Methods, systems, equipment, and media for generating scenario maps of industry digital transformation

By collecting and preprocessing multi-source heterogeneous data, and using pre-trained language models and preset graph skeletons to collaboratively generate dynamic ontology, the problem of low graph construction efficiency and difficulty in updating in industry digital transformation scenarios is solved. This achieves efficient, standardized and intelligent graph generation, and improves the standardization and application interaction capabilities of the graph.

CN123070262APending Publication Date: 2026-11-10INSPUR YUNZHOU (SHANDONG) IND INTERNET CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611577896.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-13
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies in industry digital transformation scenarios suffer from problems such as high communication costs, high map construction costs, long iteration cycles, low standardization, poor reusability, difficulty in updating, limited application forms, and insufficient user interaction capabilities, making it difficult to achieve efficient, dynamic, and intelligent map generation.

Method used

By collecting and preprocessing multi-source heterogeneous data, and utilizing pre-trained language models in collaboration with preset graph skeletons, structured output and dynamic ontology generation are achieved. Combined with graph database modeling, efficient, standardized, and intelligent generation of scene graphs is realized, and incremental updates and visual interaction are supported.

Benefits of technology

It improves the efficiency, standardization, and intelligent application capabilities of map construction, realizing the automated, standardized, dynamic, and intelligent generation of maps, and supporting incremental updates of maps and deep integration with business systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123070262A_ABST
    Figure CN123070262A_ABST
Patent Text Reader

Abstract

This invention provides a method, system, device, and medium for generating industry digital transformation scenario maps, belonging to the field of artificial intelligence technology. The method includes: preprocessing multi-source heterogeneous data to obtain standardized data; extracting four types of digital elements—industry, supply chain links, business activities, and transformation scenarios—using a pre-trained language model based on a structured output template aligned with a preset map skeleton, and obtaining scenario map association data after verification; performing fill-in-the-blank ontology generation using the preset map skeleton as a template, and obtaining an industry digital transformation scenario map through knowledge fusion and map modeling; prioritizing scenarios and writing them into corresponding scenario nodes; updating the map according to update trigger conditions; and visually rendering and outputting the map. This invention achieves automated, standardized, and dynamic generation of maps, improving map construction efficiency, data quality, and intelligent application capabilities.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, and more specifically relates to a method, system, device and medium for generating scenario maps of industry digital transformation. Background Technology

[0002] In key industry digital transformation practices, digital transformation service providers and manufacturing enterprises face high communication costs and difficulties in reaching consensus regarding understanding needs, aligning solutions, and planning pathways. The root cause lies in the barrier between industrial and digital languages, with both sides lacking a common, structured communication platform. Currently, the industry primarily relies on expert experience to manually construct digital transformation scenario maps. However, this traditional approach heavily depends on domain experts manually labeling entity relationships and business rules, resulting in high construction costs, lengthy iteration cycles, and inconsistent understanding and classification standards among different experts. Consequently, the maps suffer from low standardization and poor reusability, making it difficult to form a unified and standardized industry digital transformation scenario map.

[0003] In recent years, large language models have demonstrated capabilities in knowledge extraction and structured generation. However, existing technologies largely focus on building general knowledge graphs, lacking dedicated generation frameworks for industry-specific digital transformation scenarios. The graph content generated by general methods is disconnected from actual industry needs, with inconsistent output formats and incomplete fields, making it difficult to directly use for standardized expressions in industry scenarios. Scenario division and element association heavily rely on expert experience, resulting in inconsistent standards and difficulty in reuse. Once built, the graph is difficult to dynamically adjust and expand according to different industries, leading to high update and maintenance costs. Furthermore, existing technologies fail to form a closed loop with post-generation user interaction, secondary editing, and scenario-based solutions, hindering the continuous evolution and intelligent application of the generated graph.

[0004] In terms of scenario priority assessment, scenario map updates, and application output, existing technologies mostly rely on subjective scoring or simple weighting, making it difficult to objectively reflect implementation difficulty, expected benefits, and risk levels. Scenario map updates often employ full reconstruction, making updates difficult once new data emerges, thus hindering timeliness. Furthermore, scenario map applications are limited in form, making deep integration with business systems difficult, and user interaction and intelligent question-answering capabilities are insufficient. These issues result in low efficiency in constructing scenario maps for current industry digital transformation, difficulty in dynamic adjustment, a lack of collaborative mechanisms, and difficulties in implementing theoretical methods. Therefore, there is an urgent need for an intelligent scenario map generation method adapted to the characteristics of industry digital transformation. Summary of the Invention

[0005] To address the above issues, the present invention aims to provide a method, system, device, and medium for generating industry digital transformation scenario maps. By leveraging a map generation and dynamic maintenance mechanism driven by a large model and an industry map skeleton, the invention achieves efficient, standardized, dynamic, and intelligent generation of industry digital transformation scenario maps, thereby improving map construction efficiency, standardization, and intelligent application capabilities.

[0006] To achieve the above objectives, the present invention employs the following technical solution: In a first aspect, embodiments of this application provide a method for generating an industry digital transformation scenario map, comprising the following steps: S1: Collect multi-source heterogeneous data, including document data, table data, web page and interface data, and sensor data; preprocess the multi-source heterogeneous data to obtain standardized data; the preprocessing includes parsing, cleaning, deduplication, and entity disambiguation. S2: Based on the structured output template aligned with the preset map skeleton, the pre-trained language model is used to extract four types of digital elements from the standardized data: industry, industrial chain links, business activities, transformation scenarios, and so on. Hierarchical annotation, priority annotation, and relationship identification are performed, and the scene map association data is obtained after verification. S3: Using the preset graph skeleton as a template, perform fill-in-the-blank ontology generation to obtain a dynamic ontology. Based on the dynamic ontology, perform knowledge fusion and graph modeling on the scene graph association data, store it in the graph database, and obtain the industry digital transformation scene graph. S4: Prioritize the scenarios in the industry digital transformation scenario map, and write the obtained scenario priority as an attribute into the corresponding scenario node in the graph database; S5: Update the industry digital transformation scenario map using the newly added multi-source heterogeneous data according to the update trigger conditions; S6: Visualize and render the industry digital transformation scenario map and / or output the application.

[0007] In an optional implementation, step S1 includes: The following parsing process is performed on the multi-source heterogeneous data to generate parsing result data: Extract the text layer from document-type data. When the text layer is missing or garbled, use optical character recognition (OCR) to identify the page image. Parse table-type data row by row to obtain records corresponding to fields and values; Data from web pages and APIs can be captured by using the Hypertext Transfer Protocol interface or a data collection program based on fixed fields. The sensor data is parsed according to the sampling protocol or by timestamp and device identifier to obtain the time-stamped records, and the time-stamped records are aligned by timestamp. The parsed data is uniformly converted into UTF-8 text, and three types of metadata tags are added: source, type, and collection time. Clean and normalize the UTF-8 text, remove headers, footers, table of contents, blank lines and duplicate paragraphs, and normalize full-width / half-width and simplified / traditional Chinese characters to obtain normalized text; The content fingerprint of the normalized text is calculated for each paragraph, and only one fingerprint with the same content is retained to obtain the deduplicated text. For concepts with the same meaning but different expressions in the deduplicated text, synonym merging is first performed using a semantic dictionary; for concepts not covered by the semantic dictionary, the cosine similarity between two vectors is calculated using the following formula:

[0008] when Not lower than the preset similarity threshold When they are identified as the same entity, they are merged while retaining their respective sources, resulting in disambiguated text. The required fields and naming conventions of the disambiguation text are checked according to the rules, and samples are randomly selected according to a preset ratio. The pass rate is calculated using the following formula. :

[0009] in, This represents the number of samples that passed the random inspection. The total number of samples to be inspected; when When the pass rate falls below the preset threshold, the entire batch is returned for reprocessing; when When the disambiguation text is greater than or equal to the preset pass rate threshold, it will be used as standardized data.

[0010] In an optional implementation, step S2 includes: A predefined JavaScript object representation output template aligned with a preset graph skeleton; the preset graph skeleton includes one industry level and five types of industry chain links. The text describes a series of business activities and their corresponding transformation scenarios, along with four types of digital elements. It also mentions the JavaScript object representation output template, which includes fields for industry, supply chain links, business activities, transformation scenarios, four types of digital elements, hierarchy, and priority. The four types of digital elements include data elements, knowledge models, software tools, and talent skills. Input the standardized data and a preset number of labeled examples into the pre-trained language model, and fill the template item by item according to the JavaScript object representation method through the pre-trained language model to output the structured extraction result; The structured extraction results identify transformation scenarios, which are then divided into four levels according to granularity: first level (industry level), second level (domain level), third level (enterprise level), and fourth level (job level). The identified transformation scenarios are further categorized into main scenarios and sub-scenarios. Sub-scenarios are further decompositions of the main scenarios, inheriting the four types of digital elements from their respective main scenarios. The resulting levels are written into the level field of the corresponding transformation scenario in the structured extraction results. The level annotations are used to construct a level index. For the main scene in the structured extraction result, data elements, knowledge models, software tools, and talent skills are labeled in the four categories of digital element fields of the main scene; For the main scenario and sub-scenario in the structured extraction result, mark them as basic essential level, core key level or forward-looking innovation level in the priority field of the structured extraction result according to the preset transformation guidance; Identify the collaborative relationships across industry chain links and the integration relationships across business activities from the structured extraction results, and write the collaborative relationships and the integration relationships into the association relationship field of the structured extraction results as association relationships to be mapped as graph edges; The labeled structured extraction results are validated and disambiguated, including: checking the validity of field types and enumeration values; checking the completeness of hierarchical relationships, i.e., the fourth, third, and second levels belong to their direct parent level, and filling or marking missing hierarchical relationships; when multiple expressions of the same entity appear, the frequency of each expression in the structured extraction results is counted, and the expression with the highest frequency is selected as the expression of the entity; when the same relationship is contradictory, the number of sources of each relationship expression in the structured extraction results is counted, and the relationship expression with the most sources is selected as the expression of the relationship; obtaining the generation probability output by the pre-trained language model when generating each structured extraction result, and using the generation probability as the confidence level of the structured extraction result, those below the preset confidence threshold of 0.7 are transferred to the manual review queue; The structured extraction results after verification and disambiguation constitute the scene map association data.

[0011] In an optional implementation, step S3 includes: The entity types and relationship types defined in the preset graph skeleton are obtained, and specific instances and attribute specifications are filled into the preset graph skeleton according to the target industry to obtain a dynamic ontology. The entity types include industry, industrial chain links, business activities, transformation scenarios and four types of digital elements. The relationship types include attribution, inclusion, association, decomposition, cross-link collaboration and cross-business collaboration. Construct a graph database and align the node and edge types of the graph database with the fields of the JavaScript object representation output template; In the graph database, node types and edge types are defined according to the dynamic ontology. The main scene and sub-scenes are connected by decomposition relationship. The four types of digital elements are mounted as attributes on the main scene nodes. The sub-scenes reuse the four types of digital elements of the main scene. Based on the dynamic ontology, entities in the scene graph association data are mapped to graph nodes in the graph database and node identifiers are generated; relationships in the scene graph association data are mapped to graph edges in the graph database and edge indexes are generated. Calculate the cosine similarity between the vector of the newly extracted entity and the entity vector in the existing graph. When the cosine similarity is not lower than the preset fusion similarity threshold, merge the newly extracted entity and the existing entity into the same node to reduce redundant nodes in the graph database. For missing relation edges in the graph database, the number of historical co-occurrences of the entities at both ends of the relation edge in the existing graph is counted, and the ratio of the number of historical co-occurrences to the total number of occurrences of the entities at both ends is calculated. When the ratio is not lower than a preset co-occurrence probability threshold, the relation edge is completed between the entities at both ends, and the inferred source is marked on the completed relation edge. When the same attribute of the same entity has multiple values, the frequency of each value in the scene map association data is counted, and the value with the highest frequency is selected as the final value of the attribute; when multiple values ​​have the same frequency, the value with the latest update time is selected as the final value of the attribute. Based on the node and edge types defined in the dynamic ontology, the mapped graph nodes and edges, the merged nodes, and the completed relational edges are organized to obtain the industry digital transformation scenario graph. The industry digital transformation scenario map is stored in the graph database, where nodes represent entities and have type, name, and source attributes, and edges represent relationships and have relationship type and confidence. The graph database supports industry namespace isolation and version number management, and exports resource description framework format and / or JavaScript object representation link data format.

[0012] In an optional implementation, step S4 includes: The system reads the number of node attributes for each scene node from the graph database and calculates the digital element maturity based on this number. It also reads the number of associated edges and the layer depth for each scene node in the graph database and calculates the implementation difficulty based on these dimensions. Furthermore, it reads historical revenue statistics associated with each scene node and calculates expected revenue based on these statistics. Finally, it reads historical risk statistics associated with each scene node and calculates the implementation risk based on these statistics. The system then normalizes the digital element maturity, implementation difficulty, expected revenue, and implementation risk to a standardized format. arrive ; The maturity level of the digital elements is calculated using the following formula:

[0013] in, For the first The maturity of digital elements at each scenario node , , , The first The values ​​of data element attributes, knowledge model attributes, tool software attributes, and talent skill attributes for each scene node are taken when they exist. If it does not exist, take ; The implementation difficulty is calculated using the following formula:

[0014] in, For the first The implementation difficulty of each scenario node For the first The number of associated edges for each scene node in the graph database. This is the average number of associated edges for all scene nodes in the graph database. For the first The hierarchical depth of each scene node This is the maximum value of the hierarchy depth of all scene nodes in the graph database; The expected return is calculated using the following formula:

[0015] in, For the first Expected revenue for each scenario node For the first The number of historical revenue statistics associated with each scenario node For the first Historical earnings statistics; The implementation risk is calculated using the following formula:

[0016] in, For the first Implementation risks at each scenario node For the first The number of historical risk statistics associated with each scenario node For the first One historical risk statistical value; The calculated maturity level of the digital elements, the implementation difficulty, the expected benefits, and the implementation risks are normalized to the following formulas: arrive :

[0017] in, For the first The unnormalized index values ​​of each scene node This represents the set of values ​​for this metric for all scene nodes. It is the minimum value in the set of values. The maximum value in the set of values. The index value is the normalized value; The analytic hierarchy process (AHP) is used to conduct pairwise comparisons of the maturity, implementation difficulty, expected benefits, and implementation risks of normalized digital elements, constructing a judgment matrix. And calculate the weight vector Perform a consistency check; when the consistency ratio... Less than the preset consistency threshold When determining whether the weight allocation is reasonable, the following applies:

[0018] The judgment matrix The largest eigenvalue, The judgment matrix The order of The preset average random consistency index; According to the weight vector The normalized digital element maturity, implementation difficulty, expected benefits, and implementation risks are weighted and summed to obtain a comprehensive score for each scenario:

[0019] in, For the first The overall score of each scene node. , , , The weight vectors are respectively The weighting of digital elements maturity, implementation difficulty, expected benefits, and implementation risks in China, and meeting the following requirements. , , , , The first The maturity, implementation difficulty, expected benefits, and implementation risks of digital elements after normalization of each scenario node; All scenarios are sorted according to their overall scores to form a priority list, and the overall score and priority of each scenario are written as attributes into the corresponding scenario node in the graph database.

[0020] In an optional implementation, step S5 includes: Set timed triggers and / or event triggers as update trigger conditions, where the event triggers include new data being added to the database; When the update triggering condition is met, newly added multi-source heterogeneous data is collected, and the preprocessing is performed on the newly added multi-source heterogeneous data to obtain newly added standardized data. Based on the structured output template and the pre-trained language model, industry, industrial chain links, business activities, transformation scenarios and four types of digital elements are extracted from the newly added standardized data to obtain newly added scenario map association data; Based on the dynamic ontology, knowledge fusion and graph modeling are performed on the newly added scene graph association data to generate incremental nodes and incremental relationships. The incremental nodes and incremental relationships are merged into the industry digital transformation scenario graph, and only the newly added or updated parts are written. The indexes and adjacency tables of the graph database are updated, and a new version number is assigned.

[0021] In an optional implementation, step S6 includes: The system performs visualization rendering of the industry digital transformation scenario map. The front end renders the topology map and / or hierarchical map according to the hierarchical index of the graph database. It receives user editing operations on nodes and edges and writes them back to the graph database. It receives scenario retrieval, element query, relationship link tracing and multidimensional statistics requests and returns the corresponding results from the industry digital transformation scenario map based on the adjacency list of the graph database. The application output of the industry digital transformation scenario map is carried out. According to the template, the industry digital transformation scenario map matrix, digital element list and scenario priority list are extracted from the industry digital transformation scenario map and exported as a report. The map data in resource description framework format and / or JavaScript object representation link data format is output to the enterprise digital platform, industrial Internet platform and / or business system through the application programming interface gateway in a standard interface. The system receives user questions through a question-and-answer service. The service performs semantic parsing on the user questions to obtain search criteria. Based on the search criteria, it retrieves relevant entities, relationships, and original text fragments from the industry digital transformation scenario map and the standardized data. It then generates a natural language answer based on the relevant entities, relationships, and original text fragments, and labels the source of the content in the natural language answer.

[0022] Secondly, embodiments of this application also provide an industry digital transformation scenario map generation system, including: The data acquisition and preprocessing module is used to acquire multi-source heterogeneous data, including document data, table data, web page and interface data, and sensor data; and to preprocess the multi-source heterogeneous data to obtain standardized data; the preprocessing includes parsing, cleaning, deduplication, and entity disambiguation. The extraction module is used to extract four types of digital elements from the standardized data based on the structured output template aligned with the preset map skeleton, namely industry, industrial chain links, business activities, transformation scenarios, and others, using a pre-trained language model. It also performs hierarchical labeling, priority labeling, and relationship identification, and obtains scene map association data after verification. The graph generation module is used to perform fill-in-the-blank ontology generation using the preset graph skeleton as a template to obtain a dynamic ontology, and to perform knowledge fusion and graph modeling on the scene graph association data based on the dynamic ontology, and store it in the graph database to obtain the industry digital transformation scene graph. The priority evaluation module is used to evaluate the priority of scenarios in the industry digital transformation scenario map and write the evaluated scenario priority as an attribute into the corresponding scenario node in the graph database. The update module is used to update the industry digital transformation scenario map using newly added multi-source heterogeneous data according to the update trigger conditions; The output module is used to visualize and render the industry digital transformation scenario map and / or output it for application.

[0023] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the industry digital transformation scenario mapping generation method described in any of the above.

[0024] Fourthly, embodiments of this application also provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the industry digital transformation scenario mapping generation method as described in any of the above.

[0025] As can be seen from the above technical solutions, the present invention has the following advantages: The industry digital transformation scenario map generation method provided in this application achieves automated, standardized, dynamic, and intelligent generation of industry digital transformation scenario maps through unified collection and preprocessing of multi-source heterogeneous data, intelligent extraction based on preset map skeleton and structured output template, fill-in-the-blank dynamic ontology generation and graph database modeling, quantitative evaluation of scenario priorities and attribute write-back, incremental updates, and visual interaction and intelligent application output. This improves map construction efficiency, data quality, structural consistency, evaluation objectivity, and application interaction capabilities, providing a unified, standardized, and reusable structured carrier for industry digital transformation.

[0026] This application parses document-type data, table-type data, web page and interface-type data, and sensor-type data separately. Through cleaning and normalization, content fingerprint deduplication, semantic dictionary merging, vector similarity disambiguation, mandatory field checking, and sampling pass rate control, data from different sources and in different formats are uniformly converted into standardized data. This ensures the integrity of industry data sources and the consistency of formats, improves data traceability and usability, and provides a high-quality data foundation for graph construction.

[0027] This application constrains the generation process of the pre-trained language model by using a structured output template aligned with a predefined graph skeleton. This allows the model to fill in fields such as industry, supply chain links, business activities, transformation scenarios, four types of digital elements, hierarchy, and priority item by item. Combined with hierarchy division, decomposition of main scenarios and sub-scenarios, annotation of the four types of digital elements, priority marking, relationship identification, verification and disambiguation, and confidence verification, the graph extraction process is automated. This ensures that the output content is formatted uniformly, fields are complete, and semantics are consistent, thereby improving the standardization and reusability of graph construction.

[0028] This application uses a pre-defined graph skeleton as a template to perform fill-in-the-blank ontology generation, resulting in a dynamic ontology. Node and edge types are defined in the graph database. The main scene and sub-scenes are connected by decomposing relationships. Four types of digital elements are attached to the main scene nodes. Node merging is performed using entity vector cosine similarity, relationship completion is performed based on historical co-occurrence statistics, and attribute conflict resolution is performed based on occurrence frequency and update time. It also supports industry namespace isolation, version number management, and export of resource description framework formats and JavaScript object representation linked data formats, improving the consistency, scalability, and interoperability of the graph.

[0029] This application objectively reads the number of node attributes, the number of associated edges, the level depth, historical revenue statistics, and historical risk statistics of scene nodes from a graph database. It calculates the maturity, implementation difficulty, expected revenue, and implementation risk of digital elements according to the corresponding formulas and performs normalization processing. Then, it constructs a judgment matrix, calculates the weight vector, and performs consistency checks through the analytic hierarchy process. Finally, it calculates the weighted sum to obtain the comprehensive score of each scene and writes it back to the corresponding scene node. This improves the objectivity, interpretability, and decision support capability of scene priority assessment. At the same time, the assessment results can be used for priority construction of graph database indexes or sorting and filtering when querying graphs.

[0030] This application collects newly added multi-source heterogeneous data through timed triggering and / or event triggering. After preprocessing, extraction, knowledge fusion, and graph modeling, incremental nodes and incremental relationships are generated. Only the newly added or updated parts are written and assigned a new version number, realizing incremental updates and version management of the graph. At the same time, it supports front-end rendering of topology graphs and / or hierarchical graphs by level, secondary editing of nodes and edges and writing back to the graph database, scene retrieval, feature query, relationship link tracing, list report export, application programming interface gateway standard output, and question-and-answer service that generates natural language answers based on search results and marks the source. It forms a closed loop of graph construction, updating, interaction and application, improving the timeliness, usability and integration capability of the graph with enterprise digital platforms, industrial internet platforms and business systems. Attached Figure Description

[0031] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 A flowchart illustrating the method for generating industry digital transformation scenario maps provided in this application.

[0033] Figure 2 A schematic diagram of the structure of the industry digital transformation scenario mapping generation system provided in this application.

[0034] Figure 3 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0035] The various embodiments of this disclosure will be described more fully in the detailed steps of the method for generating industry digital transformation scenario maps described below. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0036] In the following, the terms “comprising” or “may include”, which may be used in various embodiments of this disclosure, indicate the presence of the disclosed functions, operations, or elements, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in various embodiments of this disclosure, the terms “comprising,” “having,” and their cognates are intended only to indicate a particular feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or the possibility of adding one or more combinations of the foregoing.

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Please see Figure 1 The diagram shows a flowchart of a method for generating an industry digital transformation scenario map in a specific embodiment. The method includes the following steps: S1: Collect multi-source heterogeneous data, including document data, table data, web page and interface data, and sensor data; preprocess the multi-source heterogeneous data to obtain standardized data; the preprocessing includes parsing, cleaning, deduplication, and entity disambiguation.

[0039] In a specific implementation, this step addresses the problem of how to transform data from different sources and in different formats into a unified and usable input.

[0040] First, data access and parsing are performed. For document-type data such as standard documents, research reports, and technical manuals, the text layer is extracted first; if the text layer is missing or garbled, optical character recognition is performed on the page images as a fallback to ensure the text can be processed subsequently. For example, an industry standard document is a scanned PDF, and garbled characters are obtained when the text layer is directly extracted. At this time, optical character recognition is performed on each page image of the PDF to obtain readable text. For table-type data, it is parsed row by row into records corresponding to fields and values. For example, after parsing a certain enterprise digital investment statistics table, records such as the enterprise name being a certain company and the investment amount being 5 million yuan are obtained. For web page and interface-type data, fixed fields are captured through hypertext transfer protocol interfaces or collection programs. For example, fields such as title, release date, and text are captured from the release page of a certain website. For sensor-type data, it is parsed into records with time identifiers according to the sampling protocol or by timestamp and device identifier. For example, a certain factory equipment sensor reports data such as temperature and vibration at the second level, and after parsing, records such as the device identifier being Device A, the timestamp being 10:00:00 on January 1, 2024, and the temperature being 65 degrees Celsius are obtained. The above parsed result data is uniformly converted into UTF-8 text, and three types of metadata tags including source, type, and collection time are added to facilitate subsequent traceability.

[0041] Next, text cleaning, deduplication and entity disambiguation are carried out. During cleaning, headers, footers, tables of contents, blank lines and duplicate paragraphs are deleted according to rules, and normalization is performed on full-width and half-width characters, as well as simplified and traditional Chinese characters. For example, the full-width characters MES are normalized to the half-width MES, and the traditional Chinese "製造執行系統" is normalized to the simplified Chinese manufacturing execution system. During deduplication, each piece of text is normalized first, then the content fingerprint is calculated, and only one entry is retained for the same fingerprint. For example, the same document is repeatedly published on different websites, and the content fingerprints are the same after normalization, so only one entry is retained. During entity disambiguation, for concepts with the same meaning but different expressions, synonym merging is first performed using a semantic dictionary. For those not covered, the cosine similarity of each expression vector is calculated. When the similarity is not lower than the threshold of 0.85, they are determined to be the same entity and merged, and their respective sources are retained at the same time. The cosine similarity is calculated according to the following formula:

[0042] For example, MES and manufacturing execution system are already covered in the semantic dictionary and are directly merged; the industrial internet platform and the industrial internet operating system are not covered by the dictionary, and the vector similarity calculation result is 0.88, so they are determined to be the same entity and merged. Finally, mandatory fields and naming specifications are checked according to the rules, and a fixed proportion of samples are manually sampled for inspection to calculate the pass rate :

[0043] where, is the number of samples that pass the sampling inspection, is the total number of sampled samples; when When the pass rate falls below the preset threshold, the entire batch is returned for reprocessing; when When the pass rate is greater than or equal to the preset pass rate threshold, the disambiguated text is used as standardized data. Standardized data is obtained through the above parsing, cleaning, deduplication, and entity disambiguation processes.

[0044] S2: Based on the structured output template aligned with the preset map skeleton, the pre-trained language model is used to extract four types of digital elements from the standardized data: industry, industrial chain links, business activities, transformation scenarios, and so on. Hierarchical annotation, priority annotation, and relationship identification are performed, and the scene map association data is obtained after verification.

[0045] In this specific implementation, this step transforms the preprocessed data into structured candidate results for industry, supply chain links, business activities, transformation scenarios, and four categories of digital elements. The core is to constrain the generation process with a defined output format, preventing the model from operating freely. Specifically, a JavaScript object representation output template aligned with a predefined graph skeleton is defined. Fields cover industry, supply chain links, business activities, transformation scenarios, four categories of digital elements, hierarchy, and priority. The four categories of digital elements include data elements, knowledge models, software tools, and talent skills. The predefined graph skeleton includes one industry level, five supply chain links, and... Each business activity and its corresponding transformation scenario, and four types of digital elements. The input is a positive integer. Standardized data and a small number of labeled examples are fed into a pre-trained language model, which is required to strictly fill in the template item by item. The resulting output is structured data with a controllable format, rather than natural language that is difficult to verify. Any pre-trained language model that can fill in the structured output template item by item and output the generation probability can be selected, such as BERT, GPT series, LLaMA, Qwen, ChatGLM, etc. For example, given the input of information about a smart factory in a certain industry, the model can fill in the template and output: industry is discrete manufacturing, supply chain link is production and manufacturing, business activity is equipment management, transformation scenario is predictive maintenance of equipment, data element is equipment operation data, knowledge model is fault prediction model, software tool is predictive maintenance system, talent skill is data analysis skill, level is third-level enterprise level, and priority is core and critical level.

[0046] Based on this, the extracted content includes three parts. First, scenario identification: transformation scenarios are identified from the structured extraction results, categorized into four levels of granularity from coarse to fine: Level 1 (industry-level), Level 2 (domain-level), Level 3 (enterprise-level), and Level 4 (job-level). Scenarios are further divided into main scenarios and sub-scenarios. Sub-scenarios are further decompositions of the main scenarios, also participating in level and priority labeling, but without separate labeling of the four types of digital elements; their elements are inherited from their respective main scenarios. For example, equipment management is the main scenario, while equipment status monitoring and equipment fault diagnosis are its sub-scenarios, both inheriting the four types of digital elements from the main scenario. Second, four-element labeling: each main scenario is labeled with its involved data elements, knowledge models, software tools, and talent skills. Third, priority labeling: scenarios are marked as basic mandatory, core critical, or forward-looking innovation levels based on pre-defined transformation guidelines. In addition, the relationships between scenarios are identified, especially collaborative relationships across industry chain links and integrated relationships across business activities, and these are written into the relationship field of the structured extraction results as relational edges in the graph.

[0047] Subsequently, to ensure the results are controllable and traceable, post-processing and verification are required after generation. The output is parsed, and the validity of field types and enumerated values ​​is checked item by item; the hierarchical relationships are checked for completeness, i.e., the fourth, third, and second levels belong to their respective direct superior levels, and missing ones are filled in or marked; when multiple expressions of the same entity appear, the frequency of each expression in the structured extraction results is counted, and the one with the highest frequency is selected; when contradictions exist in the same relationship, the number of sources for each relationship expression is counted, and the one with the most sources is selected; the generation probability output by the pre-trained language model when generating each structured extraction result is obtained and used as the confidence level of that result; results below the threshold of 0.7 are transferred to the manual review queue.

[0048] Ultimately, the structured extraction results after verification and disambiguation constitute the scene map association data.

[0049] S3: Using the preset graph skeleton as a template, perform fill-in-the-blank ontology generation to obtain a dynamic ontology. Based on the dynamic ontology, perform knowledge fusion and graph modeling on the scene graph association data, store it in the graph database, and obtain the industry digital transformation scene graph.

[0050] In a specific implementation, this step organizes the extracted results into a standardized graph. The ontology is not generated from scratch each time, but rather through a fill-in-the-blank generation using a preset graph skeleton as a template: the entity types and relationship types defined in the preset graph skeleton are obtained, and specific instances and attribute specifications are filled in according to the target industry to obtain a dynamic ontology. Entity types include industry, supply chain links, business activities, transformation scenarios, and four types of digital elements. Relationship types include attribution, inclusion, association, decomposition, cross-link collaboration, and cross-business collaboration. For example, if a certain industry's graph skeleton defines a four-layer structure of industry, supply chain links, business activities, and transformation scenarios, filling in specific instances of discrete manufacturing in the target industry yields the dynamic ontology for that industry. Scene entities are divided into two categories: main scenes and sub-scenes, connected through decomposition relationships. The four types of digital elements are only attached to the main scene nodes as attributes, while sub-scenes reuse higher-level elements, thus forming a two-level scene structure of main scenes and sub-scenes. Cross-link collaboration depicts the connection and cooperation between scenes in different supply chain links, while cross-business integration depicts the connection and cooperation between scenes in different business activities; together, they constitute the collaborative relationship network of the graph.

[0051] Next, a graph database is constructed, aligning its node and edge types with the fields of the JavaScript object representation output template. Node and edge types are defined in the graph database according to a dynamic ontology. The main scene and sub-scenes are connected by decomposing relationships, and four types of digital elements are attached as attributes to the main scene nodes. Based on the dynamic ontology, entities in the scene graph association data are mapped to graph nodes in the graph database, generating node identifiers; relationships are mapped to graph edges, generating edge indexes. For example, discrete manufacturing is mapped to industry nodes, production manufacturing to supply chain link nodes, equipment management to main scene nodes, and equipment status monitoring to sub-scene nodes, connected by decomposing relationship edges.

[0052] In the knowledge fusion stage, the cosine similarity between the vector of a newly extracted entity and the entity vectors in the existing graph is calculated. When the similarity is not lower than a preset fusion similarity threshold, the newly extracted entity and the existing entity are merged into the same node to reduce redundant nodes in the graph database. For example, the newly extracted device predictive maintenance vector has a similarity of 0.92 with the device predictive maintenance vector in the existing graph, so it is determined to be the same node and merged. For missing relationship edges in the graph database, the historical co-occurrence frequency of the entities at both ends of the relationship edge in the existing graph is counted, and the ratio of the historical co-occurrence frequency to the total occurrence frequency of the entities at both ends is calculated. When the ratio is not lower than a preset co-occurrence probability threshold, the relationship edge is completed between the entities at both ends, and the inference source is labeled. When the same attribute of the same entity has multiple values, the frequency of each value in the scene graph association data is counted, and the value with the highest frequency is selected as the final value; if the frequencies are the same, the value with the latest update time is selected. Based on the node and edge types defined in the dynamic ontology, the mapped graph nodes and edges, merged nodes, and completed relation edges are organized to obtain an industry digital transformation scenario graph. This graph is stored in a graph database. Nodes represent entities and have type, name, and source attributes; edges represent relationships and have relationship type and confidence. The graph database supports industry-specific namespace isolation and version number management, and can export resource description framework format or JavaScript object notation linking data format.

[0053] S4: Prioritize the scenarios in the industry digital transformation scenario map, and write the obtained scenario priority as an attribute into the corresponding scenario node in the graph database.

[0054] In this specific implementation, this step prioritizes the scenarios in the graph and writes the evaluation results back to the graph as attributes of the scenario nodes. All data required for the evaluation is objectively read from the graph database, eliminating reliance on manual scoring and ensuring the evaluation results are traceable and reproducible. Specifically, the number of node attributes, the number of associated edges, the layer depth, historical revenue statistics, and historical risk statistics for each scenario node are read from the graph database and used to calculate four dimensions: digital element maturity, implementation difficulty, expected revenue, and implementation risk.

[0055] First, the maturity of digital elements is calculated. The number of node attributes for each scene node is read from the graph database, and the maturity of digital elements is calculated based on the number of node attributes. For example, if a scene node has data element attributes, knowledge model attributes, and tool software attributes, but lacks talent skills attributes, then... , , , The maturity level of digital elements is calculated using the following formula:

[0056] in, For the first The maturity of digital elements at each scenario node , , , The first The values ​​of data element attributes, knowledge model attributes, tool software attributes, and talent skill attributes for each scene node are 1 if they exist and 0 if they do not exist.

[0057] Next, the implementation difficulty is calculated. The number of associated edges and the level depth of each scene node in the graph database are read, and the implementation difficulty is calculated based on these parameters:

[0058] in, For the first The implementation difficulty of each scenario node For the first The number of associated edges for each scene node in the graph database. This represents the average number of associated edges for all scene nodes in the graph database. For the first The hierarchical depth of each scene node This represents the maximum level depth of all scene nodes in the graph database. For example, if a scene node has 8 associated edges, the average number of associated edges for all scene nodes in the graph database is 5, the level depth is 3, and the maximum level depth for all scene nodes is 4, then the implementation difficulty is... .

[0059] Then, calculate the expected returns and implementation risks. Read the historical revenue statistics associated with each scenario node, and calculate the expected returns based on these historical statistics.

[0060] in, For the first Expected revenue for each scenario node For the first The number of historical revenue statistics associated with each scenario node For the first A set of historical revenue statistics. For example, if a node in a certain scenario is associated with three historical revenue statistics, namely 10, 12, and 14, then the expected revenue... Read the historical risk statistics associated with each scenario node, and calculate the implementation risk based on the historical risk statistics:

[0061] in, For the first Implementation risks at each scenario node For the first The number of historical risk statistics associated with each scenario node For the first A historical risk statistic. For example, if a scenario node is associated with two historical risk statistic values, 3 and 5 respectively, then the implementation risk... .

[0062] Since the four dimensions mentioned above have different dimensions and value ranges, the calculated digital element maturity, implementation difficulty, expected benefits, and implementation risks need to be normalized to 0 to 1 using the following formula:

[0063] in, For the first The unnormalized index values ​​of each scene node This represents the set of values ​​for this metric for all scene nodes. The minimum value in the set of values. To be the maximum value in the set of values, This is the normalized value of the indicator. For example, if the minimum value of an indicator across all scene nodes is 0.5 and the maximum value is 1.0, and the value of this indicator in a certain scene node is 0.75, then the normalized value is... .

[0064] After normalization, the maturity, implementation difficulty, expected benefits, and implementation risks of the normalized digital elements are compared pairwise using the analytic hierarchy process (AHP) to construct a judgment matrix. And calculate the weight vector Perform a consistency check; when the consistency ratio... The weight allocation is considered reasonable when the value is less than the preset consistency threshold of 0.1, where:

[0065] To determine the matrix The largest eigenvalue, To determine the matrix The order of This is a preset average random consistency index. For example, a 4th-order judgment matrix is ​​constructed, and the weight vector is calculated. Consistency ratio If the value is less than 0.1, the weight allocation is considered reasonable. (Based on the weight vector) The normalized digital element maturity, implementation difficulty, expected benefits, and implementation risks are weighted and summed to obtain a comprehensive score for each scenario:

[0066] in, For the first The overall score of each scene node. , , , Weight vectors The weighting of digital elements maturity, implementation difficulty, expected benefits, and implementation risks in China, and meeting the following requirements. , , , , The first The normalized digital element maturity, implementation difficulty, expected benefits, and implementation risks for each scenario node. For example, if the normalized digital element maturity for a certain scenario node is 0.75, the implementation difficulty is 0.6, the expected benefits are 0.8, and the implementation risk is 0.4, then the overall score is... .

[0067] Finally, all scenarios are sorted according to their overall scores to form a priority list, and the overall score and priority of each scenario are written as attributes into the corresponding scenario node in the graph database. The overall score and priority can be used for priority building of the graph database index or for sorting and filtering during graph queries, so that the evaluation results directly serve the storage organization and retrieval operations of the graph, rather than just outputting a separate evaluation report.

[0068] S5: Based on the update trigger conditions, update the industry digital transformation scenario map using the newly added multi-source heterogeneous data.

[0069] In specific implementations, to ensure the continuous evolution of the map, timed or event-triggered updates are pre-set as the conditions for updates. Event-triggered updates include the addition of new data to the database. For example, an update can be scheduled to trigger at 2:00 AM every day, while simultaneously monitoring publishing websites; when new data is published, the update is triggered immediately.

[0070] When the update trigger condition is met, newly added multi-source heterogeneous data is collected. Preprocessing is performed on this data to obtain standardized new data. Based on structured output templates and pre-trained language models, four types of digital elements—industry, supply chain links, business activities, and transformation scenarios—are extracted from the standardized new data to obtain new scenario graph-related data. Knowledge fusion and graph modeling are performed on the new scenario graph-related data based on a dynamic ontology, generating incremental nodes and relationships. These incremental nodes and relationships are merged into the industry digital transformation scenario graph, with only the newly added or updated portions written, and the graph database's index and adjacency table updated, assigning a new version number. For example, adding a file about edge computing, after processing, adds edge computing-related scenario nodes and relationship edges to the graph, updates the index, and upgrades the version number from v1.0 to v1.1. The graph includes version numbers, supporting multi-version backtracking, comparison, and switching.

[0071] In one implementation, self-learning optimization can also be performed. The results of manual correction are accumulated as positive samples. Based on the existing model parameters, a small number of labeled samples from the industry are used for fine-tuning, adapting the model to industry terminology and expression habits without requiring training from scratch. The fine-tuned model is then fed back into the extraction stage of step S2, forming a closed loop of use, correction, and retraining. In this way, as the results in the manual review queue continuously accumulate, the pre-trained language model can gradually improve the accuracy of extracting the target industry, and the quality of incremental updates to the language map also improves.

[0072] S6: Visualize and render the industry digital transformation scenario map and / or output the application.

[0073] In a specific implementation, on the one hand, the system visualizes and renders the industry's digital transformation scenario map. The front-end renders a topology or hierarchical graph based on the hierarchical index of the graph database, receives user edits of nodes and edges and writes them back to the graph database, and receives scenario retrieval, feature query, relationship link tracing, and multidimensional statistical requests, returning the corresponding results from the graph based on the adjacency list of the graph database. For example, if a user clicks on a main scenario node on the front-end, the system expands its subordinate sub-scenario nodes based on the hierarchical index; if a user edits the priority attribute of a node, the system writes the modification back to the graph database.

[0074] On the other hand, the application output of industry digital transformation scenario maps is performed. Following templates, the industry digital transformation scenario map matrix, digital element list, and scenario priority list are extracted from the map and exported as a report. This report is then delivered via an application programming interface (API) gateway to enterprise digital platforms, industrial internet platforms, or business systems using a resource description framework (RPA) format or JavaScript object representation (JavaScript Object Notation) linked data format. For example, a list of digital elements for a specific industry can be exported as an Excel report, while the map data is simultaneously pushed to the enterprise digital platform via API.

[0075] Simultaneously, the system receives user questions through a question-and-answer service. This service performs semantic parsing to obtain search criteria, and then retrieves relevant entities, relationships, and original text fragments from the industry digital transformation scenario map and standardized data. Based on these criteria, it generates a natural language answer, annotating the source of the content within the answer. For example, if a user asks what digital elements are needed for equipment management in a certain industry, the question-and-answer service parses the search criteria as industry, equipment management, and digital elements. It retrieves the scenario node and its four types of digital element attributes from the map, and also finds relevant original text fragments in the standardized data. The resulting answer states that the scenario requires data elements such as equipment operation data, a fault prediction model, predictive maintenance systems, and data analysis skills, with the source annotated as page 3 of a certain document.

[0076] Through the aforementioned visualization, application output, and Q&A services, the graph can be effectively integrated with enterprise digital platforms, industrial internet platforms, and business systems to meet the needs of dynamic invocation and intelligent interaction.

[0077] In this embodiment, through unified parsing and disambiguation governance of multi-source heterogeneous data, combined with automatic extraction of large models under the constraints of structured output templates, and with the help of knowledge fusion between fill-in-the-blank dynamic ontology and graph database, quantitative priority evaluation based on objective data from graph database, timed and event-driven incremental updates, and visual interaction and retrieval-enhanced question answering, the efficient, standardized, dynamic, and intelligent generation of industry digital transformation scenario maps is achieved. This significantly improves the efficiency of map construction, data quality, standardization, evaluation objectivity, dynamic adaptability, and intelligent application capabilities, providing unified, reusable, and sustainably evolving structured support for industry digital transformation.

[0078] like Figure 2As shown, the following are embodiments of the industry digital transformation scenario map generation system provided in this disclosure. This system and the industry digital transformation scenario map generation method of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the industry digital transformation scenario map generation system, please refer to the embodiments of the above industry digital transformation scenario map generation method.

[0079] An industry digital transformation scenario mapping generation system includes: The data acquisition and preprocessing module is used to acquire multi-source heterogeneous data, including document data, table data, web page and interface data, and sensor data; and to preprocess the multi-source heterogeneous data to obtain standardized data; the preprocessing includes parsing, cleaning, deduplication, and entity disambiguation.

[0080] The extraction module is used to extract four types of digital elements from the standardized data based on the structured output template aligned with the preset map skeleton, namely industry, industrial chain links, business activities, transformation scenarios, and so on, using a pre-trained language model. It also performs hierarchical labeling, priority labeling, and relationship identification, and obtains scene map association data after verification.

[0081] The graph generation module is used to perform fill-in-the-blank ontology generation using the preset graph skeleton as a template to obtain a dynamic ontology. Based on the dynamic ontology, knowledge fusion and graph modeling are performed on the scene graph association data, and the data is stored in the graph database to obtain the industry digital transformation scene graph.

[0082] The priority assessment module is used to assess the priority of scenarios in the industry digital transformation scenario map and write the assessed scenario priority as an attribute into the corresponding scenario node in the graph database.

[0083] The update module is used to update the industry digital transformation scenario map using newly added multi-source heterogeneous data based on update trigger conditions.

[0084] The output module is used to visualize and render the industry digital transformation scenario map and / or output it for application.

[0085] The industry digital transformation scenario map generation system provided in this embodiment transforms multi-source heterogeneous industry data into a unified, complete, and clearly defined industry digital transformation scenario map by leveraging an automated generation and dynamic maintenance mechanism driven by a large model and an industry map skeleton. It supports objective assessment of scenario priorities, incremental map updates, version management, visual interaction, and enhanced search and Q&A capabilities, thereby significantly improving map construction efficiency, standardization, data quality, assessment objectivity, and intelligent application capabilities. This provides a unified, standardized, and reusable structured platform for both supply and demand sides in digital transformation, reducing construction and maintenance costs and enhancing the timeliness, scalability, and integration capabilities of the map with enterprise digital platforms, industrial internet platforms, and business systems.

[0086] Figure 3 A schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.

[0087] The industry digital transformation scenario mapping generation method provided in this application can be applied to electronic devices. Those skilled in the art will understand that the electronic device structure involved in the embodiments of this invention does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this invention, electronic devices include, but are not limited to, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0088] Electronic devices may include processors, external memory interfaces, internal memory, universal serial bus (USB) interfaces, charging management modules, power management modules, batteries, wireless communication modules, audio modules, speakers, microphones, sensor modules, buttons, cameras, displays, and SIM card interfaces, etc.

[0089] A processor may include one or more processing units, such as: a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors.

[0090] The processor can serve as the nerve center and command center of an electronic device. The controller can generate operation control signals based on the instruction opcode and timing signals to control the fetching and execution of instructions.

[0091] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from this memory. This avoids repeated accesses, reduces processor latency, and thus improves system efficiency.

[0092] An external storage interface (ESI) can be used to connect external memory cards, such as microSD cards, to expand the storage capacity of electronic devices. The external memory card communicates with the processor through the ESI to perform data storage functions, such as saving music and video files on the external memory card.

[0093] Internal memory can be used to store computer executable program code, which includes instructions. The processor executes various functional applications and data processing of electronic devices by running the instructions stored in internal memory. Internal memory can include a program storage area and a data storage area. Internal memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one disk storage device, flash memory device, universal flash storage (UFS), etc.

[0094] Wireless communication functionality in electronic devices can be achieved through antennas, wireless communication modules, modem processors, and baseband processors.

[0095] Wireless communication modules can provide solutions for wireless communication applications in electronic devices, including wireless local area networks (WLANs) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), and infrared (IR) technologies.

[0096] Electronic devices can implement audio functions through audio modules, speakers, receivers, microphones, headphone jacks, and application processors.

[0097] Electronic devices can achieve shooting functions through ISPs, cameras, video codecs, GPUs, displays, and application processors.

[0098] Electronic devices can achieve display functions through GPUs, displays, and application processors.

[0099] A GPU is a microprocessor for image processing, connected to the display screen and application processor. GPUs are used to perform mathematical and geometric calculations for graphics rendering. A processor may include one or more GPUs, which execute program instructions to generate or modify display information.

[0100] A display screen is used to display images, videos, etc. A display screen includes a display panel.

[0101] The aforementioned electronic device realizes the beneficial effects of automated, standardized, and dynamic generation of industry digital transformation scenario maps through the unified governance of multi-source heterogeneous data, extraction of large model template constraints, fill-in-the-blank dynamic ontology and graph database modeling, quantitative evaluation of scenario priorities, incremental updates, and intelligent application output of the method for generating industry digital transformation scenario maps in this application.

[0102] The storage medium provided in this application stores a program product capable of generating a scenario map for industry digital transformation.

[0103] In some possible implementations, the industry digital transformation scenario mapping generation method of this disclosure can be implemented as a program product, which includes program code. When the program product is run on a terminal device, the program code is used to cause the terminal device to perform the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of this disclosure.

[0104] The storage medium disclosed herein may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0105] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating a scenario map of industry digital transformation, characterized in that, Includes the following steps: S1: Collect multi-source heterogeneous data, including document data, table data, web page and interface data, and sensor data; preprocess the multi-source heterogeneous data to obtain standardized data; the preprocessing includes parsing, cleaning, deduplication, and entity disambiguation. S2: Based on the structured output template aligned with the preset map skeleton, the pre-trained language model is used to extract four types of digital elements from the standardized data: industry, industrial chain links, business activities, transformation scenarios, and so on. Hierarchical annotation, priority annotation, and relationship identification are performed, and the scene map association data is obtained after verification. S3: Using the preset graph skeleton as a template, perform fill-in-the-blank ontology generation to obtain a dynamic ontology. Based on the dynamic ontology, perform knowledge fusion and graph modeling on the scene graph association data, store it in the graph database, and obtain the industry digital transformation scene graph. S4: Prioritize the scenarios in the industry digital transformation scenario map, and write the obtained scenario priority as an attribute into the corresponding scenario node in the graph database; S5: Update the industry digital transformation scenario map using the newly added multi-source heterogeneous data according to the update trigger conditions; S6: Visualize and render the industry digital transformation scenario map and / or output the application.

2. The method for generating industry digital transformation scenario maps according to claim 1, characterized in that, Step S1 includes: The following parsing process is performed on the multi-source heterogeneous data to generate parsing result data: Extract the text layer from document-type data. When the text layer is missing or garbled, use optical character recognition (OCR) to identify the page image. Parse the table data row by row to obtain the records corresponding to the fields and values; Data from web pages and APIs can be captured by using the Hypertext Transfer Protocol interface or a data collection program based on fixed fields. The sensor data is parsed according to the sampling protocol or by timestamp and device identifier to obtain the time-stamped records, and the time-stamped records are aligned by timestamp. The parsed data is uniformly converted into UTF-8 text, and three types of metadata tags are added: source, type, and collection time. Clean and normalize the UTF-8 text, remove headers, footers, table of contents, blank lines and duplicate paragraphs, and normalize full-width / half-width and simplified / traditional Chinese characters to obtain normalized text; The content fingerprint of the normalized text is calculated for each paragraph, and only one fingerprint with the same content is retained to obtain the deduplicated text. For concepts with the same meaning but different expressions in the deduplicated text, synonym merging is first performed using a semantic dictionary; for concepts not covered by the semantic dictionary, the cosine similarity between two vectors is calculated using the following formula: when Not lower than the preset similarity threshold When they are identified as the same entity, they are merged while retaining their respective sources, resulting in disambiguated text. The required fields and naming conventions of the disambiguation text are checked according to the rules, and samples are randomly selected according to a preset ratio. The pass rate is calculated using the following formula. : in, This represents the number of samples that passed the random inspection. The total number of samples to be inspected; when When the pass rate falls below the preset threshold, the entire batch is returned for reprocessing; when When the disambiguation text is greater than or equal to the preset pass rate threshold, it will be used as standardized data.

3. The method for generating industry digital transformation scenario maps according to claim 2, characterized in that, Step S2 includes: A predefined JavaScript object representation output template aligned with a preset graph skeleton; the preset graph skeleton includes one industry level and five types of industry chain links. The text describes a series of business activities and their corresponding transformation scenarios, along with four types of digital elements. It also mentions the JavaScript object representation output template, which includes fields for industry, supply chain links, business activities, transformation scenarios, four types of digital elements, hierarchy, and priority. The four types of digital elements include data elements, knowledge models, software tools, and talent skills. Input the standardized data and a preset number of labeled examples into the pre-trained language model, and fill the template item by item according to the JavaScript object representation method through the pre-trained language model to output the structured extraction result; The structured extraction results identify transformation scenarios, which are then divided into four levels according to granularity: first level (industry level), second level (domain level), third level (enterprise level), and fourth level (job level). The identified transformation scenarios are further categorized into main scenarios and sub-scenarios. Sub-scenarios are further decompositions of the main scenarios, inheriting the four types of digital elements from their respective main scenarios. The resulting levels are written into the level field of the corresponding transformation scenario in the structured extraction results. The level annotations are used to construct a level index. For the main scene in the structured extraction result, data elements, knowledge models, software tools, and talent skills are labeled in the four categories of digital element fields of the main scene; For the main scenario and sub-scenario in the structured extraction result, mark them as basic essential level, core key level or forward-looking innovation level in the priority field of the structured extraction result according to the preset transformation guidance; Identify the collaborative relationships across industry chain links and the integration relationships across business activities from the structured extraction results, and write the collaborative relationships and the integration relationships into the association relationship field of the structured extraction results as association relationships to be mapped as graph edges; The labeled structured extraction results are validated and disambiguated, including: checking the validity of field types and enumeration values; checking the completeness of hierarchical relationships, i.e., the fourth, third, and second levels belong to their direct parent level, and filling or marking missing hierarchical relationships; when multiple expressions of the same entity appear, the frequency of each expression in the structured extraction results is counted, and the expression with the highest frequency is selected as the expression of the entity; when the same relationship is contradictory, the number of sources of each relationship expression in the structured extraction results is counted, and the relationship expression with the most sources is selected as the expression of the relationship; obtaining the generation probability output by the pre-trained language model when generating each structured extraction result, and using the generation probability as the confidence level of the structured extraction result, those below the preset confidence threshold of 0.7 are transferred to the manual review queue; The structured extraction results after verification and disambiguation constitute the scene map association data.

4. The method for generating industry digital transformation scenario maps according to claim 3, characterized in that, Step S3 includes: The entity types and relationship types defined in the preset graph skeleton are obtained, and specific instances and attribute specifications are filled into the preset graph skeleton according to the target industry to obtain a dynamic ontology. The entity types include industry, industrial chain links, business activities, transformation scenarios and four types of digital elements. The relationship types include attribution, inclusion, association, decomposition, cross-link collaboration and cross-business collaboration. Construct a graph database and align the node and edge types of the graph database with the fields of the JavaScript object representation output template; In the graph database, node types and edge types are defined according to the dynamic ontology. The main scene and sub-scenes are connected by decomposition relationship. The four types of digital elements are mounted as attributes on the main scene nodes. The sub-scenes reuse the four types of digital elements of the main scene. Based on the dynamic ontology, entities in the scene graph association data are mapped to graph nodes in the graph database and node identifiers are generated; relationships in the scene graph association data are mapped to graph edges in the graph database and edge indexes are generated. Calculate the cosine similarity between the vector of the newly extracted entity and the entity vector in the existing graph. When the cosine similarity is not lower than the preset fusion similarity threshold, merge the newly extracted entity and the existing entity into the same node to reduce redundant nodes in the graph database. For missing relation edges in the graph database, the number of historical co-occurrences of the entities at both ends of the relation edge in the existing graph is counted, and the ratio of the number of historical co-occurrences to the total number of occurrences of the entities at both ends is calculated. When the ratio is not lower than a preset co-occurrence probability threshold, the relation edge is completed between the entities at both ends, and the inferred source is marked on the completed relation edge. When the same attribute of the same entity has multiple values, the frequency of each value in the scene map association data is counted, and the value with the highest frequency is selected as the final value of the attribute; when multiple values ​​have the same frequency, the value with the latest update time is selected as the final value of the attribute. Based on the node and edge types defined in the dynamic ontology, the mapped graph nodes and edges, the merged nodes, and the completed relational edges are organized to obtain the industry digital transformation scenario graph. The industry digital transformation scenario map is stored in the graph database, where nodes represent entities and have type, name, and source attributes, and edges represent relationships and have relationship type and confidence. The graph database supports industry namespace isolation and version number management, and exports resource description framework format and / or JavaScript object representation link data format.

5. The method for generating industry digital transformation scenario maps according to claim 4, characterized in that, Step S4 includes: The system reads the number of node attributes for each scene node from the graph database and calculates the digital element maturity based on this number. It also reads the number of associated edges and the layer depth for each scene node in the graph database and calculates the implementation difficulty based on these dimensions. Furthermore, it reads historical revenue statistics associated with each scene node and calculates expected revenue based on these statistics. Finally, it reads historical risk statistics associated with each scene node and calculates the implementation risk based on these statistics. The system then normalizes the digital element maturity, implementation difficulty, expected revenue, and implementation risk to a standardized format. arrive ; The maturity level of the digital elements is calculated using the following formula: in, For the first The maturity of digital elements at each scenario node , , , The first The values ​​of data element attributes, knowledge model attributes, tool software attributes, and talent skill attributes for each scene node are taken when they exist. If it does not exist, take ; The implementation difficulty is calculated using the following formula: in, For the first The implementation difficulty of each scenario node For the first The number of associated edges for each scene node in the graph database. This is the average number of associated edges for all scene nodes in the graph database. For the first The hierarchical depth of each scene node This is the maximum value of the hierarchy depth of all scene nodes in the graph database; The expected return is calculated using the following formula: in, For the first Expected revenue for each scenario node For the first The number of historical revenue statistics associated with each scenario node For the first Historical earnings statistics; The implementation risk is calculated using the following formula: in, For the first Implementation risks at each scenario node For the first The number of historical risk statistics associated with each scenario node For the first One historical risk statistical value; The calculated maturity level of the digital elements, the implementation difficulty, the expected benefits, and the implementation risks are normalized to the following formulas: arrive : in, For the first The unnormalized index values ​​of each scene node This represents the set of values ​​for this metric for all scene nodes. It is the minimum value in the set of values. The maximum value in the set of values. These are the normalized index values; The analytic hierarchy process (AHP) is used to conduct pairwise comparisons of the maturity, implementation difficulty, expected benefits, and implementation risks of normalized digital elements, constructing a judgment matrix. And calculate the weight vector Perform a consistency check; when the consistency ratio... Less than the preset consistency threshold When determining whether the weight allocation is reasonable, the following applies: The judgment matrix The largest eigenvalue, The judgment matrix The order of The preset average random consistency index; According to the weight vector The normalized digital element maturity, implementation difficulty, expected benefits, and implementation risks are weighted and summed to obtain a comprehensive score for each scenario: in, For the first The overall score of each scene node. , , , The weight vectors are respectively The weighting of digital elements maturity, implementation difficulty, expected benefits, and implementation risks in China, and meeting the following requirements. , , , , The first The maturity, implementation difficulty, expected benefits, and implementation risks of digital elements after normalization of each scenario node; All scenarios are sorted according to their overall scores to form a priority list, and the overall score and priority of each scenario are written as attributes into the corresponding scenario node in the graph database.

6. The method for generating industry digital transformation scenario maps according to claim 5, characterized in that, Step S5 includes: Set timed triggers and / or event triggers as update trigger conditions, where the event triggers include new data being added to the database; When the update triggering condition is met, newly added multi-source heterogeneous data is collected, and the preprocessing is performed on the newly added multi-source heterogeneous data to obtain newly added standardized data. Based on the structured output template and the pre-trained language model, industry, industrial chain links, business activities, transformation scenarios and four types of digital elements are extracted from the newly added standardized data to obtain newly added scenario map association data; Based on the dynamic ontology, knowledge fusion and graph modeling are performed on the newly added scene graph association data to generate incremental nodes and incremental relationships. The incremental nodes and incremental relationships are merged into the industry digital transformation scenario graph, only the newly added or updated parts are written, and the indexes and adjacency tables of the graph database are updated, and a new version number is assigned.

7. The method for generating industry digital transformation scenario maps according to claim 6, characterized in that, Step S6 includes: The system performs visualization rendering of the industry digital transformation scenario map. The front end renders the topology map and / or hierarchical map according to the hierarchical index of the graph database. It receives user editing operations on nodes and edges and writes them back to the graph database. It receives scenario retrieval, element query, relationship link tracing and multidimensional statistics requests and returns the corresponding results from the industry digital transformation scenario map based on the adjacency list of the graph database. The application output of the industry digital transformation scenario map is carried out. According to the template, the industry digital transformation scenario map matrix, digital element list and scenario priority list are extracted from the industry digital transformation scenario map and exported as a report. The map data in resource description framework format and / or JavaScript object representation link data format is output to the enterprise digital platform, industrial Internet platform and / or business system through the application programming interface gateway in a standard interface. The system receives user questions through a question-and-answer service. The service performs semantic parsing on the user questions to obtain search criteria. Based on the search criteria, it retrieves relevant entities, relationships, and original text fragments from the industry digital transformation scenario map and the standardized data. It then generates a natural language answer based on the relevant entities, relationships, and original text fragments, and labels the source of the content in the natural language answer.

8. A system for generating scenario maps for industry digital transformation, characterized in that, The system employs the industry digital transformation scenario map generation method as described in any one of claims 1 to 7; The system includes: The data acquisition and preprocessing module is used to acquire multi-source heterogeneous data, including document data, table data, web page and interface data, and sensor data; and to preprocess the multi-source heterogeneous data to obtain standardized data; the preprocessing includes parsing, cleaning, deduplication, and entity disambiguation. The extraction module is used to extract four types of digital elements from the standardized data based on the structured output template aligned with the preset map skeleton, namely industry, industrial chain links, business activities, transformation scenarios, and others, using a pre-trained language model. It also performs hierarchical labeling, priority labeling, and relationship identification, and obtains scene map association data after verification. The graph generation module is used to perform fill-in-the-blank ontology generation using the preset graph skeleton as a template to obtain a dynamic ontology, and to perform knowledge fusion and graph modeling on the scene graph association data based on the dynamic ontology, and store it in the graph database to obtain the industry digital transformation scene graph. The priority evaluation module is used to evaluate the priority of scenarios in the industry digital transformation scenario map and write the evaluated scenario priority as an attribute into the corresponding scenario node in the graph database. The update module is used to update the industry digital transformation scenario map using newly added multi-source heterogeneous data according to the update trigger conditions; The output module is used to visualize and render the industry digital transformation scenario map and / or output it for application.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the industry digital transformation scenario mapping generation method as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the industry digital transformation scenario mapping generation method as described in any one of claims 1 to 7.