Industrial chain atlas construction and visualization method and system based on man-machine cooperation

Abstract

The invention discloses an industrial chain atlas construction and visualization method and system based on man-machine cooperation. The method comprises the following steps: S1, constructing an industrial knowledge network by using a natural language processing technology, and performing confidence verification and manual operation auditing; s2, constructing an industrial chain atlas skeleton based on expert knowledge, hanging lower nodes for key industrial concept nodes by extracting associated nodes and calculating semantic similarity, and perfecting node portraits; s3, searching and associating industrial chain element nodes to industrial concept nodes in the industrial knowledge network, calculating semantic similarity as confidence, and performing manual operation auditing to complete association mounting of the industrial chain element nodes; and S4, displaying the industrial chain atlas according to a visual form specified by a user. According to the method, the industrial chain maps in different fields can be accurately and efficiently constructed, the method has universality for different industries, and manpower and economic cost consumed in the industrial chain knowledge map construction process can be greatly reduced.

Description

Technical Field

[0001] The present invention belongs to the field of knowledge graphs, and specifically relates to a method and system for constructing and visualizing an industrial chain graph based on human-machine collaboration. Background Art

[0002] The industrial chain describes the entire process from raw materials to final products or services to consumers, covering multiple links such as raw material procurement, production, processing, and sales. These links are based on certain technical and economic connections, and form a chain-like relationship based on specific logical relationships and time-space layout relationships. The industrial chain map is a visualization tool that describes the various market segments of the industrial chain and their relationships. It integrates data such as enterprises, transactions, industries, information, and popularity to present each link and its relationship in the industrial chain, or its development status and trends in a graphical way, providing users with a comprehensive perspective on the entire industrial chain.

[0003] In the prior art, the methods for constructing industrial chain maps include a variety of methods, such as a construction model based on data collection and organization, a construction model based on industry classification standards, and a construction model based on data mining and knowledge extraction.

[0004] (1) Building an industrial chain map based on data collection and organization

[0005] First, data is collected from public information channels, such as the annual reports and prospectuses of listed companies, to obtain the business scope, product information, upstream and downstream customers of the enterprises, so as to determine the enterprises in the industrial chain and their relationships. At the same time, reference is made to industry research reports, industrial statistics data released by government departments, professional databases, etc., to enrich the data required for the map. After that, professionals screen, classify and organize the collected data, and identify the key links of the industrial chain, the position of enterprises in the industrial chain, and the upstream and downstream relationships between them based on their own industry knowledge and experience, and then construct the basic framework of the industrial chain map.

[0006] (2) Constructing an industrial chain map based on industry classification standards

[0007] First, use existing industry classification standards, such as the National Economic Industry Classification and Shenwan Industry Classification, to classify companies into different industry categories. On this basis, sort out the upstream and downstream relationships between industries and build an industrial chain map. For example, according to the National Economic Industry Classification standards, determine the relationship between a sub-industry in the manufacturing industry and other industries such as the raw materials industry and the service industry. Based on the general classification standards, combined with the professional knowledge and characteristics of specific industries, refine or adjust the classification to more accurately reflect the actual situation of the industrial chain. For example, in the new energy vehicle industry chain, in addition to determining the relationship between automobile manufacturers and parts manufacturers based on standard classification, it will also deeply analyze the integration relationship between emerging fields such as battery technology and autonomous driving technology and the traditional automobile industry to supplement and improve the industrial chain map.

[0008] (3) Building an industrial chain map based on data mining and knowledge extraction

[0009] Using natural language processing technology and information extraction algorithms, automatically extract entity and relationship information related to the industrial chain from a large amount of text data. For example, extract key information such as company name, product name, upstream and downstream relationship description from news reports, corporate official website introductions, industry forums, etc. as the basic data for building an industrial chain map. Using the construction concept and technology of knowledge graphs, the extracted entities and relationships are integrated and modeled to form a structured industrial chain knowledge graph. By defining entity types, relationship types, and attributes, a semantically related industrial chain map is constructed to achieve more effective representation and utilization of industrial chain knowledge.

[0010] The process of constructing the industrial chain map in the above-mentioned existing technologies all relies on a large number of operators and expert knowledge. It is basically difficult to share basic data between different industrial chains, and the human resources and economic costs consumed are relatively large. Therefore, providing a technology that can reduce the input of operators and expert knowledge and automatically construct the industrial chain map has a high practical demand for the construction scenario of the industrial chain map. Summary of the invention

[0011] The purpose of the present invention is to solve the many deficiencies in the construction and visualization of industrial chain maps in the prior art, and to provide an industrial chain map construction and visualization method and system based on human-computer collaboration.

[0012] The specific technical solutions adopted by the present invention are as follows:

[0013] In a first aspect, the present invention provides a method for constructing and visualizing an industrial chain map based on human-machine collaboration, which comprises:

[0014] S1. Obtain the pre-constructed industrial knowledge base in the target industry field, and use natural language processing technology to extract information from the data in the industrial knowledge base, and construct an industrial knowledge network based on the extracted objects and the relationship between the objects; the objects of information extraction include at least industrial concepts and industrial chain elements, and each extracted relationship has a confidence level; verify the constructed industrial knowledge network, send the tuples with a confidence level lower than the first threshold or conflicting tuples to the operation review end for manual operation review, and modify the industrial knowledge network based on the results returned by the operation review end;

[0015] S2. For the target industrial chain, obtain the industrial chain map skeleton composed of key industrial concept nodes pre-constructed based on expert knowledge; for each key industrial concept node in the industrial chain map skeleton, extract the industrial concept nodes associated with this key industrial concept node from the industrial knowledge network and add them to the first node set, and at the same time, pair this key industrial concept node with each industrial concept node in the industrial knowledge network respectively, generate node representation using the big model and calculate the semantic similarity between nodes, then add all industrial concept nodes whose semantic similarity exceeds the second threshold to the second node set, and finally attach the nodes in the first node set and the second node set as the subordinate nodes of this key industrial concept node to the industrial chain map skeleton; after all the key industrial concept nodes in the industrial chain map skeleton complete the growth of subordinate nodes, use the big model technology to improve the node portrait for each industrial concept node and update it to the node attribute, obtain the complete node hierarchy structure of the industrial chain map and send it to the expert end for expert inspection, and correct the node hierarchy according to the results returned by the expert end;

[0016] S3. For each industrial concept node in the node hierarchy of the industrial chain map, search for industrial chain element nodes associated with this industrial concept node in the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with this industrial concept node, then use the big model to generate node representation and calculate the semantic similarity between nodes, use the semantic similarity between each industrial chain element node and this industrial concept node as the confidence of this industrial chain element node, send the industrial chain element nodes whose confidence does not exceed the third threshold to the operation review end for manual operation review, and correct the industrial chain element nodes finally associated and mounted on the industrial concept node according to the results returned by the operation review end; after all industrial concept nodes complete the association and mounting of industrial chain element nodes, the construction of the industrial chain map is completed;

[0017] S4. Visually display the industrial chain map according to the visualization format specified by the user.

[0018] As a preferred embodiment of the above-mentioned first aspect, the method for constructing the industrial knowledge base is: based on the industrial knowledge system pre-constructed for the target industry field, knowledge data of different dimensions in the industrial knowledge system are collected from external data sources through data collection and access technology, and structured storage is performed after data processing and fusion to form an industrial knowledge base.

[0019] As a preferred embodiment of the above-mentioned first aspect, in S1, the method for constructing an industrial knowledge network based on the industrial knowledge base is: for knowledge data of different dimensions in the industrial knowledge base, tuples consisting of objects and relationships are extracted therefrom using an information extraction model based on natural language processing technology, wherein the extracted objects include concepts, entities, events, and attributes related to the industry, and the relationships between objects are all with confidence; the objects obtained by information extraction are used as network nodes, and the network nodes are associated and coupled according to the relationships obtained by information extraction to construct an industrial knowledge network.

[0020] As a preferred embodiment of the first aspect, in S2, for each pair of paired key industry concept nodes and industry concept nodes, the node attribute fields of the key industry concept nodes are spliced ​​and sent to the big model to generate a first representation vector, and then the node attribute fields of the industry concept nodes are spliced ​​and sent to the big model to generate a second representation vector, and the similarity between the first representation vector and the second representation vector is calculated and used as the semantic similarity between the nodes of the current paired nodes;

[0021] As a preferred embodiment of the first aspect mentioned above, the node portrait generated for each industry concept node using the big model technology includes an introduction to the industry concept, synonyms, industry classification, upstream products and downstream products.

[0022] As a preferred embodiment of the above-mentioned first aspect, in S3, for each industrial concept node in the industrial chain map and the industrial chain element node found in the industrial knowledge network, the node attribute fields of this industrial concept node need to be spliced ​​and sent to the big model to generate a third representation vector, and then the attribute fields of this industrial chain element node are also spliced ​​and sent to the big model to generate a fourth representation vector, and the similarity between the third representation vector and the fourth representation vector is calculated and used as the confidence of this industrial chain element node.

[0023] As a preferred embodiment of the first aspect, in S4, when the industrial chain map is visualized, the visualization dimensions include one or more of the industrial chain structure dimension, the industrial chain distribution situation dimension and the industrial chain change trend dimension.

[0024] As a preferred embodiment of the above-mentioned first aspect, the manual review results of the operation review end and the expert end need to be saved and constructed as an incremental learning data set of the artificial intelligence algorithm used to generate the original data to be reviewed, so as to optimize and improve the artificial intelligence algorithm.

[0025] In a second aspect, the present invention provides an industrial chain map construction and visualization system based on human-machine collaboration, which includes:

[0026] The industrial knowledge network construction module is used to obtain the pre-constructed industrial knowledge base in the target industry field, and use natural language processing technology to extract information from the data in the industrial knowledge base, and construct an industrial knowledge network based on the extracted objects and the relationship between the objects; the objects of information extraction include at least industrial concepts and industrial chain elements, and each extracted relationship has a confidence level; the constructed industrial knowledge network is verified, and the tuples with a confidence level lower than the first threshold or conflicts are sent to the operation review end for manual operation review, and the industrial knowledge network is corrected according to the results returned by the operation review end;

[0027] The industrial concept node expansion module is used to obtain the industrial chain map skeleton composed of key industrial concept nodes pre-built based on expert knowledge for the target industrial chain; for each key industrial concept node in the industrial chain map skeleton, extract the industrial concept nodes associated with this key industrial concept node from the industrial knowledge network and add them to the first node set, and at the same time, after pairing this key industrial concept node with each industrial concept node in the industrial knowledge network, use the big model to generate node representation and calculate the semantic similarity between nodes, and then add all industrial concept nodes with semantic similarity exceeding the second threshold to the second node set, and finally attach the nodes in the first node set and the second node set as the lower nodes of this key industrial concept node to the industrial chain map skeleton; after all the key industrial concept nodes in the industrial chain map skeleton complete the lower node attachment growth, use the big model technology to improve the node portrait for each industrial concept node and update it to the node attribute, obtain the complete node hierarchy structure of the industrial chain map and send it to the expert end for expert inspection, and modify the node hierarchy according to the results returned by the expert end;

[0028] The industrial chain element mounting module is used to sequentially search for industrial chain element nodes associated with each industrial concept node in the node hierarchy of the industrial chain map in the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with the industrial concept node, then use the large model to generate node representations and calculate the semantic similarity between nodes, use the semantic similarity between each industrial chain element node and the industrial concept node as the confidence of the industrial chain element node, send the industrial chain element nodes whose confidence does not exceed the third threshold to the operation review end for manual operation review, and correct the industrial chain element nodes finally associated and mounted on the industrial concept node according to the results returned by the operation review end; after all industrial concept nodes complete the association and mounting of industrial chain element nodes, the construction of the industrial chain map is completed;

[0029] The graph visualization module is used to visualize the industrial chain graph according to the visualization form specified by the user.

[0030] In the third aspect, the present invention provides a computer-readable storage medium, characterized in that a computer program is stored on the storage medium. When the computer program is executed by a processor, it can implement the industrial chain map construction and visualization method based on human-computer collaboration as described in any one of the solutions in the first aspect above.

[0031] In a fourth aspect, the present invention provides a computer electronic device, characterized in that it includes a memory and a processor;

[0032] The memory is used to store computer programs;

[0033] The processor is used to implement the industrial chain map construction and visualization method based on human-computer collaboration as described in any one of the solutions of the first aspect above when executing the computer program.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] (1) The present invention provides a method for semi-automatically constructing an industrial chain map through human-machine collaboration, which fully combines big data intelligence and group intelligence. First, based on big data intelligence, natural language processing technology, large language model and other artificial intelligence technologies are used to extract core concepts, entities, documents and other knowledge from massive industrial big data and associate them. Then, the group intelligence of operators and the expert wisdom of industry experts are used to adjust and review the constructed industrial chain map, and the accuracy and scientificity of the industrial chain map are improved by repeated iterations. The present invention can quickly, accurately and efficiently construct industrial chain maps in different fields, has universality for different industries, and can greatly reduce the human resources and economic costs consumed in the process of constructing the industrial chain knowledge map.

[0036] (2) The present invention uses a combination of multiple visualization forms to present the map. Under human-machine collaboration, the machine selects the form and initially lays out the mapping data, and the human designer optimizes the visual effect. The network diagram presents the structure, the tree diagram shows the hierarchy, the heat map expresses the heat, and the radial diagram highlights the core relationship. The map information is clearly presented in multiple dimensions, which enhances the user's rapid, intuitive, and in-depth exploration and understanding of different dimensions of the industrial chain map.

[0037] (2) The knowledge graph constructed by the present invention is based on the perspective of the industrial chain. In practical applications, it can deeply deconstruct the internal organizational structure of the industry through visualization technology, so as to facilitate enterprises and government departments to sort out the complex upstream and downstream relationships and clearly present the factors affecting competitiveness hidden behind the connection, coordination, and checks and balances of various links in the industrial chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of an industrial chain map construction and visualization method based on human-machine collaboration;

[0039] Figure 2 An exemplary tree diagram for visualizing the industrial chain structure;

[0040] Figure 3 An exemplary radial diagram for visualizing the industrial chain structure;

[0041] Figure 4 An exemplary Sankey diagram for visualizing the industrial chain structure;

[0042] Figure 5 An exemplary octopus diagram for visualizing the industrial chain structure;

[0043] Figure 6 An exemplary regional distribution map for visualizing the distribution trend of the industrial chain;

[0044] Figure 7 An exemplary enterprise characteristic distribution diagram for visualizing the distribution situation of the industrial chain;

[0045] Figure 8 An exemplary industrial chain node distribution diagram for visualizing the distribution situation of the industrial chain;

[0046] Fig. 9 An exemplary distribution map of key enterprises for visualizing the distribution situation of the industrial chain;

[0047] Fig.10 An exemplary enterprise change trend diagram for visualizing the change trend of the industrial chain;

[0048] Fig.11 An exemplary patent authorization trend chart for visualizing the trend of changes in the industrial chain;

[0049] Fig.12 An exemplary technology life cycle change diagram for visualizing the trend of industry chain changes;

[0050] Fig.13 An exemplary enterprise type change diagram for visualizing the changing trend of the industrial chain. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation mode of the present invention is described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the present invention can be combined accordingly without conflicting with each other.

[0052] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for the purpose of distinguishing descriptions, and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0053] The present invention provides a method for constructing and visualizing an industrial chain map based on human-machine collaboration, which includes steps S1 to S4. The method uses a human-machine collaborative industrial chain map construction method. On the one hand, it is necessary to introduce artificial intelligence technologies such as natural language processing, large models, and knowledge maps to automatically process basic data in the process of constructing the industrial chain knowledge map. On the other hand, it is necessary to introduce experts and operators to check the data generated by artificial intelligence technology, eliminate unqualified data, and use the feedback results to further optimize the model used in artificial intelligence technology, thereby continuously improving the accuracy of the model, gradually reducing the degree of participation of experts and operators, and realizing the automatic update of the industrial chain map.

[0054] In conjunction with a preferred embodiment of the present invention, the specific implementation of steps S1 to S4 in the present invention is described in detail below.

[0055] S1. Obtain a pre-constructed industrial knowledge base in the target industry field, and use natural language processing technology to extract information from the data in the industrial knowledge base, and construct an industrial knowledge network based on the extracted objects and the relationships between objects; the objects of information extraction include at least industrial concepts and industrial chain elements, and each extracted relationship has a confidence level; verify the constructed industrial knowledge network, and send the tuples with a confidence level lower than the first threshold or conflicts to the operation review end for manual operation review, and modify the industrial knowledge network based on the results returned by the operation review end.

[0056] It should be noted that the above-mentioned industry knowledge base is pre-constructed based on expert knowledge and industry experience. In the present invention, the pre-constructed industry knowledge base can be directly read, or it can be optimized and constructed for the target industry field. In an embodiment of the present invention, the method for constructing the industry knowledge base is: based on the pre-constructed industry knowledge system for the target industry field, the knowledge data of different dimensions in the industry knowledge system is collected from external data sources through data collection and access technology, and structured storage is performed after data processing and fusion to form an industry knowledge base.

[0057] The following is a preferred example to further illustrate how to construct an industry knowledge base in an embodiment of the present invention. In this example, the construction of the industry knowledge base includes two steps: industry knowledge system design and industry knowledge collection and import.

[0058] Step 1-1. Design of industrial knowledge system: Construct an industrial knowledge system for specific industrial fields. Based on the general industrial knowledge system, manually expand the knowledge system according to the different characteristics and special data of specific industries, and form a knowledge system including concepts (industry classification system, product concept nodes, technology concept nodes), entities (enterprises, talents, parks, investment institutions), documents (patents, papers, scientific research projects, information, research reports, policies), events (investment and financing events, risk events), documents (patents, papers, scientific research projects, information, research reports, policies), indicators (macroeconomic indicators, corporate financial indicators), etc., and complete the database table design and field design of the industrial knowledge base.

[0059] Step 1-2, industrial knowledge collection and import: Corresponding to the industrial knowledge system in a specific field, use data collection and access technology to access from the industrial database and collect relevant industrial data from the Internet, and import it into the corresponding industrial knowledge base through field mapping and fusion.

[0060] The data collected in the above step 1-1 is from public and legal data sources on the Internet, such as the National Enterprise Credit Information Publicity System to collect enterprise information, the National Intellectual Property Administration to collect patent information, university teachers' public homepages to collect talent information, commodity platforms to collect product information, governments at all levels to collect policy information, industry information websites to collect information, etc., and are collected and sorted regularly through crawlers, manual collection, etc.

[0061] The access technology in the above steps 1-2 is a data service API interface provided by a third-party data source, and HTTP technology is used to periodically and automatically obtain corresponding data in batches from the data service API.

[0062] The present invention needs to perform data fusion because the fields of the same type of data (such as information) obtained from different data sources and different data channels are inconsistent (for example, the title field, one is title, the other is tl, which need to be unified into title). In order to import the same type of data from different sources into the same database, data fusion is required, that is, the collected access data fields and the fields designed by the database are matched and converted. At the same time, it is also necessary to specify fusion rules, including the priority of the data source, the format verification of the field, the update strategy, etc. For example, a field (address) of the same enterprise has two data sources, but they are inconsistent. At this time, it is necessary to formulate fusion rules to select the optimal data source field. Through data fusion, the same type of data governance from multiple sources and multiple types can be realized to form a unified and high-quality knowledge base.

[0063] Therefore, based on the above process, an industry knowledge base can be constructed for the target industry field. Subsequently, according to the above S1 step, the industry knowledge network can be semi-automatically constructed through the human-machine collaboration method of operation + artificial intelligence.

[0064] It should be noted that in the above-mentioned S1 step of the present invention, the core of constructing an industrial knowledge network is to use an information extraction model based on natural language processing technology to extract triples consisting of objects and relationships from various knowledge data in the industrial knowledge base, and these triples record the associations between different objects. In an embodiment of the present invention, the method for constructing an industrial knowledge network based on an industrial knowledge base is: for different dimensional knowledge data in the industrial knowledge base, using an information extraction model based on natural language processing technology to extract triples consisting of objects and relationships, wherein the extracted objects at least include concepts related to the industry (industry classification system, product concept nodes, technical concept nodes), entities (enterprises, talents, parks, investment institutions), events (investment and financing events, risk events), attributes (specific information of entities, events, etc.), and can also extract documents (patents, papers, scientific research projects, information, research reports, policies), indicators (macroeconomic indicators, corporate financial indicators) and other object types as needed, and the extracted triples also record the relationship between objects and each triple relationship has a confidence output by the information extraction model. After completing the information extraction of all knowledge data in the industrial knowledge base, a series of triples are obtained. The objects obtained by information extraction are used as network nodes. The network nodes are associated and coupled according to the relationships obtained by information extraction to construct an industrial knowledge network.

[0065] The following is a preferred example to further illustrate how to construct an industrial knowledge network in an embodiment of the present invention. The example includes two steps: automatic extraction of industrial knowledge and operation of the industrial knowledge network.

[0066] Step 1-3, automatic extraction of industrial knowledge: Based on the industrial knowledge base, natural language processing technology is used to extract information, including concept extraction (extracting product concepts, technical concepts, etc.), entity extraction (extracting entities such as enterprises, talents, investment institutions, etc.), event extraction (extracting events such as enterprise cooperation, investment integration, product development, talent cooperation, etc.), relationship extraction (extracting the relationship between entities, such as the cooperative and competitive relationship between enterprises, the production and supply relationship between enterprises and products, the appointment relationship between enterprises and talents, the cooperative relationship between talents, etc.), attribute extraction (attribute supplement of entities, events, etc.), subject classification (such as the subject classification of information and research reports), etc., to deeply mine and associate the independent industrial knowledge base to form a network-like industrial knowledge network, where the nodes of the network are the concepts, entities, documents, events, and indicators in the industrial system design, and the edges of the network are the relationships between the above objects. The confidence output by the information extraction model is also recorded on the corresponding nodes or edges of the industrial knowledge network to facilitate subsequent manual operations for verification.

[0067] Steps 1-4, industrial knowledge network operation: After obtaining the initial industrial knowledge network, the results extracted by the industrial information extraction model with low confidence or conflicts can be automatically pushed to the operator client for operation and review, and the industrial knowledge network will be modified according to the operation review results.

[0068] It should be noted that, since the aforementioned industrial information extraction model outputs the confidence of the result (probability between 0-1) while outputting the extraction result, the introduction of the result with too low confidence into the industrial knowledge network may cause the industrial knowledge network to be unreliable. Therefore, the triples with low confidence in the industrial knowledge network need to be reviewed by the operator and can only exist in the final industrial knowledge network after confirmation. In addition, in addition to considering the confidence, the present invention also needs to consider the result conflict problem caused by different data sources. The so-called result conflict means that the result of information extraction conflicts with the existing knowledge base or other data sources. The legal representative of the company extracted from the information is A, but the legal representative of the company in the knowledge base is B. There is a conflict between these two pieces of information, and the results of information extraction need to be reviewed by the operator to retain the correct information and delete the wrong information.

[0069] In addition, the manual review results of the operation review end can be saved and constructed as an incremental learning data set of the information extraction model, which can be used to optimize the information extraction model. This will continuously optimize the information extraction model and reduce the frequency of subsequent operator intervention.

[0070] It should be noted that the operation audit end in the present invention may be a client running in the form of software, for operation audit personnel to manually verify the data to be audited sent by the system and issue correction instructions to the system.

[0071] S2. For the target industrial chain, obtain the industrial chain map skeleton composed of key industrial concept nodes pre-constructed based on expert knowledge; for each key industrial concept node in the industrial chain map skeleton, extract the industrial concept nodes associated with this key industrial concept node from the industrial knowledge network and add them to the first node set, and at the same time, pair this key industrial concept node with each industrial concept node in the industrial knowledge network respectively, use the large model (LLM) to generate node representation and calculate the semantic similarity between nodes, then add all industrial concept nodes with semantic similarity exceeding the second threshold to the second node set, and finally attach the nodes in the first node set and the second node set as the subordinate nodes of this key industrial concept node to the industrial chain map skeleton; after all the key industrial concept nodes in the industrial chain map skeleton complete the growth of subordinate nodes, use the large model technology to improve the node portrait for each industrial concept node and update it to the node attribute, obtain the complete node hierarchy structure of the industrial chain map and send it to the expert end for expert inspection, and correct the node hierarchy according to the results returned by the expert end.

[0072] It should be noted that the above-mentioned industrial chain map skeleton is a pre-constructed map skeleton based on the expert knowledge in the target industry field, in which some key industrial concept nodes in the target industry field are marked in a hierarchical form, but the subdivided industrial concept nodes may be missing, and need to be further supplemented and linked using the industrial knowledge network.

[0073] In an embodiment of the present invention, each key industry concept node in the industrial chain map skeleton needs to search for industry concepts (such as product concepts or technical concepts) related to this key industry concept node from the industrial knowledge network, and expand it to the industrial chain map skeleton. The industry concepts that can be expanded to the industrial chain map skeleton in the industrial knowledge network are divided into two categories. One category is the industrial concept nodes extracted from the industrial knowledge network that are associated with this key industry concept node (that is, there are edge connections), and the other category is some industrial concept nodes in the industrial knowledge network that have a high semantic similarity with this key industry concept node. The two categories of industrial concepts are recorded in the aforementioned first node set and the second node set, respectively. Of course, the two node sets obtained for each key industry concept node may be repeated, so they can be removed and attached to the industrial chain map skeleton.

[0074] In addition, it is particularly important to note that the industrial concept nodes in the above-mentioned second node set are screened using the semantic similarity between nodes, which depends on the original attribute similarity in each node. In order to improve the accuracy of attribute similarity calculation, the present invention introduces a large model technology to generate accurate node attribute representations. In the above-mentioned S2 step of the present invention, for each pair of paired key industrial concept nodes and industrial concept nodes, the method of calculating the semantic similarity between nodes by the large model representation technology is: the node attribute fields of the key industrial concept node are spliced ​​and sent to the large model to generate a first representation vector, and then the node attribute fields of the industrial concept node are spliced ​​and sent to the large model to generate a second representation vector, and the similarity between the first representation vector and the second representation vector is calculated and used as the semantic similarity between nodes of the current paired node.

[0075] In addition, the present invention also uses big model technology to generate node portraits for each industrial concept node in the complete node hierarchy of the industrial chain map (such as the introduction, synonyms, industry classification, upstream products and downstream products of this industrial concept), so as to facilitate the subsequent connection of industrial chain elements related to this industrial concept based on the node portraits of the industrial concept nodes.

[0076] The following is a preferred example to further illustrate how to construct a complete node hierarchy of the industrial chain map in the embodiment of the present invention, which mainly includes the steps of identifying key nodes, constructing a node hierarchy and improving node portraits. In this example, the industry concept mainly considers the product concept.

[0077] Step 2-1. Identify key nodes: For specific industrial fields, experts first form the main skeleton structure of the industrial chain system of the specific industry based on the classification system standards within the industry, that is, the skeleton of the industrial chain map composed of a series of product concepts.

[0078] Step 2-2, construct a node hierarchy: Based on the product concept library composed of all product concept nodes in the above industrial knowledge network, use the big model technology to expand the industrial chain map skeleton constructed in the previous steps, and calculate the semantic association strength and hierarchical relationship between each product concept in the product concept library and the existing nodes of the industrial chain map skeleton. Screen product concepts whose semantic association strength with existing nodes reaches the set threshold and product concepts with hierarchical relationships as subordinate nodes of each existing key industrial chain concept node in the industrial chain map skeleton.

[0079] Step 2-3, improve the node portrait: use the big model technology to improve the node portrait for each industrial chain node in the expanded industrial chain map skeleton, including node introduction, synonyms, industry classification, upstream products and downstream products. After the node system is initially constructed through the above series of steps, it is sent to the expert side, and experts are introduced to check whether the internal logic of the node system is reasonable, including whether the hierarchical structure is clear, whether the relationship between nodes is accurate, etc., and evaluate whether the node system can accurately reflect the actual operation of the industrial chain, and correct the unreasonable parts.

[0080] In this example, when improving the node portrait of the industrial chain node, taking the GPU industrial chain node as an example, the large model needs to generate the following information:

[0081] Introduction: GPU (Graphics Processing Unit) is a chip specially designed for processing graphics-related tasks. It plays a key role in computer graphics rendering and can quickly process and render complex 3D graphics and images. GPU has many processing units and is good at parallel computing. In addition to graphics processing, it is also widely used in computing-intensive fields such as deep learning.

[0082] Synonyms: graphics processor, graphics chip;

[0083] Industry classification: C397 Electronic device manufacturing;

[0084] Upstream products: wafers, lithography equipment, and specialty gases;

[0085] Downstream products: graphics cards, smart security equipment, and self-driving cars.

[0086] S3. For each industrial concept node in the node hierarchy of the industrial chain map, search for industrial chain element nodes associated with this industrial concept node in the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with this industrial concept node, then use the big model to generate node representation and calculate the semantic similarity between nodes, use the semantic similarity between each industrial chain element node and this industrial concept node as the confidence of this industrial chain element node, send the industrial chain element nodes whose confidence does not exceed the third threshold to the operation review end for manual operation review, and correct the industrial chain element nodes finally associated and mounted on the industrial concept node according to the results returned by the operation review end; after all industrial concept nodes complete the association and mounting of industrial chain element nodes, the construction of the industrial chain map is completed.

[0087] It should be noted that the industrial concept nodes that need to be connected to the industrial chain element nodes in the present invention are mainly product concepts, and the industrial chain elements that need to be connected include enterprises, patents, talents, investment institutions and policies related to the product concept. In order to improve the accuracy of the calculation of semantic similarity between nodes, the present invention also introduces the big model technology in the S3 step to generate accurate node attribute representation. In an embodiment of the present invention, the specific method of calculating semantic similarity through the big model representation technology is as follows: for each industrial concept node in the industrial chain map and the industrial chain element node found in the industrial knowledge network, the node attribute field of this industrial concept node needs to be spliced ​​and sent to the big model to generate a third representation vector, and then the attribute field of this industrial chain element node is also spliced ​​and sent to the big model to generate a fourth representation vector, and the similarity between the third representation vector and the fourth representation vector is calculated and used as the confidence of this industrial chain element node.

[0088] The following is a preferred example to further illustrate how to achieve automatic connection of industrial chain elements in the embodiment of the present invention. The automatic connection process relies on the industrial knowledge base constructed in the above steps, including enterprises, patents, talents, investment institutions, policies, etc., combined with the semantic characteristics of industrial chain nodes, and based on the association algorithm of large model representation learning, through node association, representation alignment and enhancement, semantic association calculation, review and verification and other processes, to achieve matching association between these industrial chain elements and nodes, and generate an industrial chain knowledge map composed of industrial chain concepts and industrial chain elements. The specific process is as follows:

[0089] Step 3-1) Alignment and enhancement of feature semantic representation: For each industrial chain concept node, find the industrial chain element nodes related to the industrial chain concept node from the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with this industrial concept node.

[0090] Step 3-2) Concatenate the field features (name, synonyms, introduction, industry classification) of the concept nodes of the industrial chain and convert them into a 1024-dimensional vector using the Embedding technology of the big model. Perform similar node attribute representation processing on the elements of the industrial chain to be attached. For example, concatenate the name, introduction, business scope, industry classification, and main products of the enterprise and convert them into a 1024-dimensional vector. The semantic representation alignment of the nodes and elements of the industrial chain can be achieved through the above method.

[0091] Step 3-3): Semantic association calculation: Calculate the cosine similarity of the semantic vector of the industrial chain concept node and the semantic vector of the industrial chain element as the confidence level of the similarity between the two.

[0092] Step 3-4): Audit and verification: The results output by the algorithm are manually verified and audited through random sampling. At the same time, the results with lower confidence are also manually verified and audited.

[0093] In the actual crowdsourcing operation process of the industrial chain map, the association between the industrial elements and the industrial chain nodes generated in the above S3 step can be pushed to the crowdsourcing personnel for verification and operation by screening the confidence threshold. The matching association results with low confidence can also be fed back to the artificial intelligence algorithms and models used in this process for subsequent optimization and iteration of the algorithms and models. When tuning the algorithm, you can manually construct a test set and a verification set as the basis for algorithm tuning. Algorithm tuning mainly includes the selection and processing of semantic features (whether the semantics of the industrial chain nodes need to include industry classifications, whether the introduction needs to do certain exclusion word processing, which large model is used for Embedding, and the dimension of the vector, etc.).

[0094] S4. Visually display the industrial chain map according to the visualization format specified by the user.

[0095] It should be noted that the method steps shown in S1 to S4 above can essentially be implemented in the form of computer programs. In actual implementation, the above steps can be automatically scheduled through automated software and tools to continuously improve the accuracy of the model, gradually reduce the degree of participation of experts and operators, and realize the automatic update of the industrial chain map. The whole process is as follows Figure 1 As shown, through continuous iteration, the frequency of operations and expert intervention can be continuously reduced, and artificial intelligence technology will process data in most links efficiently and accurately.

[0096] It should be noted that the specific form and logic of visualizing the industrial chain map can be reasonably designed and optimized according to the actual application scenario. In an embodiment of the present invention, when visualizing the industrial chain map, the visualization dimension includes one or more of the industrial chain structure dimension, the industrial chain distribution situation dimension, and the industrial chain change trend dimension.

[0097] When visualizing the industrial chain structure, the industrial chain system can be visualized in a tree diagram ( Figure 2 ), radial graph( Figure 3 )、Sankey diagram( Figure 4 ), Octopus( Figure 5 ) and other forms of display, the industrial chain system architecture can be viewed from multiple dimensions and perspectives.

[0098] When visualizing the distribution of the industrial chain, you can use the user-specified industrial chain system and related enterprises to visualize the regional distribution ( Figure 6 ), enterprise characteristics distribution ( Figure 7 ), industry chain node distribution ( Figure 8 )、Key Enterprise Distribution( Fig. 9) and other distribution methods to show the distribution trend of the industrial chain.

[0099] When visualizing the trend of industrial chain changes, you can use the user-specified industrial chain system and related enterprises, patents, investment and financing information to visualize the trend of enterprise changes ( Fig.10 )、Patent Licensing Trends( Fig.11 ), Technology Life Cycle( Fig.12 ) and investment and financing trends ( Fig.13 ) for visual display.

[0100] It can be seen that the present invention can perform flexible visualization based on different industrial chain perspectives, can deeply deconstruct the internal organizational structure of the industry, sort out the complex upstream and downstream connections, and clearly present the factors affecting competitiveness hidden behind the connection, coordination, and checks and balances of each link in the industrial chain.

[0101] Therefore, based on the same inventive concept, the present invention also provides an industrial chain map construction and visualization system based on human-machine collaboration corresponding to the industrial chain map construction and visualization method based on human-machine collaboration provided in the above embodiment, which includes:

[0102] The industrial knowledge network construction module is used to obtain the pre-constructed industrial knowledge base in the target industry field, and use natural language processing technology to extract information from the data in the industrial knowledge base, and construct an industrial knowledge network based on the extracted objects and the relationship between the objects; the objects of information extraction include at least industrial concepts and industrial chain elements, and each extracted relationship has a confidence level; the constructed industrial knowledge network is verified, and the tuples with a confidence level lower than the first threshold or conflicts are sent to the operation review end for manual operation review, and the industrial knowledge network is corrected according to the results returned by the operation review end;

[0103] The industrial concept node expansion module is used to obtain the industrial chain map skeleton composed of key industrial concept nodes pre-built based on expert knowledge for the target industrial chain; for each key industrial concept node in the industrial chain map skeleton, extract the industrial concept nodes associated with this key industrial concept node from the industrial knowledge network and add them to the first node set, and at the same time, after pairing this key industrial concept node with each industrial concept node in the industrial knowledge network, use the big model to generate node representation and calculate the semantic similarity between nodes, and then add all industrial concept nodes with semantic similarity exceeding the second threshold to the second node set, and finally attach the nodes in the first node set and the second node set as the lower nodes of this key industrial concept node to the industrial chain map skeleton; after all the key industrial concept nodes in the industrial chain map skeleton complete the lower node attachment growth, use the big model technology to improve the node portrait for each industrial concept node and update it to the node attribute, obtain the complete node hierarchy structure of the industrial chain map and send it to the expert end for expert inspection, and modify the node hierarchy according to the results returned by the expert end;

[0104] The industrial chain element mounting module is used to sequentially search for industrial chain element nodes associated with each industrial concept node in the node hierarchy of the industrial chain map in the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with the industrial concept node, then use the large model to generate node representations and calculate the semantic similarity between nodes, use the semantic similarity between each industrial chain element node and the industrial concept node as the confidence of the industrial chain element node, send the industrial chain element nodes whose confidence does not exceed the third threshold to the operation review end for manual operation review, and correct the industrial chain element nodes finally associated and mounted on the industrial concept node according to the results returned by the operation review end; after all industrial concept nodes complete the association and mounting of industrial chain element nodes, the construction of the industrial chain map is completed;

[0105] The graph visualization module is used to visualize the industrial chain graph according to the visualization form specified by the user.

[0106] In addition, based on the same inventive concept, the present invention also provides a computer electronic device corresponding to the method for constructing and visualizing an industrial chain map based on human-machine collaboration provided in the above embodiment, which includes a memory and a processor;

[0107] The memory is used to store computer programs;

[0108] The processor is used to implement the above-mentioned method for constructing and visualizing an industrial chain map based on human-machine collaboration when executing the computer program;

[0109] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0110] In addition, based on the same inventive concept, the present invention provides a computer-readable storage medium corresponding to an industrial chain map construction and visualization method based on human-computer collaboration, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it can realize the industrial chain map construction and visualization method based on human-computer collaboration as described above.

[0111] In addition, based on the same inventive concept, the present invention provides a computer program product, including a computer program / instructions, which, when executed by a processor, can implement the industrial chain map construction and visualization method based on human-computer collaboration as described above.

[0112] Specifically, in the computer-readable storage medium of each of the above embodiments, the stored computer program is executed by the processor to perform the above steps S1 to S4.

[0113] It is understandable that the above storage medium may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. The storage medium may also be a U disk, a mobile hard disk, a magnetic disk or an optical disk, etc., which can store program codes.

[0114] It can be understood that the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0115] It should also be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the various embodiments provided in this application, the division of steps or modules in the system and method is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or steps can be combined or integrated together, and a module or step can also be split.

[0116] The above-described embodiments are only some preferred implementations of the present invention, but are not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for constructing and visualizing an industrial chain map based on human-machine collaboration, characterized in that: include: S1. Obtain the pre-constructed industrial knowledge base in the target industry field, and use natural language processing technology to extract information from the data in the industrial knowledge base, and construct an industrial knowledge network based on the extracted objects and the relationship between the objects; the objects of information extraction include at least industrial concepts and industrial chain elements, and each extracted relationship has a confidence level; verify the constructed industrial knowledge network, send the tuples with a confidence level lower than the first threshold or conflicting tuples to the operation review end for manual operation review, and modify the industrial knowledge network based on the results returned by the operation review end; S2. For the target industrial chain, obtain the industrial chain map skeleton composed of key industrial concept nodes pre-constructed based on expert knowledge; for each key industrial concept node in the industrial chain map skeleton, extract the industrial concept nodes associated with this key industrial concept node from the industrial knowledge network and add them to the first node set, and at the same time, pair this key industrial concept node with each industrial concept node in the industrial knowledge network respectively, generate node representation using the big model and calculate the semantic similarity between nodes, then add all industrial concept nodes whose semantic similarity exceeds the second threshold to the second node set, and finally attach the nodes in the first node set and the second node set as the subordinate nodes of this key industrial concept node to the industrial chain map skeleton; after all the key industrial concept nodes in the industrial chain map skeleton complete the growth of subordinate nodes, use the big model technology to improve the node portrait for each industrial concept node and update it to the node attribute, obtain the complete node hierarchy structure of the industrial chain map and send it to the expert end for expert review, and modify the node hierarchy according to the results returned by the expert end; S3. For each industrial concept node in the node hierarchy of the industrial chain map, search for industrial chain element nodes associated with this industrial concept node in the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with this industrial concept node, then use the big model to generate node representation and calculate the semantic similarity between nodes, use the semantic similarity between each industrial chain element node and this industrial concept node as the confidence of this industrial chain element node, send the industrial chain element nodes whose confidence does not exceed the third threshold to the operation review end for manual operation review, and correct the industrial chain element nodes finally associated and mounted on the industrial concept node according to the results returned by the operation review end; after all industrial concept nodes complete the association and mounting of industrial chain element nodes, the construction of the industrial chain map is completed; S4. Visually display the industrial chain map according to the visualization format specified by the user.

2. The method for constructing and visualizing an industrial chain map based on human-machine collaboration according to claim 1, characterized in that: The method for constructing the industrial knowledge base is as follows: based on the industrial knowledge system pre-constructed for the target industry field, knowledge data of different dimensions in the industrial knowledge system are collected from external data sources through data collection and access technology, and structured storage is performed after data processing and fusion to form an industrial knowledge base.

3. The method for constructing and visualizing an industrial chain map based on human-machine collaboration according to claim 1, characterized in that: In S1, the method for constructing an industrial knowledge network based on an industrial knowledge base is as follows: for knowledge data of different dimensions in the industrial knowledge base, an information extraction model is used to extract tuples consisting of objects and relationships based on natural language processing technology, wherein the extracted objects include concepts, entities, events, and attributes related to the industry, and the relationships between objects all have confidence levels; the objects obtained by information extraction are used as network nodes, and the network nodes are associated and coupled according to the relationships obtained by information extraction to construct an industrial knowledge network.

4. The method for constructing and visualizing an industrial chain map based on human-machine collaboration according to claim 1, characterized in that: In S2, for each pair of paired key industry concept nodes and industry concept nodes, the node attribute fields of the key industry concept nodes are spliced ​​and sent to the big model to generate a first representation vector, and then the node attribute fields of the industry concept nodes are spliced ​​and sent to the big model to generate a second representation vector, and the similarity between the first representation vector and the second representation vector is calculated and used as the semantic similarity between the nodes of the current paired nodes; Preferably, the node portrait generated for each industry concept node using the big model technology includes an introduction to the industry concept, synonyms, industry classification, upstream products, and downstream products.

5. The method for constructing and visualizing an industrial chain map based on human-machine collaboration according to claim 1, characterized in that: In S3, for each industrial concept node in the industrial chain map and the industrial chain element node found in the industrial knowledge network, the node attribute fields of the industrial concept node need to be spliced ​​and sent to the big model to generate a third representation vector, and then the attribute fields of the industrial chain element node are also spliced ​​and sent to the big model to generate a fourth representation vector, and the similarity between the third representation vector and the fourth representation vector is calculated and used as the confidence of this industrial chain element node.

6. The method for constructing and visualizing an industrial chain map based on human-machine collaboration according to claim 1, characterized in that: In S4, when the industrial chain map is visualized, the visualization dimensions include one or more of the industrial chain structure dimension, the industrial chain distribution situation dimension and the industrial chain change trend dimension.

7. The method for constructing and visualizing an industrial chain map based on human-machine collaboration according to claim 1, characterized in that: The manual review results of the operational review end and the expert end need to be saved and constructed as an incremental learning data set for the artificial intelligence algorithm used to generate the original data to be reviewed, so as to optimize and improve the artificial intelligence algorithm.

8. An industrial chain map construction and visualization system based on human-machine collaboration, characterized in that: include: The industrial knowledge network construction module is used to obtain the pre-constructed industrial knowledge base in the target industry field, and use natural language processing technology to extract information from the data in the industrial knowledge base, and construct an industrial knowledge network based on the extracted objects and the relationship between the objects; the objects of information extraction include at least industrial concepts and industrial chain elements, and each extracted relationship has a confidence level; the constructed industrial knowledge network is verified, and the tuples with a confidence level lower than the first threshold or conflicts are sent to the operation review end for manual operation review, and the industrial knowledge network is corrected according to the results returned by the operation review end; The industrial concept node expansion module is used to obtain the industrial chain map skeleton composed of key industrial concept nodes pre-built based on expert knowledge for the target industrial chain; for each key industrial concept node in the industrial chain map skeleton, extract the industrial concept nodes associated with this key industrial concept node from the industrial knowledge network and add them to the first node set, and at the same time, after pairing this key industrial concept node with each industrial concept node in the industrial knowledge network, use the big model to generate node representation and calculate the semantic similarity between nodes, and then add all industrial concept nodes with semantic similarity exceeding the second threshold to the second node set, and finally attach the nodes in the first node set and the second node set as the lower nodes of this key industrial concept node to the industrial chain map skeleton; after all the key industrial concept nodes in the industrial chain map skeleton complete the lower node attachment growth, use the big model technology to improve the node portrait for each industrial concept node and update it to the node attribute, obtain the complete node hierarchy structure of the industrial chain map and send it to the expert end for expert inspection, and modify the node hierarchy according to the results returned by the expert end; The industrial chain element mounting module is used to sequentially search for industrial chain element nodes associated with each industrial concept node in the node hierarchy of the industrial chain map in the industrial knowledge network, add these industrial chain element nodes to the industrial chain map and associate them with the industrial concept node, then use the large model to generate node representations and calculate the semantic similarity between nodes, use the semantic similarity between each industrial chain element node and the industrial concept node as the confidence of the industrial chain element node, send the industrial chain element nodes whose confidence does not exceed the third threshold to the operation review end for manual operation review, and correct the industrial chain element nodes finally associated and mounted on the industrial concept node according to the results returned by the operation review end; after all industrial concept nodes complete the association and mounting of industrial chain element nodes, the construction of the industrial chain map is completed; The graph visualization module is used to visualize the industrial chain graph according to the visualization form specified by the user.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the method for constructing and visualizing an industrial chain map based on human-machine collaboration as described in any one of claims 1 to 7 is implemented.

10. A computer electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is used to implement the industrial chain map construction and visualization method based on human-computer collaboration as described in any one of claims 1 to 7 when executing the computer program.

Images