Method, system and device for creating knowledge graph in industrial field, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a method and a device for creating a knowledge graph in an industrial field, electronic equipment and a medium. The method comprises the steps of obtaining first unstructured data from a first data source in a first sub-field, wherein the first unstructured data carries knowledge annotations; executing machine learning on the first unstructured data to generate a first model suitable for extracting knowledge; based on the first model, extracting first knowledge from second unstructured data provided by the first data source; obtaining first structured data and first semi-structured data from a second data source in the second sub-field; extracting second knowledge from the first structured data; extracting third knowledge from the first semi-structured data; and constructing a knowledge graph fusing the first sub-field and the second sub-field based on the first knowledge, the second knowledge and the third knowledge. This facilitates the construction of a knowledge graph that fuses a plurality of sub-fields.

Description

Technical Field

[0001] The present invention relates to the industrial field, and in particular to methods, devices, electronic devices and media for creating knowledge graphs in the industrial field. Background Art

[0002] Knowledge Graph describes concepts, entities and their relationships in the objective world in a structured form. It is an interdisciplinary subject that integrates cognitive computing, knowledge representation and reasoning, information retrieval and extraction, natural language processing, Web technology, machine learning and big data mining.

[0003] In recent years, the industrial field has begun to favor the use of knowledge graphs as a graphical display method to achieve applications such as industrial equipment data collection, knowledge classification, knowledge storage, and knowledge analysis. Introducing knowledge graphs into the industrial field solves the problem of knowledge construction in the industrial field. However, information in the industrial field is more chaotic than in other fields, and much of the information comes from unstructured text. Therefore, it is extremely important to perform semantic analysis on this information, extract knowledge units that can be used to draw knowledge graphs, and find the connections between knowledge units.

[0004] However, building a knowledge graph requires a lot of work. Summary of the invention

[0005] The embodiments of the present invention propose a method, device, electronic device and medium for creating a knowledge graph in the industrial field.

[0006] In a first aspect, a method for creating a knowledge graph in an industrial field is provided. The method comprises:

[0007] Acquire first unstructured data from a first data source in a first sub-field of an industrial field, wherein the first unstructured data is annotated with knowledge;

[0008] performing machine learning on the first unstructured data to generate a first model suitable for extracting knowledge;

[0009] Extracting first knowledge from second unstructured data provided by the first data source based on the first model, wherein the second unstructured data does not have knowledge annotations;

[0010] Acquire first structured data and first semi-structured data from a second data source in a second sub-domain of the industrial domain;

[0011] extracting second knowledge from the first structured data;

[0012] extracting tertiary knowledge from the first semi-structured data; and

[0013] Based on the first knowledge, the second knowledge and the third knowledge, a knowledge graph integrating the first sub-field and the second sub-field is constructed, wherein the first knowledge, the second knowledge and the third knowledge are all represented in the form of triples.

[0014] Therefore, a knowledge graph that integrates multiple sub-fields can be constructed.

[0015] Preferably, the proportion of unstructured data in the first sub-field is greater than a predetermined threshold, and the proportion of unstructured data in the second sub-field is less than the threshold.

[0016] Therefore, for sub-fields where unstructured data accounts for a large proportion, knowledge is extracted from unstructured data, and for sub-fields where unstructured data accounts for a small proportion, knowledge is extracted from structured data or semi-structured data. Therefore, the knowledge graph obtained through fusion achieves a good compromise between the breadth of the field, the comprehensiveness of the data, and the difficulty of implementation.

[0017] Preferably, the first sub-field and the second sub-field belong to the same industrial category.

[0018] Therefore, for sub-fields that are close to each other, a good compromise effect can be achieved through this differentiated knowledge extraction strategy.

[0019] Preferably, the first sub-field and the second sub-field belong to the same industrial sub-category.

[0020] Therefore, for sub-fields that are particularly close to each other, a better compromise effect can be achieved through this differentiated knowledge extraction strategy.

[0021] Preferably, the method further comprises:

[0022] Acquire second structured data and second semi-structured data from the first data source;

[0023] extracting fourth knowledge from the second structured data;

[0024] Extracting fifth knowledge from second semi-structured data;

[0025] Based on the first knowledge, the second knowledge and the third knowledge, the knowledge graph integrating the first sub-field and the second sub-field is constructed, including:

[0026] Based on the first knowledge, the second knowledge, the third knowledge, the fourth knowledge and the fifth knowledge, a knowledge graph is constructed.

[0027] Therefore, the knowledge graph can be further enriched.

[0028] Preferably, constructing a knowledge graph integrating the first sub-field and the second sub-field includes:

[0029] Based on the first knowledge, the fourth knowledge and the fifth knowledge, a knowledge graph of the first sub-field is constructed;

[0030] Based on the second knowledge and the third knowledge, construct a knowledge graph of the second sub-field;

[0031] Combining the knowledge graph of the first subfield and the knowledge graph of the second subfield into the knowledge graph of the first subfield and the second subfield;

[0032] The attributes of the entities in the knowledge graph of the first subfield are compared with the corresponding attributes of the entities in the knowledge graph of the second subfield to determine the similarity between the entities in the knowledge graph of the first subfield and the entities in the knowledge graph of the second subfield; when the similarity is higher than a preset threshold, the entities in the knowledge graph of the first subfield and the entities in the knowledge graph of the second subfield are combined.

[0033] Therefore, by combining entities with high similarity, the accuracy of the knowledge graph is improved.

[0034] In a second aspect, a device for creating a knowledge graph in an industrial field is provided, the device comprising:

[0035] A first acquisition module is configured to acquire first unstructured data from a first data source in a first sub-field of an industrial field, wherein the first unstructured data is annotated with knowledge;

[0036] an execution module configured to perform machine learning on the first unstructured data to generate a first model suitable for extracting knowledge;

[0037] A first extraction module is configured to extract first knowledge from second unstructured data provided by the first data source based on the first model, wherein the second unstructured data does not have knowledge annotations;

[0038] A second acquisition module configured to acquire the first structured data and the first semi-structured data from a second data source in a second sub-field of the industrial field;

[0039] a second extraction module configured to extract second knowledge from the first structured data;

[0040] a third extraction module configured to extract third knowledge from the first semi-structured data; and

[0041] A building module is configured to build a knowledge graph integrating the first sub-field and the second sub-field based on the first knowledge, the second knowledge and the third knowledge, wherein the first knowledge, the second knowledge and the third knowledge are all represented in the form of triples.

[0042] Therefore, a knowledge graph that integrates multiple sub-fields can be constructed.

[0043] Preferably, the proportion of unstructured data in the first sub-field is greater than a predetermined threshold, and the proportion of unstructured data in the second sub-field is less than the threshold.

[0044] Therefore, for sub-fields where unstructured data accounts for a large proportion, knowledge is extracted from unstructured data, and for sub-fields where unstructured data accounts for a small proportion, knowledge is extracted from structured data or semi-structured data. Therefore, the knowledge graph obtained through fusion achieves a good compromise between the breadth of the field, the comprehensiveness of the data, and the difficulty of implementation.

[0045] Preferably, the first sub-field and the second sub-field belong to the same industrial category.

[0046] Therefore, for sub-fields that are close to each other, a good compromise effect can be achieved through this differentiated knowledge extraction strategy.

[0047] Preferably, the first sub-field and the second sub-field belong to the same industrial sub-category.

[0048] Therefore, for sub-fields that are particularly close to each other, a better compromise effect can be achieved through this differentiated knowledge extraction strategy.

[0049] Preferably, wherein:

[0050] The first acquisition module is configured to acquire second structured data and second semi-structured data from the first data source;

[0051] The first extraction module is configured to extract fourth knowledge from the second structured data and fifth knowledge from the second semi-structured data;

[0052] The building module is configured to build a knowledge graph based on the first knowledge, the second knowledge, the third knowledge, the fourth knowledge and the fifth knowledge.

[0053] Therefore, the knowledge graph can be further enriched.

[0054] Preferably, the construction module is configured to construct a knowledge graph of the first sub-field based on the first knowledge, the fourth knowledge and the fifth knowledge; construct a knowledge graph of the second sub-field based on the second knowledge and the third knowledge; combine the knowledge graph of the first sub-field and the knowledge graph of the second sub-field into knowledge graphs of the first sub-field and the second sub-field; wherein the attributes of the entities in the knowledge graph of the first sub-field are compared with the corresponding attributes of the entities in the knowledge graph of the second sub-field to determine the similarity between the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field; when the similarity is higher than a preset threshold, combine the entities in the knowledge graph of the first sub-field with the entities in the knowledge graph of the second sub-field.

[0055] Therefore, by combining entities with high similarity, the accuracy of the knowledge graph is improved.

[0056] In a third aspect, an electronic device is provided, comprising a processor and a memory, wherein an application executable by the processor is stored in the memory, so as to enable the processor to execute the method for creating a knowledge graph in an industrial field according to any one of the above items.

[0057] In a fourth aspect, a computer-readable medium is provided, comprising computer-readable instructions stored thereon, wherein the computer-readable instructions are used to execute a method for creating a knowledge graph in an industrial field according to any one of the above items.

[0058] In a fifth aspect, a computer program product is provided, comprising a computer program, which, when executed by a processor, performs a method for creating a knowledge graph in an industrial field according to any one of the above items. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to make the technical solutions of the examples of the present disclosure clearer, the drawings used in the description of the examples are briefly introduced below. Obviously, the drawings described below are only some examples of the present disclosure. Those skilled in the art can obtain other drawings based on these drawings without creative work.

[0060] Figure 1 It is a flowchart of a method for creating a knowledge graph in an industrial field according to an embodiment of the present invention.

[0061] Figure 2 It is a schematic diagram of the process of creating a knowledge graph in an industrial field according to an embodiment of the present invention.

[0062] Figure 3 is a schematic diagram of extracting knowledge according to an embodiment of the present invention.

[0063] Figure 4 It is a structural diagram of a device for creating a knowledge graph in an industrial field according to an embodiment of the present invention.

[0064] Figure 5 is a structural diagram of an electronic device according to an embodiment of the present invention.

[0065] List of reference numerals:

[0066]

[0067] DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following examples are given to further illustrate the present invention in detail.

[0069] For the sake of brevity and intuitiveness in description, the scheme of the present invention is described below by describing several representative embodiments. A large number of details in the embodiments are only used to help understand the scheme of the present invention. However, it is obvious that the technical solution of the present invention may not be limited to these details when implemented. In order to avoid unnecessarily obscuring the scheme of the present invention, some embodiments are not described in detail, but only a framework is given. Hereinafter, "including" means "including but not limited to", and "according to..." means "at least according to..., but not limited to...". Due to the language habits of Chinese, when the number of components is not specifically stated below, it means that the components can be one or more, or can be understood as at least one.

[0070] With the rapid development of digitalization, unstructured data plays an increasingly important role. However, knowledge extraction based on unstructured data is difficult. For industrial fields with a wide variety of industries, it is of great significance to achieve a knowledge graph construction solution that strikes a balance between the breadth of the field, the comprehensiveness of the data, and the difficulty of implementation.

[0071] Figure 1 It is a flowchart of a method for creating a knowledge graph in an industrial field according to an embodiment of the present invention.

[0072] like Figure 1 As shown, the method includes:

[0073] Step 101: Acquire first unstructured data from a first data source in a first sub-field of an industrial field, wherein the first unstructured data is annotated with knowledge.

[0074] The industrial field can be divided into multiple sub-fields according to different classification methods.

[0075] For example, industry can be divided into the following sub-fields: metallurgical industry, electric power industry, coal and coking industry, petroleum industry, chemical industry, machinery industry, building materials industry, forestry industry, food industry, textile industry, sewing industry, leather industry and other industries.

[0076] For example, based on the nature of the products, the industrial sector can be divided into the following sub-sectors: light industry and heavy industry.

[0077] For example, based on the relative concentration of the three production factors of labor, capital and technology in each industry, the industrial sector can be divided into the following sub-sectors: labor-intensive, capital-intensive and technology-intensive industries.

[0078] For example, according to the commonly used industrial classification catalog in the industry, the industrial field can be divided into 39 industrial categories, 191 medium categories and 525 subcategories. Any category, medium category, subcategory or combination thereof can constitute a subfield.

[0079] Moreover, those skilled in the art may further divide the sub-fields according to actual needs, and the embodiments of the present invention are not limited in this regard.

[0080] The first unstructured data obtained from the first data source in the first sub-field may specifically be text data such as operation specification descriptions. The first unstructured data has knowledge annotations. The knowledge annotations may include at least one of entity annotations, entity relationship annotations, and entity attribute annotations. Based on the first unstructured data with knowledge annotations, an artificial intelligence model is trained through machine learning, and the trained artificial intelligence model is suitable for extracting knowledge from the unstructured data.

[0081] The unstructured data may be obtained by crawling the network, for example, by crawling according to keywords, and the unstructured data may be manually marked to obtain the first unstructured data.

[0082] Step 102: Perform machine learning on the first unstructured data to generate a first model suitable for extracting knowledge.

[0083] Here, the machine learning algorithm trains an artificial neural network model into a first model using the first unstructured data.

[0084] Step 103: extracting first knowledge from second unstructured data provided by the first data source based on the first model, wherein the second unstructured data does not carry knowledge annotations.

[0085] Step 104: Acquire first structured data and first semi-structured data from a second data source in a second sub-field of the industrial field.

[0086] Here, the unstructured data in the second sub-domain may not be acquired.

[0087] Step 105: extracting second knowledge from the first structured data.

[0088] Step 106: extracting the third knowledge from the first semi-structured data. For example, the third knowledge may be extracted from the first semi-structured data by using a regular expression.

[0089] Step 107: Based on the first knowledge, the second knowledge and the third knowledge, a knowledge graph integrating the first sub-field and the second sub-field is constructed, wherein the first knowledge, the second knowledge and the third knowledge are all represented in the form of triples.

[0090] For example, a triple may contain: (entity-entity relationship-entity) or (entity-attribute-attribute value).

[0091] An ontology library in the industrial field can be pre-built, or an ontology library built based on an existing mechanism can be obtained. The ontology library includes an ontology built according to the knowledge system of the first sub-field and an ontology built according to the knowledge system of the second sub-field. The knowledge system is a regulation of knowledge naming and knowledge classification, wherein the knowledge annotations in the first unstructured data match the ontology library.

[0092] Specifically, the process of constructing a knowledge graph based on the first knowledge, the second knowledge, and the third knowledge may include: establishing entity nodes corresponding to entities in the first knowledge, the second knowledge, and the third knowledge; establishing connection relationships between entity nodes based on entity relationships in the first knowledge, the second knowledge, and the third knowledge; and setting attributes and attribute values ​​of entity nodes based on attributes and attribute values ​​in the first knowledge, the second knowledge, and the third knowledge. Since the knowledge annotations of the first unstructured data carry a large amount of credible knowledge, the knowledge annotations can be added to the knowledge graph at the same time.

[0093] The knowledge graph generated in step 107 includes knowledge of both the first sub-field and the second sub-field. There is no need to perform unstructured knowledge extraction processing that requires a lot of calculations on the second sub-field. A good compromise is achieved between data comprehensiveness and implementation difficulty.

[0094] In one embodiment, the proportion of unstructured data in the first sub-field is greater than a predetermined threshold, and the proportion of unstructured data in the second sub-field is less than the threshold. For example, the threshold is 80%. For example, the ratio of the amount of unstructured data in the first data source to the total amount of data in the first data source is used as a basis for judgment. When the ratio exceeds 80%, it is determined that the proportion of unstructured data in the first sub-field is greater than the threshold. Similarly, the ratio of the amount of unstructured data in the second data source to the total amount of data in the second data source is also used as a basis for judgment. When the ratio is less than 80%, it is determined that the proportion of unstructured data in the second sub-field is less than the threshold.

[0095] For the first sub-field where the proportion of unstructured data in the total data volume is relatively large, knowledge is extracted from the unstructured data through machine learning-based modeling. However, for the second sub-field where the proportion of unstructured data in the total data volume is relatively small, it is not necessary to extract knowledge from the unstructured data through modeling, but directly extract structured data and unstructured data, thereby achieving a knowledge graph creation method that achieves a good compromise between the breadth of the field and the difficulty of implementation. In addition, the knowledge graph obtained by combination also has comprehensive data.

[0096] In one embodiment, the first subfield and the second subfield belong to the same industrial category. Therefore, for two subfields in the same industrial category, considering the homogeneity of the two categories, a good compromise effect can be achieved through this differentiated knowledge extraction method.

[0097] In one embodiment, the first subfield and the second subfield belong to the same industrial subcategory. Therefore, for two subfields in the same industrial subcategory, considering the significant homogeneity of the two subcategories, a better compromise effect can be achieved through this differentiated knowledge extraction method.

[0098] In one embodiment, the method further includes: acquiring second structured data and second semi-structured data from the first data source; extracting fourth knowledge from the second structured data; extracting fifth knowledge from the second semi-structured data; wherein constructing a knowledge graph integrating the first sub-field and the second sub-field based on the first knowledge, the second knowledge and the third knowledge includes: constructing a knowledge graph based on the first knowledge, the second knowledge, the third knowledge, the fourth knowledge and the fifth knowledge. Therefore, the knowledge graph can be further enriched.

[0099] In one embodiment, constructing a knowledge graph integrating the first sub-field and the second sub-field includes: constructing a knowledge graph of the first sub-field based on the first knowledge, the fourth knowledge and the fifth knowledge; constructing a knowledge graph of the second sub-field based on the second knowledge and the third knowledge; combining the knowledge graph of the first sub-field and the knowledge graph of the second sub-field into the knowledge graph of the first sub-field and the knowledge graph of the second sub-field; wherein the attributes of the entities in the knowledge graph of the first sub-field are compared with the corresponding attributes of the entities in the knowledge graph of the second sub-field to determine the similarity between the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field; when the similarity is higher than a preset threshold, combining the entities in the knowledge graph of the first sub-field with the entities in the knowledge graph of the second sub-field. Therefore, by combining entities with high similarity, the accuracy of the knowledge graph is improved.

[0100] Figure 2 It is a schematic diagram of the process of creating a knowledge graph in the industrial field according to an embodiment of the present invention.

[0101] like Figure 2 As shown, unstructured data 21, structured data 22, and semi-structured data 23 are obtained from the first data source 11 in the first sub-field of the industrial field. The unstructured data 21, structured data 22, and semi-structured data 23 are all data in the first sub-field. For example, the unstructured data 21 may include industrial behavior specification text. The structured data 23 may be data in a database. The semi-structured data 23 may be partially structured data such as tables and encyclopedia data.

[0102] The structured data 24 and the semi-structured data 25 are obtained from the second data source 12 in the second sub-domain. The structured data 24 and the semi-structured data 25 are both data in the second sub-domain. For example, the structured data 24 can be data in a database. The semi-structured data 25 can be partially structured data such as tables and encyclopedia data.

[0103] Extract part of the data from the unstructured data 21 provided by the first data source 11 as a training set. For example, the unstructured data 21 is divided into a ratio of 1:3, and the part with a ratio of 1 is the training set. The part with a weight of 3 is used to extract knowledge. Add knowledge annotations to the data in the training set, and the knowledge annotations match the ontology library containing the corresponding knowledge system of the first sub-field and the second sub-field. For example, the ontology library is an ontology library established for all industries. Knowledge annotations may include at least one of entity annotations, entity relationship annotations, and entity attribute annotations.

[0104] The machine learning model is trained using the data in the training set to obtain the knowledge extraction model 26. Then, knowledge is extracted from the data other than the training set in the unstructured data 21 using the knowledge extraction model 26. The extracted knowledge can be used as the first knowledge 31, or the extracted knowledge can be combined with the knowledge annotations in the training set into the first knowledge 31.

[0105] The fourth knowledge 32 is extracted from the structured data 22 provided by the first data source 11. The fifth knowledge 33 is extracted from the semi-structured data 23 provided by the first data source 11. The second knowledge 34 is extracted from the structured data 24 originating from the second data source 12, and the third knowledge 35 is extracted from the semi-structured data 25 originating from the second data source 12. The first knowledge 31, the second knowledge 34, the third knowledge 35, the fourth knowledge 32 and the fifth knowledge 33 are all represented in the form of triples.

[0106] Based on the first knowledge 31, the fourth knowledge 32 and the fifth knowledge 33, a first knowledge graph 36 of the first sub-domain is constructed. Based on the second knowledge 34 and the third knowledge 35, a second knowledge graph 37 of the second sub-domain is constructed. The process of constructing a knowledge graph based on knowledge may include: establishing entity nodes corresponding to entities in the knowledge; establishing connection relationships between entity nodes based on entity relationships in the knowledge; and setting attributes and attribute values ​​of entity nodes based on attributes and attribute values ​​in the knowledge.

[0107] The first knowledge graph 36 and the second knowledge graph 37 are combined to obtain a third knowledge graph 38 applicable to the first sub-field and the second sub-field. It is understandable that an entity has attributes, and if the attributes of two entities are highly similar, the two entities can be determined to be one entity, and therefore combined. In the combination process, for any entity in the first knowledge graph 36: the attributes of the entity in the knowledge graph of the first sub-field are compared with the corresponding attributes of the entity in the knowledge graph of the second sub-field to determine the similarity between the entity in the knowledge graph of the first sub-field and the entity in the knowledge graph of the second sub-field; when the similarity is higher than a preset threshold, the entity in the knowledge graph of the first sub-field is combined with the entity in the knowledge graph of the second sub-field.

[0108] For example, assume that the first knowledge graph 36 includes entity A, and the second knowledge graph 36 includes entity B, entity C, and entity D. The attributes of entity A are compared with the corresponding attributes of entity B, entity C, and entity D, and it is found that the similarity between entity A and entity B is greater than a predetermined threshold (e.g., 85%), then entity A and entity B are combined into one entity in the third knowledge graph 38.

[0109] Figure 3 FIG. 1 is a schematic diagram of extracting knowledge according to an embodiment of the present invention. Figure 3 In the example, a data source that provides unstructured and structured data is used for illustration.

[0110] exist Figure 3 In the process, data is first obtained from a data source 40. Then, a data parsing process 41 is performed on the data.

[0111] When the parsing result indicates that the data is unstructured data 42 , the unstructured data 42 is input into a trained knowledge extraction model 43 to obtain knowledge 44 represented by triples. The knowledge 44 is stored in a Neo4j database 45 .

[0112] When the parsing result indicates that the data is structured data 50, knowledge extraction is performed to extract knowledge 51 from the structured data 50. Manual knowledge addition 52 is performed on the knowledge 51 to obtain knowledge 53 including the knowledge 51 and the manually added knowledge represented by triples. Knowledge 44 is retrieved from the Neo4j database 45. Knowledge 44 is combined with knowledge 53 into aggregated knowledge 54 of the data source 40.

[0113] Figure 4 1 is a structural diagram of a device for creating a knowledge graph in an industrial field according to an embodiment of the present invention. The device includes:

[0114] A first acquisition module 401 is configured to acquire first unstructured data from a first data source in a first sub-field of an industrial field, wherein the first unstructured data is annotated with knowledge;

[0115] An execution module 402 configured to perform machine learning on the first unstructured data to generate a first model suitable for extracting knowledge;

[0116] A first extraction module 403 is configured to extract first knowledge from second unstructured data provided by the first data source based on the first model, wherein the second unstructured data does not have knowledge annotations;

[0117] A second acquisition module 404 is configured to acquire the first structured data and the first semi-structured data from a second data source in a second sub-field of the industrial field;

[0118] A second extraction module 405 configured to extract second knowledge from the first structured data;

[0119] a third extraction module 406 configured to extract third knowledge from the first semi-structured data; and

[0120] The construction module 407 is configured to construct a knowledge graph integrating the first sub-field and the second sub-field based on the first knowledge, the second knowledge and the third knowledge, wherein the first knowledge, the second knowledge and the third knowledge are all represented in the form of triples.

[0121] In one embodiment, the proportion of unstructured data in the first sub-field is greater than a predetermined threshold, and the proportion of unstructured data in the second sub-field is less than the threshold.

[0122] In one embodiment, the first sub-field and the second sub-field belong to the same industrial category. In one embodiment, the first sub-field and the second sub-field belong to the same industrial sub-category.

[0123] In one embodiment, the first acquisition module 401 is configured to acquire second structured data and second semi-structured data from the first data source; the first extraction module 403 is configured to extract fourth knowledge from the second structured data and fifth knowledge from the second semi-structured data; the construction module 407 is configured to construct a knowledge graph based on the first knowledge, the second knowledge, the third knowledge, the fourth knowledge and the fifth knowledge.

[0124] In one embodiment, the construction module 407 is configured to construct a knowledge graph of the first sub-field based on the first knowledge, the fourth knowledge and the fifth knowledge; construct a knowledge graph of the second sub-field based on the second knowledge and the third knowledge; combine the knowledge graph of the first sub-field and the knowledge graph of the second sub-field into knowledge graphs of the first sub-field and the second sub-field; compare the attributes of the entities in the knowledge graph of the first sub-field with the corresponding attributes of the entities in the knowledge graph of the second sub-field to determine the similarity between the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field; when the similarity is higher than a preset threshold, combine the entities in the knowledge graph of the first sub-field with the entities in the knowledge graph of the second sub-field.

[0125] An embodiment of the present invention further provides an electronic device having a processor-memory architecture. Figure 5 is a structural diagram of an electronic device according to an embodiment of the present invention.

[0126] like Figure 5 As shown, the electronic device 500 includes a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the computer program is executed by the processor 501, any of the above methods for creating a knowledge graph is implemented. The memory 502 can be specifically implemented as a variety of storage media such as an electrically erasable programmable read-only memory (EEPROM), a flash memory (Flash memory), and a programmable program read-only memory (PROM). The processor 801 can be implemented to include one or more central processing units or one or more field programmable gate arrays, wherein the field programmable gate array integrates one or more central processing unit cores. Specifically, the central processing unit or the central processing unit core can be implemented as a CPU or an MCU or a DSP, and so on.

[0127] It should be noted that not all steps and modules in the above process and structure diagram are necessary, and some steps or modules can be ignored according to actual needs. The execution logic of each step is not fixed and can be adjusted as needed. The division of each module is only for the convenience of describing the functional division adopted. In actual implementation, a module can be implemented by multiple modules, and the functions of multiple modules can also be implemented by the same module. These modules can be located in the same device or in different devices.

[0128] The hardware modules in each embodiment can be implemented mechanically or electronically. For example, the hardware module may include a specially designed permanent circuit or logic device (e.g., a dedicated processor, such as an FPGA or ASIC) to perform a specific operation. The hardware module may also include a programmable logic device or circuit (e.g., including a general-purpose processor or other programmable processor) temporarily configured by software to perform the operation. As for whether to use a mechanical method, a dedicated permanent circuit, or a temporarily configured circuit (e.g., configured by software) to implement the hardware module, it can be decided based on cost and time considerations.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for creating a knowledge graph in an industrial field, characterized in that: include: Acquire first unstructured data from a first data source in a first sub-field of the industrial field, wherein the first unstructured data is annotated with knowledge (101); Performing machine learning on the first unstructured data to generate a first model suitable for extracting knowledge (102); Extracting first knowledge from second unstructured data provided by the first data source based on the first model, wherein the second unstructured data does not have knowledge annotations (103); Acquire first structured data and first semi-structured data from a second data source in a second sub-domain of the industrial domain (104); Extracting second knowledge from the first structured data (105); extracting third knowledge from the first semi-structured data (106); as well as Based on the first knowledge, the second knowledge and the third knowledge, a knowledge graph integrating the first sub-field and the second sub-field is constructed, wherein the first knowledge, the second knowledge and the third knowledge are all represented in the form of triples (107).

2. The method according to claim 1, characterized in that The proportion of unstructured data in the first sub-field is greater than a predetermined threshold, and the proportion of unstructured data in the second sub-field is less than the threshold.

3. The method according to claim 2, characterized in that The first sub-field and the second sub-field belong to the same industrial category.

4. The method according to claim 3, characterized in that The first sub-field and the second sub-field belong to the same industrial sub-category.

5. The method according to any one of claims 1 to 4, characterized in that Also includes: Acquire second structured data and second semi-structured data from the first data source; extracting fourth knowledge from the second structured data; extracting fifth knowledge from the second semi-structured data; Wherein, based on the first knowledge, the second knowledge and the third knowledge, constructing a knowledge graph integrating the first sub-field and the second sub-field (107) comprises: The knowledge graph is constructed based on the first knowledge, the second knowledge, the third knowledge, the fourth knowledge and the fifth knowledge.

6. The method according to claim 5, characterized in that The constructing of a knowledge graph integrating the first sub-field and the second sub-field includes: Based on the first knowledge, the fourth knowledge and the fifth knowledge, construct a knowledge graph of the first sub-field; Based on the second knowledge and the third knowledge, construct a knowledge graph of the second sub-field; Combining the knowledge graph of the first subfield and the knowledge graph of the second subfield into the knowledge graphs of the first subfield and the second subfield; The attributes of the entities in the knowledge graph of the first sub-field are compared with the corresponding attributes of the entities in the knowledge graph of the second sub-field to determine the similarity between the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field; when the similarity is higher than a preset threshold, the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field are combined.

7. A device for creating a knowledge graph in an industrial field, characterized in that: include: A first acquisition module (401) is configured to acquire first unstructured data from a first data source in a first sub-field of the industrial field, wherein the first unstructured data is annotated with knowledge; an execution module (402) configured to perform machine learning on the first unstructured data to generate a first model suitable for extracting knowledge; A first extraction module (403) is configured to extract first knowledge from second unstructured data provided by the first data source based on the first model, wherein the second unstructured data does not have knowledge annotations; A second acquisition module (404) is configured to acquire the first structured data and the first semi-structured data from a second data source in a second sub-domain of the industrial domain; a second extraction module (405), configured to extract second knowledge from the first structured data; a third extraction module (406), configured to extract third knowledge from the first semi-structured data; as well as A construction module (407) is configured to construct a knowledge graph integrating the first sub-field and the second sub-field based on the first knowledge, the second knowledge and the third knowledge, wherein the first knowledge, the second knowledge and the third knowledge are all represented in the form of triples.

8. The device according to claim 7, characterized in that The proportion of unstructured data in the first sub-field is greater than a predetermined threshold, and the proportion of unstructured data in the second sub-field is less than the threshold.

9. The device according to claim 8, characterized in that The first sub-field and the second sub-field belong to the same industrial category.

10. The device according to claim 9, characterized in that The first sub-field and the second sub-field belong to the same industrial sub-category.

11. The device according to claim 7, characterized in that: The first acquisition module (401) is configured to acquire second structured data and second semi-structured data from the first data source; The first extraction module (403) is configured to extract fourth knowledge from the second structured data and fifth knowledge from the second semi-structured data; The construction module (407) is configured to construct the knowledge graph based on the first knowledge, the second knowledge, the third knowledge, the fourth knowledge and the fifth knowledge.

12. The device according to claim 11, characterized in that The construction module (407) is configured to construct a knowledge graph of the first sub-field based on the first knowledge, the fourth knowledge and the fifth knowledge; construct a knowledge graph of the second sub-field based on the second knowledge and the third knowledge; combine the knowledge graph of the first sub-field and the knowledge graph of the second sub-field into knowledge graphs of the first sub-field and the second sub-field; The attributes of the entities in the knowledge graph of the first sub-field are compared with the corresponding attributes of the entities in the knowledge graph of the second sub-field to determine the similarity between the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field; when the similarity is higher than a preset threshold, the entities in the knowledge graph of the first sub-field and the entities in the knowledge graph of the second sub-field are combined.

13. An electronic device, characterized in that: It comprises a processor (501) and a memory (502), wherein an application program executable by the processor (501) is stored in the memory (502) so as to enable the processor (501) to execute a method for creating a knowledge graph for an industrial field according to any one of claims 1 to 6.

14. A computer readable medium, characterized in that The invention comprises computer-readable instructions stored thereon, wherein the computer-readable instructions are used for a method for creating a knowledge graph in an industrial field according to any one of claims 1 to 6.

15. A computer program product, characterized in that It comprises a computer program which, when executed by a processor, performs a method for creating a knowledge graph in an industrial field according to any one of claims 1 to 6.

Images