object identification rules, basic attribute set, and general association template, a general framework for the metamodel is generated.

[0045] Based on the set of industry demand parameters, a hash mapping algorithm is used to match the corresponding industry-specific feature set from the cross-industry feature library;

[0046] The industry-specific feature sets are dynamically added to the extended structure of the meta-model general framework to obtain the specific feature meta-models corresponding to each target industry.

[0047] Cross-industry consistency verification is performed on the exclusive feature meta-models corresponding to each target industry, and customized meta-models corresponding to each target industry are generated based on the results of the consistency verification.

[0048] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0049] First, unlike related technologies where cross-industry meta-models have poor reusability and cannot achieve automated modeling, this invention adopts a dynamic combination mode of a general meta-model base library and a cross-industry feature library. Without redesigning the meta-model structure, it achieves automated construction of a cross-industry adaptable meta-model framework from scratch, solves the problem of non-reusability of industry-specific meta-models, and improves the success rate of model adaptation.

[0050] Secondly, unlike related technologies where semantic alignment of model elements relies on manual intervention, this invention achieves automatic unification of terminology across multiple industries by employing industry-specific word segmentation, core word similarity matching, and industry-weighted ambiguity resolution. This facilitates cross-industry model interoperability without the need for manual definition of mapping rules, thereby improving the accuracy of semantic alignment of cross-industry elements and avoiding the problem of cross-industry models being unable to interoperate due to "different meanings of the same word".

[0051] Third, unlike the low efficiency of model structure assembly in related technologies, this invention takes the user-specified model coverage as the core and eliminates redundant attributes through correlation calculation, which can accurately screen industry needs and improve the elimination rate of redundant attributes. Furthermore, driven by needs, it realizes the automated assembly of "non-redundant" and "non-general" model structures, reduces the amount of data for single model construction, improves the efficiency of subsequent data modeling, and shortens the construction cycle of multi-industry information models.

[0052] It is worth noting that this invention is mainly applicable to the fully automated construction of cross-industry adaptation, semantic definition, lightweight structural assembly, and intelligent syntax verification of information models in multiple fields such as equipment manufacturing (e.g., CNC machine tools), urban construction (e.g., building energy consumption), and green and low-carbon (e.g., photovoltaic power stations). It mainly solves the problems of long development cycles, inconsistent semantics, and redundant structures of cross-industry models, without involving any data collection, data transfer, or interactive functions.

[0053] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0054] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0055] Figure 1 This is a flowchart illustrating the semantic construction and automatic generation method for multi-industry information models according to an embodiment of the present invention.

[0056] Figure 2 This is a schematic diagram of the main modules of the semantic construction and automatic generation system for multi-industry information models according to an embodiment of the present invention. Detailed Implementation

[0057] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0058] A specific embodiment of the present invention discloses a method for semantic construction and automatic generation of multi-industry information models, such as... Figure 1 As shown, the steps S1 to S5 are as follows:

[0059] Step S1: Generate cross-industry standard format requirement parameters based on the cross-industry model construction request; decompose the standard format requirement parameters into industry requirement parameter sets corresponding to each target industry.

[0060] During implementation, standard format requirement parameters are generated across industries, including: generating standard format requirement parameters in JSON format based on the target industry, model coverage, core model elements, and model file output format specified by the user in the cross-industry model building request.

[0061] It should be noted that the standard format requirement parameters are the internal unified format when the backend processes user requirements. Regardless of the output format of the model file subsequently selected by the user, the internal format of the standard format requirement parameters is fixed as JSON, thereby ensuring that the parameter format is uniform and unambiguous.

[0062] Specifically, a cross-industry model building request consists of input parameters submitted by the user to the front-end model building module. The back-end model building module then performs industry-specific requirement parsing to obtain standard-format requirement parameters, thus clarifying the user's cross-industry model building goals. Along with generating the standard-format requirement parameters, the back-end module also returns a model building task ID for subsequent progress tracking and result association. For example, when a user initiates a cross-industry model building request on the front-end, they send the user-specified input parameters to the system interface ` / api / internal / model / build / cross-industry`. After generating the standard-format requirement parameters, the back-end module returns a cross-industry model building task ID, such as `TASK-CROSS-IND-202509`.

[0063] It is understandable that user input parameters are strictly limited to the model building dimension. In other words, users can only input information related to model building, and will not input any data-related content (such as data collection frequency, data format, data threshold, etc.). Its essence is to clarify the model building goals and output requirements.

[0064] In some implementations, the user's operation process for requesting the build module on the front end specifically includes:

[0065] (1) Select the target industry: such as "equipment manufacturing" and "green and low-carbon";

[0066] (2) Fill in the model coverage (i.e., application scenario): such as "CNC machine tool operation monitoring" and "photovoltaic power station energy consumption statistics" and other specific scenarios; at the same time, the model coverage must match the selected industry;

[0067] (3) Specify the core elements of the model: Select / input by industry, such as "spindle speed" and "machining accuracy" for equipment manufacturing, and "photovoltaic conversion rate" for green and low-carbon industries;

[0068] (4) Select the output format of the model file: such as JSON (a lightweight data exchange format) or XML (an extensible markup language) format, the default is JSON format.

[0069] The model coverage is used to clarify the specific business problem or service scenario that the model is to solve. It is the target boundary of model construction and can be used as a screening criterion to ensure that the model only contains elements that are strongly related to the target scenario and eliminates irrelevant and redundant attributes. For example, the model coverage of the equipment manufacturing industry includes: CNC machine tool operation monitoring, that is, the model only serves the goal of monitoring the operating status of CNC machine tools; the model coverage of the green and low-carbon industry includes: photovoltaic power station energy consumption statistics, that is, the model only serves the goal of statistical photovoltaic power station energy consumption.

[0070] The core elements of the model are one of the key inputs for subsequent steps such as requirements parsing, semantic alignment, and model assembly, and they are present throughout the model building process. In the requirements parsing stage, after the user-input parameters are converted into standard format requirements parameters, the user-specified core elements will be extracted, which can serve as the core constraints for model building; for example, core elements such as "spindle speed" and "photovoltaic panel conversion rate" will be marked as mandatory elements.

[0071] Furthermore, the user's front-end operations are mapped to an automatic conversion logic of standardized JSON structure, specifically including the following operation process:

[0072] (1) When the user selects “Equipment Manufacturing” and “Green and Low Carbon” in the front-end visualization interface, the parameters will be mapped to the “target_industries” field in JSON format and automatically filled with the corresponding values. This field is used to specify the target industry that needs to be built into the model. It is a required field with the constraint that it must contain at least one industry. Cross-industry input is supported. At the same time, the selected target industry must be in the system’s preset industry library.

[0073] (2) The model coverage scope input by the user will be directly written into the “model_scopes” field in the JSON format to verify the matching of the target industry and the scope. For example, “photovoltaic power station” must belong to the “green and low-carbon” industry, otherwise an error will be prompted. Its constraint is to correspond one-to-one with the “target_industries” field. For example, “photovoltaic power station” cannot correspond to the “equipment manufacturing” industry.

[0074] (3) The core elements of the model input by the user are sorted by industry and generated into an array of "core_elements" objects in JSON format. That is, each industry contains the core elements of the model to ensure that the field structure is standardized. This array of objects is an optional parameter. If the user does not fill it in, the core elements of the industry will be automatically matched from the cross-industry feature library. If the user fills it in, it must be ensured that it belongs to the corresponding industry. For example, "spindle speed" belongs to "equipment manufacturing".

[0075] (4) The output format of the model file input by the user corresponds to the string “output_format”, which only supports the two standardized model file output formats, JSON and XML.

[0076] Furthermore, decomposing the standard format requirement parameters into industry requirement parameter sets corresponding to each target industry means breaking down the overall cross-industry requirement into independent requirements for each individual industry. When operating on the front end, users may simultaneously select multiple target industries and their corresponding cross-industry requirements. For example, if a user selects "equipment manufacturing" and "green and low-carbon" as target industries, and the selected requirement is "CNC machine tool operation monitoring + photovoltaic power station energy consumption statistics," then these requirements will be converted into standard format requirement parameters in JSON format containing information from multiple industries. The decomposition process involves breaking down the overall cross-industry requirement into individual industry-specific requirement parameter packages according to each target industry selected by the user, facilitating subsequent industry-specific model building.

[0077] After decomposition, the set of industry demand parameters for a single target industry includes two parts:

[0078] The first part describes the application scenarios of the model in the target industry, such as "CNC machine tool operation monitoring" for the equipment manufacturing industry and "photovoltaic power station energy consumption statistics" for the green and low-carbon industry.

[0079] The second part is a list of core elements for the target industry, such as "spindle speed and machining accuracy" for the equipment manufacturing industry, and "photovoltaic panel conversion rate" for the green and low-carbon industry.

[0080] For example, if a user selects "equipment manufacturing + green and low-carbon" as the target industry, the model coverage is "CNC machine tool operation monitoring + photovoltaic power station energy consumption statistics". The core elements of the model are "spindle speed + machining accuracy + photovoltaic panel conversion rate", and thus form a standard format requirement parameter in JSON format. After decomposing the standard format requirement parameter, two specific industry requirements are obtained: (1) Equipment manufacturing industry: scenario = CNC machine tool operation monitoring, core elements = spindle speed, machining accuracy; (2) Green and low-carbon industry: scenario = photovoltaic power station energy consumption statistics, core elements = photovoltaic panel conversion rate, thus forming a set of industry requirement parameters for each industry.

[0081] Step S2: Combine the industry demand parameter set, the pre-set meta-model base library, and the cross-industry feature library to generate customized meta-models for each target industry.

[0082] It is understandable that the process of generating a customized meta-model does not involve data collection and exchange. The industry requirement parameter set, meta-model base library, and cross-industry feature library are all static preset data, rather than dynamic business data.

[0083] During implementation, the meta-model base library includes at least: object identification rules, basic attribute sets, and general association templates. The object identification rules define a unified naming convention for model objects, such as "IND-Industry-Type-001" to identify a unique model object. The basic attribute set defines general basic attributes of the model, excluding industry-specific attributes, such as model name and ID, creation time, and industry. The general association template defines general logical relationships between attributes, such as triggering rule B if attribute A exceeds a threshold, but does not include industry-specific association rules, such as overload warnings if the spindle speed exceeds 2000 r / min, which is characteristic of the equipment manufacturing industry.

[0084] Generating customized meta-models for each target industry specifically includes: loading the object identification rules, basic attribute sets, and general association templates corresponding to each target industry from the meta-model base library; generating a general meta-model framework based on the object identification rules, basic attribute sets, and general association templates; matching corresponding industry-specific feature sets from the cross-industry feature library based on the industry requirement parameter set using a hash mapping algorithm; dynamically adding the industry-specific feature sets to the extended structure of the general meta-model framework to obtain the exclusive feature meta-models for each target industry; performing cross-industry consistency verification on the exclusive feature meta-models for each target industry, and generating customized meta-models for each target industry based on the consistency verification results.

[0085] In addition, the process will also output a cross-industry adaptation consistency report. For example, the equipment manufacturing meta-model contains the "spindle speed" attribute, and the green low-carbon meta-model contains the "conversion rate" attribute. The cross-industry adaptation consistency report (text / JSON format) shows that the adaptation success rate is ≥99.5%, displays the anomalies that failed the verification (such as incompatible attribute units), and provides automatic correction suggestions.

[0086] Specifically, the customized metamodel is implemented based on a cross-industry metamodel dynamic adaptation algorithm. The algorithm aims to automatically generate customized metamodels for multiple industries based on the system's general metamodel library and industry feature library, without requiring redesign of the model structure. The specific process includes the following:

[0087] (1) Loading of general infrastructure: The system internal library calls the object identification rules (such as "IND-industry-type-001"), basic attribute sets (such as "name", "ID", "creation time"), and general association templates (such as "attribute A > threshold → trigger rule B") from the metamodel base library, and obtains the general framework of the metamodel; that is, the general framework is a blank framework without industry characteristics;

[0088] (2) Dynamic injection of industry features: The “industry-feature” hash mapping algorithm is adopted. The hash mapping algorithm is a function that can map input data of arbitrary length to a fixed-length seemingly random output value through a specific mathematical transformation.

[0089] Applying this algorithm to this embodiment allows for the matching of specific attributes based on the target industry. For example, the equipment manufacturing industry matches attributes such as "spindle diameter" and "machining accuracy." These attributes are automatically embedded into the expandable attribute slots of the general framework of the meta-model, avoiding attribute conflicts. For instance, if both industries contain the "power" attribute, the unified semantic is "equipment rated power." This yields a specific feature meta-model for each target industry. For example, the specific feature meta-model for equipment manufacturing includes attributes such as "spindle speed" and "machining accuracy," while the specific feature meta-model for green and low-carbon industries includes attributes such as "photovoltaic panel conversion rate" and "inverter efficiency." This solves the problem of the general framework lacking industry-specific characteristics.

[0090] (3) Cross-industry consistency verification: Check the compatibility between the injected industry features and the general structure. For example, the attribute type of "photovoltaic panel conversion rate" is "float(0-100)", which conforms to the numerical type specification of the general meta-model. Generate an adaptation consistency report and automatically prompt correction for inconsistencies.

[0091] Therefore, based on the customized meta-model, the prototype of object-attribute-association rules for various industry models is defined, gradually realizing the integration of "general adaptation and industry characteristics". This can provide a basic framework file for subsequent steps such as semantic alignment and model assembly, and lay the foundation for cross-industry model construction.

[0092] Step S3: Based on the pre-set list of industry terms and the cross-industry semantic rule library, perform semantic alignment on each of the customized meta-models to obtain a cross-industry model element semantic alignment table.

[0093] During implementation, semantic alignment is performed on each of the customized meta-models, specifically including:

[0094] The list of industry terms is broken down into industry keywords and core terms corresponding to each target industry;

[0095] Based on the industry keywords, the statistical feature values ​​of the corresponding core words in the target industry are calculated, and based on the statistical feature values, the core word set corresponding to each target industry is selected.

[0096] Based on the cross-industry semantic rule base and the user-specified model core elements, the semantic similarity between core words expressing the same concept in the core word set of different target industries is calculated; the semantic similarity calculation results are corrected by preset weight coefficients, and the cross-industry model element semantic alignment table is obtained based on the corrected semantic similarity calculation results.

[0097] For example, the core term for "motor power" in the equipment manufacturing industry and "photovoltaic panel power" in the green and low-carbon industry are both "power" (a cross-industry core term). In this case, the semantic similarity between the two "power" terms is calculated.

[0098] Furthermore, obtaining the cross-industry model element semantic alignment table based on the corrected semantic similarity calculation results includes: mapping all target industry core words with high similarity to the same unified standardized element in the customized meta-model based on the corrected semantic similarity calculation results, and mapping all target industry core words with low similarity to different industry-specific elements in the customized meta-model; obtaining the cross-industry model element semantic alignment table based on the unified standardized element and the industry-specific element; and generating an alignment accuracy report for subsequent analysis.

[0099] In some implementations, the generation process of the cross-industry model element semantic alignment table is based on a cross-industry model element semantic alignment algorithm. The algorithm aims to automatically align the original business terms from multiple industries within the system into unified model elements, thereby solving the problem of cross-industry semantic heterogeneity. Specifically, the process includes the following steps:

[0100] First, the terminology feature extraction preprocessing stage includes:

[0101] (1) Industry-specific word segmentation: Call the corresponding industry word segmentation library and split the list of industry terms into industry keywords and core words corresponding to each target industry; for example, the equipment manufacturing word segmentation library contains "spindle" and "motor", and green and low carbon contains "photovoltaic panel" and "inverter". The "motor power" in the industry term list is split into "equipment manufacturing + power", and "photovoltaic panel power" is split into "green and low carbon + power".

[0102] (2) TF-IDF statistical characteristic value calculation: Quantify the importance of each core word in industry terminology. For example, if the TF-IDF value of "power" is high in both types of industry terminology, it indicates that the core word is a core common word.

[0103] Second, the core word similarity matching stage includes:

[0104] (1) Based on the cross-industry semantic rule base (pre-stored terminology-unified element mapping relationship) and the user-specified model core elements (base words for cross-industry semantic matching), the semantic similarity of core words of different industry terms is calculated through a similarity matching algorithm.

[0105] (2) Matching logic: If the similarity of core words is not less than the threshold (e.g., similarity ≥ 95%) and the industry keywords do not conflict, then the core words of these two different industries will be mapped to the same unified standardized element in the customized meta-model; for example, the semantic similarity between the core word "power" of "motor power" and "photovoltaic panel power" is 100%, then the two "power" will be mapped to the same unified standardized element "equipment rated power".

[0106] Third, the post-ambiguity resolution processing stage includes: if the same term matches multiple unified standardized elements, such as "efficiency" which can be mapped to "equipment efficiency" and "system efficiency", then the unique mapping can be determined by combining the model coverage and industry relevance weights; for example, in the green and low-carbon industry, "photovoltaic power plant efficiency" corresponds to "photovoltaic panel conversion efficiency".

[0107] Preferably, the present invention employs a semantic similarity algorithm that combines semantic vectors with industry weights to calculate the semantic similarity between core terms across industries, specifically including:

[0108] (1) Semantic vector construction: Cross-industry core words (such as "power" and "speed") are mapped to a pre-trained industry semantic vector space, and the semantic vector of the core words is obtained after mapping. It can be understood that this semantic vector space is obtained by training professional texts of multiple industries such as water affairs, green and low carbon, and equipment manufacturing through the Word2Vec model. The construction and training method of the Word2Vec model can be referred to the existing technology, which will not be elaborated here.

[0109] (2) Similarity calculation: Calculate the similarity between the semantic vectors of two cross-industry core words (e.g., using the cosine similarity algorithm). The similarity value ranges from [0,1], and the closer it is to 1, the more consistent the semantics are.

[0110] (3) Industry weight correction: If the semantics of core words differ slightly in different industries, a preset industry weight coefficient (such as industry relevance weight of 0.7 and semantic similarity weight of 0.3) is introduced, and the similarity results are corrected in combination with the model coverage to avoid mapping errors caused by "different meanings of the same word"; for example, "load" in equipment manufacturing refers to "equipment load", while in ERP enterprise resource planning system it refers to "material weight". The semantics of the two are different. The semantic vector similarity of "load" in equipment manufacturing and ERP system is 0.8. The corrected similarity = 0.8 × 0.3 + 0 (low industry relevance) = 0.24. At this time, mapping is not triggered and the exclusive definition is retained.

[0111] (4) Matching threshold: Set a similarity threshold of ≥0.9. If the similarity is lower than the threshold, it is determined that the cross-industry core term is a non-common term and retains the industry-specific definition. Otherwise, it is mapped to the same unified standardized element.

[0112] Furthermore, the cross-industry model element semantic alignment table, obtained based on unified standardized elements and industry-specific elements, can record the mapping relationship between the original terms and model elements of each industry. For example, in the equipment manufacturing industry, the original term "motor power" corresponds to the unified standardized element "equipment rated power," which is semantically defined as "the rated power of the equipment during normal operation, in kW." In the green and low-carbon industry, "photovoltaic panel power" corresponds to the unified standardized element "equipment rated power," which is semantically defined as "the rated power of the equipment during normal operation, in kW." The "spindle speed" exclusive element in the equipment manufacturing industry and the "conversion rate" exclusive element in the green and low-carbon industry retain their exclusive definitions of the original terms in their respective industries.

[0113] Step S4: Based on the cross-industry model element semantic alignment table, construct the lightweight model structure corresponding to each target industry.

[0114] In practice, this includes: calculating the correlation between each alignment element in the cross-industry model element semantic alignment table and each target scene in the model coverage area specified by the user;

[0115] Based on the correlation calculation results and the user-specified model core elements, a set of core elements is obtained by filtering from the semantic alignment table; at the same time, a redundancy removal report is generated.

[0116] Based on the core element set and the preset rule base, a lightweight model structure corresponding to each target industry is constructed.

[0117] In some implementations, lightweight model structures are built based on lightweight assembly algorithms. The goal of these algorithms is to automatically assemble “redundant” model structures based on semantic alignment results and industry requirements, eliminating attributes that are irrelevant to the requirements.

[0118] During lightweight assembly, it is necessary to calculate the correlation between each aligned element in the cross-industry model element semantic alignment table and each target scene, specifically including two correlation calculation methods:

[0119] First, algorithms based on cosine similarity include:

[0120] (1) Demand correlation calculation: The attribute-demand cosine similarity algorithm is used to calculate the correlation between each aligned element and the target scenario in the model coverage. For example, the correlation between "spindle speed" and the "operation monitoring" scenario is 0.92, and the correlation between "spindle speed" and the "procurement date" scenario is 0.15.

[0121] (2) Redundant attribute removal: Automatically delete attributes with a correlation degree below the threshold, retain core elements and form a core element set; for example, in the "energy consumption statistics" of the green and low-carbon industry, retain "daily power generation and conversion rate" and remove "photovoltaic panel production date", etc.

[0122] (3) Embedding of association rules: Call matching rules from the preset rule library of meta-model association in the system (such as "spindle speed > 2000 r / min → trigger overload warning" in the equipment manufacturing industry) and embed them into the model structure to form a complete link of object-attribute-association.

[0123] Second, the industry-weighted semantic-demand relevance algorithm, which incorporates "industry feature weights" and "demand scenario weights" into the cosine similarity algorithm, including:

[0124] (1) Semantic similarity: The semantic vector similarity between each aligned element and each scenario covered by the model is calculated based on the industry semantic vector library. For example, the semantic similarity between the "spindle speed" and "operation monitoring" scenarios is 0.92.

[0125] (2) Industry feature matching degree: Quantify whether the aligned element is a core feature of the industry; for example, "spindle speed" is a core feature of the equipment manufacturing industry, and its matching degree is 1.0; "purchase date" is a non-core feature, and its matching degree is 0.1;

[0126] (3) Demand scenario fit: Based on the preset industry scenario rule library, determine whether the core element set directly serves the model coverage; for example, the "operation monitoring" scenario requires core elements such as "spindle speed" and "operation status", and its fit is 1.0; "purchase date" does not serve this scenario, and its fit is 0.05.

[0127] Therefore, the aforementioned industry-weighted semantic-demand relevance algorithm not only combines industry characteristics to avoid non-core elements being misjudged as relevant, but also better fits the demand scenario, ensuring that the core elements retained directly serve the model objective. At the same time, it uses the model coverage as the screening criterion for "lightweight assembly," ensuring that the model only contains elements strongly related to the target scenario and eliminates irrelevant and redundant attributes.

[0128] The following will take the "Equipment Manufacturing - CNC Machine Tool Operation Monitoring" scenario as an example to explain in detail the process of calculating the correlation degree of each element through the industry weighted semantic-demand correlation degree algorithm;

[0129] (1) The specific parameters required for the calculation are as follows:

[0130] Alignment elements: spindle speed, machining accuracy, purchase date, and operating status;

[0131] Model coverage: CNC machine tool operation monitoring;

[0132] Industry Feature Library: The core features of equipment manufacturing are "spindle speed, machining accuracy, and operating status", while the non-core feature is "purchase date".

[0133] Scene rule base: The "Running Monitoring" scene requires real-time status elements, but not static attribute elements.

[0134] (2) The calculation logic is as follows: Relevance = (Semantic Similarity × 0.4) + (Industry Feature Matching Degree × 0.3) + (Demand Scenario Fit × 0.3); The calculation results are as follows:

[0135] The correlation coefficient of "spindle speed" = 0.92 × 0.4 + 1.0 × 0.3 + 1.0 × 0.3 = 0.968;

[0136] The correlation coefficient of "machining accuracy" = 0.88 × 0.4 + 1.0 × 0.3 + 0.95 × 0.3 = 0.937;

[0137] The correlation of "purchase date" = 0.15 × 0.4 + 0.1 × 0.3 + 0.05 × 0.3 = 0.105;

[0138] The correlation of "operational status" = 0.95 × 0.4 + 1.0 × 0.3 + 1.0 × 0.3 = 0.98;

[0139] (3) Screening results: The redundancy screening threshold is set to 0.3. Then, the "purchase date" with a correlation of less than 0.3 is removed, and "spindle speed", "machining accuracy" and "operating status" are retained as the core element set, thereby realizing the model lightweighting.

[0140] It should be noted that the semantic similarity algorithm used in the semantic alignment stage of this invention has the same essential computational logic as the correlation algorithm used in the lightweight assembly stage, but they target different computational objects. The semantic alignment stage is used to calculate the similarity between the original core words of different industries to achieve semantic alignment and unified mapping of cross-industry terms; the lightweight assembly stage is used to calculate the similarity between the aligned model elements and specific business requirement scenarios to drive the accurate assembly of the lightweight model.

[0141] Step S5: Perform syntax verification on each of the lightweight model structures and generate standardized model files corresponding to each target industry.

[0142] Implementation specifically includes: performing syntax validation on the lightweight model structure based on the LSTM (Long Short-Term Memory) network model, thereby obtaining a complete model structure that passes validation without attribute type errors and with closed-loop related logic, and generating a syntax validation report containing suggestions for correcting anomalies or those that are not present. Then, based on the validation results and the user-specified model file output format (such as JSON or XML), standardized model files corresponding to each target industry are generated. The training and construction methods of the LSTM network model can refer to existing technologies and will not be elaborated here.

[0143] Furthermore, during the model archiving and delivery phase, standardized model files for different industries, download links within the system, and model archiving records (which can be stored in a cross-industry model library) are delivered to users, completing the closed loop from requirement input to model output. It is understandable that the construction process for standardized model files is consistent across different industries, but their content varies, which will not be elaborated upon here.

[0144] To further illustrate the semantic model construction and automatic generation method of the present invention, a specific embodiment is provided below.

[0145] Taking the simultaneous construction of equipment manufacturing (CNC machine tools) and green and low-carbon (photovoltaic power station) models as an example, the specific implementation process is as follows:

[0146] (1) Users input cross-industry requirements in the front-end request building module: the target industries are "equipment manufacturing" and "green and low-carbon", and the model coverage is "CNC machine tool operation monitoring" and "photovoltaic power station energy consumption statistics", respectively. The core elements of the model include "spindle speed" and "photovoltaic panel conversion rate".

[0147] (2) Cross-industry meta-model adaptation: The meta-model general framework is loaded through the backend model building module. Based on the cross-industry feature library, the "spindle type" and "processing accuracy" attributes are injected into the "equipment manufacturing" industry, and the "photovoltaic panel material" and "inverter efficiency" attributes are injected into the "green and low-carbon" industry. The adaptation success rate is 100%.

[0148] (3) Semantic alignment of elements: The core terms with high similarity in the industry terminology list, such as “CNC machine tool motor power” and “photovoltaic power station inverter power”, are aligned to “equipment rated power”, resulting in customized meta-models for the two industries with an alignment accuracy of 99.5%.

[0149] (4) Lightweight assembly: In the customized meta-model of equipment manufacturing, the alignment element with low correlation, "purchase date", is removed, and alignment elements such as "spindle speed" and "operating status" are retained to form the core element set of the industry; in the customized meta-model of green and low carbon, the alignment element with low correlation, "photovoltaic panel model", is removed, and alignment elements such as "conversion rate" and "daily power generation" are retained to form the core element set of the industry; then, based on the core element set and the preset rule base, the lightweight model structure corresponding to the two target industries is constructed.

[0150] (5) Syntax validation: The LSTM syntax model is used to check the structure of the two lightweight models above. There are no attribute type errors. Two standardized model files in JSON format are generated respectively, such as "Equipment Manufacturing - CNC Machine Tool Model.json" and "Green Low Carbon - Photovoltaic Power Station Model.json".

[0151] (6) Model file delivery: After the front-end requests the construction module to download, the above two standardized model files can be directly used for subsequent data modeling in the system, such as CNC machine tool operation data modeling, photovoltaic power station energy consumption analysis modeling, etc.

[0152] It is understood that the above implementation methods are only for ease of understanding and simplification of description, and should not be construed as limitations on the present invention. The present invention does not impose specific limitations on the type of target industry, meta-model adaptation, semantic alignment, lightweight assembly, and syntax verification.

[0153] Therefore, it can be seen that the embodiments of the present invention can achieve one of the following beneficial effects:

[0154] First, it improves the efficiency of cross-industry model development: it shortens the parallel construction cycle of multi-industry models, eliminates the need for manual intervention, reduces modeling costs, and solves the pain point of having to redesign new industry models.

[0155] Second, ensuring semantic consistency of the model: improves the accuracy of semantic alignment of elements across industries and avoids the problem of cross-industry models being unable to communicate due to "different meanings of the same word";

[0156] Third, the model's lightweightness has been optimized: the rate of redundant attribute removal has been improved, the amount of data in a single model has been reduced, and the efficiency of subsequent data modeling has been improved.

[0157] Fourth, the technology boundaries are clear: it only covers the construction of cross-industry models "from scratch", without involving any data collection, data flow or interaction functions within the system, which reduces the technical complexity and is more conducive to the automated construction of multi-domain information models, while providing support for subsequent data modeling of the system.

[0158] In another embodiment of the present invention, a semantic construction and automatic generation system for multi-industry information models is proposed, such as... Figure 2 As shown, it specifically includes the following modules: front-end request building module and back-end model building module;

[0159] It is understandable that the front-end request building module, as the user interaction entry point, is responsible for receiving the input of requirements and the receipt of results. Its specific functions include receiving the model building requirements submitted by users (such as target industry, model scope and core elements, etc.), displaying the model building progress to users in real time (such as "metamodel adaptation in progress" "semantic alignment completed", etc.), and providing a download link for the standardized model file after the model is generated. It does not participate in any substantial model building logic operations.

[0160] The backend model building module serves as the core of the entire model building process, undertaking the core role of requirement processing and model generation. Its specific functions include parsing the standard format requirement parameters input from the front end, sequentially executing steps such as metamodel adaptation, cross-industry semantic alignment, lightweight structure assembly, and syntax verification, and finally generating a usable standardized model file for archiving and management. At the same time, it pushes a file download link to the front end. All its processing flows form a closed loop within the system and are completed independently without communicating with external data modules, ensuring the security and consistency of the building process.

[0161] The front-end request building module is used to build requests based on cross-industry models and generate cross-industry standard format requirement parameters;

[0162] The backend model building module specifically includes:

[0163] The requirement parsing unit is used to decompose the standard format requirement parameters into a set of industry requirement parameters corresponding to each target industry.

[0164] The dynamic adaptation unit is used to combine the industry requirement parameter set, the pre-set meta-model base library and the cross-industry feature library to generate customized meta-models corresponding to each target industry.

[0165] The semantic alignment unit is used to perform semantic alignment on each of the customized meta-models based on a pre-set list of industry terms and a cross-industry semantic rule library, so as to obtain a cross-industry model element semantic alignment table.

[0166] The lightweight assembly unit is used to construct a lightweight model structure corresponding to each target industry based on the cross-industry model element semantic alignment table.

[0167] The verification and generation unit is used to perform syntax verification on each of the lightweight model structures and generate standardized model files corresponding to each target industry.

[0168] Furthermore, the dynamic adaptation unit is also used for:

[0169] Load the object identification rules, basic attribute sets, and general association templates corresponding to each target industry from the metamodel base library;

[0170] Based on the object identification rules, basic attribute set, and general association template, a general framework for the metamodel is generated.

[0171] Based on the set of industry demand parameters, a hash mapping algorithm is used to match the corresponding industry-specific feature set from the cross-industry feature library;

[0172] The industry-specific feature sets are dynamically added to the extended structure of the meta-model general framework to obtain the specific feature meta-models corresponding to each target industry.

[0173] Cross-industry consistency verification is performed on the exclusive feature meta-models corresponding to each target industry, and customized meta-models corresponding to each target industry are generated based on the results of the consistency verification.

[0174] The above-described method and system embodiments are based on the same principles, and their related aspects can be referenced from each other to achieve the same technical effects. For specific implementation processes, please refer to the foregoing embodiments, which will not be repeated here.

[0175] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0176] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for semantic construction and automatic generation of multi-industry information models, characterized in that, include: Based on the cross-industry model building request, generate cross-industry standard format requirement parameters; The standard format requirement parameters are broken down into industry requirement parameter sets corresponding to each target industry; By combining the aforementioned set of industry demand parameters, a pre-set meta-model base library, and a cross-industry feature library, customized meta-models corresponding to each target industry are generated. Based on a pre-set list of industry terms and a cross-industry semantic rule library, semantic alignment is performed on each of the customized meta-models to obtain a cross-industry model element semantic alignment table. Based on the cross-industry model element semantic alignment table, construct a lightweight model structure corresponding to each target industry; The lightweight model structures described are syntax-validated, and standardized model files corresponding to each target industry are generated.

2. The method according to claim 1, characterized in that, The step of generating cross-industry standard format requirement parameters based on the cross-industry model construction request includes: Based on the target industry, model coverage, core model elements, and model file output format specified by the user in the cross-industry model building request, standard format requirement parameters in JSON format are generated.

3. The method according to claim 2, characterized in that, The metamodel base library includes at least: object identification rules, basic attribute sets, and general association templates; The object identification rules are used to define a unified naming convention for model objects; the basic attribute set is used to define the common basic attributes of the model; and the general association template is used to define the general logical relationships between the attributes.

4. The method according to claim 3, characterized in that, The process of combining the industry demand parameter set, the pre-set meta-model base library, and the cross-industry feature library to generate customized meta-models for each target industry includes: Load the object identification rules, basic attribute sets, and general association templates corresponding to each target industry from the metamodel base library; Based on the object identification rules, basic attribute set, and general association template, a general framework for the metamodel is generated. Based on the set of industry demand parameters, a hash mapping algorithm is used to match the corresponding industry-specific feature set from the cross-industry feature library; The industry-specific feature sets are dynamically added to the extended structure of the meta-model general framework to obtain the specific feature meta-models corresponding to each target industry. Cross-industry consistency verification is performed on the exclusive feature meta-models corresponding to each target industry, and customized meta-models corresponding to each target industry are generated based on the results of the consistency verification.

5. The method according to claim 1, characterized in that, The semantic alignment of each customized meta-model based on a pre-set list of industry terms and a cross-industry semantic rule library includes: The list of industry terms is broken down into industry keywords and core terms corresponding to each target industry; Based on the industry keywords, the statistical feature values ​​of the corresponding core words in the target industry are calculated, and based on the statistical feature values, the core word set corresponding to each target industry is selected. Based on the cross-industry semantic rule base and the user-specified model core elements, the semantic similarity between core words expressing the same concept in the core word set of different target industries is calculated; the semantic similarity calculation results are corrected by preset weight coefficients, and the cross-industry model element semantic alignment table is obtained based on the corrected semantic similarity calculation results.

6. The method according to claim 5, characterized in that, The semantic similarity calculation results based on the corrected semantic similarity yield the cross-industry model element semantic alignment table, which includes: Based on the corrected semantic similarity calculation results, the core words of each target industry with high similarity are mapped to the same unified standardized element in the customized meta-model, and the core words of each target industry with low similarity are mapped to different industry-specific elements in the customized meta-model. Based on the unified standardized elements and the industry-specific elements, the semantic alignment table of the cross-industry model elements is obtained.

7. The method according to claim 1, characterized in that, The step of constructing a lightweight model structure corresponding to each target industry based on the cross-industry model element semantic alignment table includes: Calculate the correlation between each alignment element in the cross-industry model element semantic alignment table and each target scene in the model coverage specified by the user; Based on the correlation calculation results and the user-specified model core elements, a set of core elements is obtained by filtering from the semantic alignment table; Based on the core element set and the preset rule base, a lightweight model structure corresponding to each target industry is constructed.

8. The method according to claim 1, characterized in that, The step of performing syntax validation on each of the lightweight model structures and generating standardized model files corresponding to each target industry includes: The lightweight model structure is syntax-validated based on the LSTM (Long Short-Term Memory) network model, and standardized model files corresponding to each target industry are generated according to the validation results and the model file output format specified by the user.

9. A semantic construction and automatic generation system for multi-industry information models, characterized in that, include: Front-end request building module and back-end model building module; The front-end request building module is used to build requests based on cross-industry models and generate cross-industry standard format requirement parameters; The backend model building module specifically includes: The requirement parsing unit is used to decompose the standard format requirement parameters into a set of industry requirement parameters corresponding to each target industry. The dynamic adaptation unit is used to combine the industry requirement parameter set, the pre-set meta-model base library and the cross-industry feature library to generate customized meta-models corresponding to each target industry. The semantic alignment unit is used to perform semantic alignment on each of the customized meta-models based on a pre-set list of industry terms and a cross-industry semantic rule library, so as to obtain a cross-industry model element semantic alignment table. The lightweight assembly unit is used to construct a lightweight model structure corresponding to each target industry based on the cross-industry model element semantic alignment table. The verification and generation unit is used to perform syntax verification on each of the lightweight model structures and generate standardized model files corresponding to each target industry.

10. The system according to claim 9, characterized in that, The dynamic adaptation unit is also used for: Load the object identification rules, basic attribute sets, and general association templates corresponding to each target industry from the metamodel base library; Based on the object identification rules, basic attribute set, and general association template, a general framework for the metamodel is generated. Based on the set of industry demand parameters, a hash mapping algorithm is used to match the corresponding industry-specific feature set from the cross-industry feature library; The industry-specific feature sets are dynamically added to the extended structure of the meta-model general framework to obtain the specific feature meta-models corresponding to each target industry. Cross-industry consistency verification is performed on the exclusive feature meta-models corresponding to each target industry, and customized meta-models corresponding to each target industry are generated based on the results of the consistency verification.

Images