Industrial SOP identification system and method based on electronic industry large model

Abstract

The invention relates to the technical field of data processing, and discloses an industrial SOP recognition system and method based on an electronic industry large model, and the method comprises the steps: extracting knowledge data from an electronic industry data source, and constructing an original knowledge graph according to the knowledge data; the node degree of each original node in the original knowledge graph is calculated, all isolated nodes are screened out according to the size of the node degrees, and after a plurality of associated nodes related to the isolated nodes are obtained, relation mining is conducted on the isolated nodes and the associated nodes to obtain a plurality of second relations; performing structure optimization on the original mapping knowledge domain according to the second relationship, and adding a new relationship edge for the isolated node to obtain a large mapping knowledge domain model applied to the electronic industry; inputting the to-be-queried request into the knowledge graph large model to identify corresponding SOP data; according to the method and the device, the query accuracy of the SOP knowledge graph is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to an industrial SOP recognition system and method based on a large model of the electronics industry. Background Art

[0002] Standard operating procedures (SOPs) are an important means to achieve standardized operations and ensure product quality in industrial production. SOPs provide systematic guidance for each link of the production process through clear operating steps and standardized operating requirements. However, the existing SOP management model mainly relies on static document forms, requiring operators to manually review paper SOP documents frequently to confirm operating steps or verify execution results. This is not only time-consuming and reduces operating efficiency, but also easily increases the risk of misoperation due to omissions or misunderstandings in manual review, thereby having a negative impact on production quality and efficiency.

[0003] With the development of knowledge graph technology, although it has promoted the transformation of paper SOP documents to industrial digitalization, the current construction method of SOP knowledge graph is still mainly rule-driven, and needs to rely on predefined rules to extract entities and relationships. However, unlike data in other fields, SOP data is large in volume and covers a wide range. This leads to incomplete entity and relationship construction in the existing SOP knowledge graph, and fails to cover the diverse knowledge points in the SOP data, further resulting in low query accuracy of the SOP knowledge graph, which limits its actual application effect in complex industrial scenarios and makes it difficult to meet the needs of efficient operation guidance and intelligent management. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an industrial SOP recognition system and method based on a large model of the electronics industry.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An industrial SOP recognition method based on a large model of the electronics industry, the method comprising:

[0007] Extracting knowledge data from electronic industry data sources and constructing an original knowledge graph based on the knowledge data; the knowledge data includes entities, relationships and attributes;

[0008] Calculate the node degree of each original node in the original knowledge graph, and filter out all isolated nodes according to the node degree. After obtaining multiple associated nodes related to the isolated nodes, perform relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships.

[0009] The original knowledge graph is structurally optimized according to the second relationship, and new relationship edges are added to isolated nodes to obtain a large knowledge graph model applied to the electronics industry.

[0010] Obtain the query request, input the query request into the knowledge graph model to identify the corresponding SOP data, and provide visual feedback; the query request at least contains keywords of entities or relationships.

[0011] Further, the entity includes a first entity and a second entity, and the relationship includes a first relationship and a second relationship;

[0012] The extraction is from knowledge data in the electronics industry data source, including:

[0013] Mark the words in the data source that are not the first entity as words to be selected;

[0014] Calculate the TF_IDF value of each candidate word in the data source;

[0015] The knowledge data is entity expanded according to the TF_IDF value to obtain multiple second entities.

[0016] Furthermore, the entity expansion of the knowledge data according to the TF_IDF value includes:

[0017] Compare the TF_IDF value of each candidate word with the preset TF_IDF value threshold;

[0018] If the TF_IDF value is greater than or equal to the preset TF_IDF threshold, the corresponding candidate word is marked as the second entity;

[0019] If the TF_IDF value is less than the preset TF_IDF threshold, the corresponding candidate word is marked as non-second entity;

[0020] All candidate words whose TF_IDF values ​​are greater than or equal to a preset TF_IDF threshold are counted to obtain a plurality of second entities.

[0021] Further, after obtaining the plurality of second entities, the method further comprises:

[0022] Extract all second entities and first entities;

[0023] Using the Word2Vec bag-of-words model, the first entity is converted into a first word vector, and the second entity is converted into a second word vector;

[0024] Calculate the semantic distance between all second word vectors and each first word vector, and compare the semantic distance with a preset semantic distance threshold; wherein the semantic distance is calculated by a cosine similarity algorithm;

[0025] If the semantic distance is greater than or equal to a preset semantic distance threshold, the second entity corresponding to the second word vector is removed from the multiple second entities;

[0026] If the semantic distance is less than a preset semantic distance threshold, the second entity corresponding to the second word vector is retained in the plurality of second entities.

[0027] Furthermore, constructing the original knowledge graph according to the knowledge data includes:

[0028] Construct entity-relationship-attribute triples based on knowledge data;

[0029] The first and second entities in the triple are taken as original nodes, and the first and second relations are taken as original edges between the original nodes to form an original knowledge graph.

[0030] Furthermore, the step of calculating the node degree of each original node in the original knowledge graph and filtering out isolated nodes according to the node degree includes:

[0031] Traverse all the original nodes in the original knowledge graph, and count the number of incoming edges for each original node and the number of outgoing edges ;

[0032] Number of incoming edges and the number of outgoing edges Perform sum calculation to obtain the node degree of each original node;

[0033] The node degree is compared with the preset node degree threshold. If the node degree is less than or equal to the preset node degree threshold, the corresponding original node is marked as an isolated node. If the node degree is greater than the preset node degree threshold, the corresponding original node is not marked as an isolated node.

[0034] Count all original nodes whose node degrees are less than or equal to the node degree threshold and obtain all isolated nodes.

[0035] Further, the acquiring of multiple associated nodes related to the isolated node includes:

[0036] Extract the isolated node and all the original nodes, and obtain the text distance between the isolated node and each original node in the data source;

[0037] Compare the text spacing with a preset text spacing threshold;

[0038] If the text spacing is less than or equal to the text spacing threshold, it is determined that the isolated node and the corresponding original node co-occur in the data source, and the corresponding original node is marked as an associated node related to the isolated node;

[0039] If the text spacing is greater than the text spacing threshold, it is determined that the isolated node and the corresponding original node do not co-occur in the data source, and the corresponding original node is not marked as an associated node related to the isolated node.

[0040] Furthermore, the relationship mining between isolated nodes and associated nodes includes:

[0041] Locate the target sentence where isolated nodes and associated nodes co-occur in the data source;

[0042] Construct a dependency tree for the target sentence after word segmentation to obtain dependency trees about isolated nodes and associated nodes;

[0043] Obtain the shortest dependency path of the isolated node and the associated node in the dependency tree, obtain all predicate verbs in the dependency path, and respectively parse the dependency relationship of each predicate verb with respect to the isolated node and the associated node, and obtain the weight of each dependency relationship to obtain multiple weights;

[0044] The multiple weights are sorted from large to small according to the numerical values, and the predicate verb corresponding to the first weight is used as the second relationship.

[0045] An industrial SOP recognition system based on a large model of the electronics industry is implemented based on the industrial SOP recognition method based on a large model of the electronics industry described above, and the system comprises:

[0046] A data processing module is used to extract knowledge data from electronic industry data sources and construct an original knowledge graph based on the knowledge data; the knowledge data includes entities, relationships and attributes;

[0047] The relationship mining module is used to calculate the node degree of each original node in the original knowledge graph, and filter out all isolated nodes according to the node degree size, and after obtaining multiple associated nodes related to the isolated nodes, perform relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships;

[0048] A data optimization module is used to optimize the structure of the original knowledge graph according to the second relationship, add new relationship edges to isolated nodes, and obtain a large knowledge graph model applied to the electronics industry;

[0049] The SOP identification module is used to obtain the query request, input the query request into the knowledge graph model to identify the corresponding SOP data, and provide visual feedback; the query request contains at least keywords of entities or relationships.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The present application discloses an industrial SOP identification system and method based on a large model of the electronics industry, comprising: extracting knowledge data from a data source of the electronics industry, and constructing an original knowledge graph based on the knowledge data; calculating the node degree of each original node in the original knowledge graph, and filtering out all isolated nodes based on the node degree, and after obtaining multiple associated nodes related to the isolated nodes, performing relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships; structurally optimizing the original knowledge graph based on the second relationships, adding new relationship edges to the isolated nodes, and obtaining a large model of the knowledge graph applied to the electronics industry; inputting a request to be queried into the large model of the knowledge graph to identify the corresponding SOP data; based on the above features, the present invention significantly reduces the operator's reliance on cumbersome static document review, and reduces the operational risks caused by missing information or misunderstandings. In addition, by dynamically expanding entities and relationships, and combining deep mining and optimization techniques of semantic associations, the problems of incomplete knowledge coverage and insufficient query accuracy caused by the limitations of rule base design in existing knowledge graph construction methods are overcome. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A flowchart of an industrial SOP identification method based on a large model of the electronics industry provided by the present invention;

[0053] Figure 2 A schematic diagram of the structure of a dependency tree provided by the present invention;

[0054] Figure 3 A schematic diagram of the module structure of an industrial SOP identification system based on a large model of the electronics industry provided by the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Example 1

[0057] See also Figure 1 As shown, this embodiment discloses an industrial SOP identification method based on a large model of the electronics industry, the method comprising:

[0058] S101: extracting knowledge data from electronic industry data sources, and constructing an original knowledge graph based on the knowledge data; the knowledge data includes entities, relationships, and attributes;

[0059] Wherein, the entity includes a first entity and a second entity, and the relationship includes a first relationship and a second relationship;

[0060] It should be noted that the data sources include but are not limited to data documents in paper or electronic document form in the electronics industry, such as technical manuals, operating procedures, product manuals, and equipment manuals, among which the data sources in paper document form are manually entered by technicians into electronic document form;

[0061] In a specific embodiment, the extracting of knowledge data from an electronics industry data source includes: calling a predefined rule, and extracting a first entity, a first relationship, and an attribute from the data source according to the predefined rule (such as a regular expression);

[0062] It should be understood that the predefined rules limit and define the extraction rules of the first entity, the first relationship and the attribute; for example, in the predefined rules, the word "temperature controller" is taken as the first entity, and there is a sentence in the data source that is "the temperature controller needs to be adjusted to 120°C", then the "temperature controller" in the corpus is extracted as the first entity, and similarly, the extraction logic of the first relationship and the attribute is also the same, which will not be elaborated in detail;

[0063] It should also be noted that: before extracting knowledge data from the electronic industry data source, the data source data will be preprocessed, and the preprocessing operation includes but is not limited to word segmentation, and the removal of meaningless words and symbols, such as repeated words, stop words, onomatopoeia, onomatopoeia, "," "!" and ".", etc.;

[0064] In another specific implementation, the extracting of knowledge data from an electronics industry data source further includes:

[0065] Mark the words in the data source that are not the first entity as words to be selected;

[0066] Calculate the TF_IDF value of each candidate word in the data source;

[0067] It should be noted that the calculation formula of TF_ID value is: ; Where: TF_ID represents the TF-IDF value of the word to be selected, Indicates the frequency of the candidate word appearing in the data document. Represents the inverse document frequency of the candidate word; where, ; Where: Represents the total number of data documents in the data source. represents the logarithmic function, DF is the document frequency, and its calculation formula is: DF=N(d); where: N(d) represents the number of documents containing the word to be selected;

[0068] Expand the knowledge data entity according to the TF_IDF value to obtain multiple second entities;

[0069] Specifically, the entity expansion of knowledge data according to the TF_IDF value includes:

[0070] Compare the TF_IDF value of each candidate word with the preset TF_IDF value threshold;

[0071] If the TF_IDF value is greater than or equal to the preset TF_IDF threshold, the corresponding candidate word is marked as the second entity;

[0072] If the TF_IDF value is less than the preset TF_IDF threshold, the corresponding candidate word is marked as non-second entity;

[0073] Count all candidate words whose TF_IDF values ​​are greater than or equal to a preset TF_IDF threshold to obtain multiple second entities;

[0074] Specifically, after obtaining the plurality of second entities, the method includes:

[0075] Extract all second entities and first entities;

[0076] Using the Word2Vec bag-of-words model, the first entity is converted into a first word vector, and the second entity is converted into a second word vector;

[0077] Calculate the semantic distance between all second word vectors and each first word vector, and compare the semantic distance with a preset semantic distance threshold; wherein the semantic distance is calculated by a cosine similarity algorithm;

[0078] If the semantic distance is greater than or equal to a preset semantic distance threshold, the second entity corresponding to the second word vector is removed from the multiple second entities;

[0079] If the semantic distance is less than a preset semantic distance threshold, retaining the second entity corresponding to the second word vector in the plurality of second entities;

[0080] It should be understood that: in order to avoid entity ambiguity between the automatically acquired second entity and the first entity, the entities are disambiguated by measuring the similarity between them;

[0081] In implementation, the construction of the original knowledge graph based on the knowledge data includes:

[0082] Construct entity-relationship-attribute triples based on knowledge data;

[0083] Taking the first and second entities in the triple as original nodes, and the first and second relations as original edges between the original nodes, to form an original knowledge graph;

[0084] By digitizing standard operating procedures in paper or electronic document form, it helps to improve the efficiency of production equipment use and operational compliance during the production process, thereby effectively improving the overall efficiency of the production line, reducing human errors, and ensuring stable product quality.

[0085] S102: Calculate the node degree of each original node in the original knowledge graph, and filter out all isolated nodes according to the node degree, and after obtaining multiple associated nodes related to the isolated nodes, perform relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships;

[0086] It should be understood that the relationship extraction based on predefined rules is subject to the richness of the rule base. However, due to the large volume and variety of data sources in the electronics industry, it is difficult to ensure the richness of the content of the knowledge graph model based on the artificially formulated rule base. This is why when using the knowledge graph model to guide operators in standard operating methods, it is impossible to cover all knowledge points as much as possible, so it is still impossible to avoid the confusion of novice operators about the standard operating methods, and thus it is impossible to improve the significant convenience brought by the digitization of standard operating procedures. Therefore, necessary relationship supplementation is needed;

[0087] In implementation, the node degree of each original node in the original knowledge graph is calculated, and isolated nodes are screened out according to the node degree, including:

[0088] Traverse all the original nodes in the original knowledge graph, and count the number of incoming edges for each original node and the number of outgoing edges ;

[0089] Number of incoming edges and the number of outgoing edges Perform sum calculation to obtain the node degree of each original node;

[0090] The calculation formula of the node degree is specifically as follows: , where: is the node degree of each original node;

[0091] The node degree is compared with the preset node degree threshold. If the node degree is less than or equal to the preset node degree threshold, the corresponding original node is marked as an isolated node. If the node degree is greater than the preset node degree threshold, the corresponding original node is not marked as an isolated node.

[0092] Count all original nodes whose node degrees are less than or equal to the node degree threshold, and get all isolated nodes;

[0093] In implementation, the step of obtaining a plurality of associated nodes related to the isolated node includes:

[0094] Extract the isolated node and all the original nodes, and obtain the text distance between the isolated node and each original node in the data source;

[0095] Compare the text spacing with a preset text spacing threshold;

[0096] If the text spacing is less than or equal to the text spacing threshold, it is determined that the isolated node and the corresponding original node co-occur in the data source, and the corresponding original node is marked as an associated node related to the isolated node;

[0097] If the text spacing is greater than the text spacing threshold, it is determined that the isolated node and the corresponding original node do not co-occur in the data source, and the corresponding original node is not marked as an associated node related to the isolated node;

[0098] For example, suppose there is a sentence in the data source that is "When the temperature signal of the temperature control module is 85°C, start the cooling system", and continue to assume that "temperature control module" is an isolated node and "cooling system" is an original node. After preprocessing the corpus, "temperature control module / temperature signal / is / 85°C / when / start / cooling system" is obtained, where the text distance between "temperature control module" and "cooling system" is 6. Assuming that the text spacing threshold is 10, "cooling system" is taken as the associated node of "temperature control module";

[0099] In implementation, the relationship mining between isolated nodes and associated nodes includes:

[0100] Locate the target sentence where isolated nodes and associated nodes co-occur in the data source;

[0101] Construct a dependency tree for the target sentence after word segmentation to obtain dependency trees about isolated nodes and associated nodes;

[0102] Obtain the shortest dependency path of the isolated node and the associated node in the dependency tree, obtain all predicate verbs in the dependency path, and respectively parse the dependency relationship of each predicate verb with respect to the isolated node and the associated node, and obtain the weight of each dependency relationship to obtain multiple weights;

[0103] It should be noted that the construction of the dependency tree, the generation of the shortest dependency path and the parsing of the dependency relationship are implemented based on the existing dependency syntax parsing tools, which include but are not limited to Stanford CoreNLP or spaCy; wherein the dependency relationship at least includes a subject relationship (nsubj), a conditional modification relationship (advmod) and a verb-object relationship (dobj), wherein the weight of the subject relationship is 0.5, the weight of the conditional modification relationship is 0.7, and the weight of the verb-object relationship is 1.0;

[0104] Sort multiple weights by value from large to small, and use the predicate verb corresponding to the first weight as the second relation;

[0105] For example, following the above example, assume that there is a sentence in the data source that is "When the temperature signal of the temperature control module is 85°C, start the cooling system", and continue to assume that "temperature control module" is an isolated node and "cooling system" is an associated node. After word segmentation of the corpus, we get "temperature control module / temperature signal / is / 85°C / when / start / cooling system". Then, we construct a dependency tree for the target sentence and get a dependency tree about isolated nodes and associated nodes (such as Figure 2 (As shown in the structural diagram of the dependency tree), according to the dependency tree, it is found that the shortest dependency path of isolated nodes and associated nodes in the dependency tree is: temperature control module → detected → when → start → cooling system, among which, the shortest dependency path includes two predicate verbs "detected" and "start", and among them, "detected" is a subject relationship for "temperature control module", so the weight is 0.5, and "detected" has no dependency relationship for "cooling system", when there is no dependency relationship, the weight is 0, and "start" is a subject-object relationship for "cooling system", the weight is 1, and "start" has no dependency relationship for "temperature control module", the weight is 0, therefore, after sorting according to the weight, "start" is taken as the second relationship.

[0106] S103: Optimize the structure of the original knowledge graph according to the second relationship, add new relationship edges to isolated nodes, and obtain a large knowledge graph model applied to the electronics industry;

[0107] It should be understood that: after completing the structural optimization of the original knowledge graph, the original knowledge graph after structural optimization needs to be imported into a preset graph database, which is achieved by using the graph database API or batch import tool; the graph database can be pre-selected by technical personnel based on data scale, complexity, and performance requirements. The graph database includes but is not limited to a specific one of Neo4j, Amazon Neptune or ArangoDB.

[0108] S104: Obtain a query request, input the query request into the knowledge graph model to identify the corresponding SOP data, and provide visual feedback; the query request at least includes keywords of entities or relationships;

[0109] It should be understood that the query request can be obtained by manual input or by identifying the state of the production equipment; for example, when the system identifies that the current interface of a certain production equipment is in the "welding process stage", the name of the production equipment and "welding process stage" are used as keywords and input into the knowledge graph model to obtain the corresponding SOP data; wherein the SOP data includes but is not limited to standard operating parameters or operating steps, etc.;

[0110] By constructing a large knowledge graph model, the present invention transforms SOP management from a traditional static, manually dependent mode to a dynamic, intelligent management system, significantly reducing operators' reliance on cumbersome static document consultation and reducing operational risks caused by missing information or misunderstandings; with the help of intelligent query and feedback mechanisms, the present invention comprehensively improves work efficiency and guidance accuracy, and further optimizes the standardization and consistency of the production process; in addition, the present invention overcomes the problems of incomplete knowledge coverage and insufficient query accuracy caused by the limitations of rule base design in existing knowledge graph construction methods by dynamically expanding entities and relationships and combining deep mining and optimization technology of semantic associations, and realizes comprehensive extraction and accurate expression of SOP data in the electronics industry; the present invention not only significantly improves the structural integrity and applicability of the knowledge graph, but also ensures its application effect in complex industrial scenarios, providing strong support for the intelligent upgrade of modern industrial production.

[0111] Example 2

[0112] See also Figure 3 As shown, based on the same inventive concept, this embodiment discloses an industrial SOP identification system based on a large model of the electronics industry. For details not provided in this embodiment, please refer to the description of the relevant parts in Example 1. The system includes:

[0113] The data processing module 210 is used to extract knowledge data from the electronic industry data source and construct an original knowledge graph based on the knowledge data; the knowledge data includes entities, relationships and attributes;

[0114] The relationship mining module 220 is used to calculate the node degree of each original node in the original knowledge graph, and filter out all isolated nodes according to the node degree, and after obtaining multiple associated nodes related to the isolated nodes, perform relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships;

[0115] The data optimization module 230 is used to optimize the structure of the original knowledge graph according to the second relationship, add new relationship edges to the isolated nodes, and obtain a large knowledge graph model applied to the electronics industry;

[0116] The SOP identification module 240 is used to obtain the query request, input the query request into the knowledge graph model to identify the corresponding SOP data, and provide visual feedback; the query request at least contains keywords of entities or relationships.

[0117] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0118] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0119] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0121] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0122] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0123] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0124] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0125] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An industrial SOP recognition method based on a large model of the electronics industry, characterized in that: The method comprises: Extracting knowledge data from electronic industry data sources and constructing an original knowledge graph based on the knowledge data; the knowledge data includes entities, relationships and attributes; Calculate the node degree of each original node in the original knowledge graph, and filter out all isolated nodes according to the node degree. After obtaining multiple associated nodes related to the isolated nodes, perform relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships. The original knowledge graph is structurally optimized according to the second relationship, and new relationship edges are added to isolated nodes to obtain a large knowledge graph model applied to the electronics industry. Obtain the query request, input the query request into the knowledge graph model to identify the corresponding SOP data, and provide visual feedback; the query request at least contains keywords of entities or relationships.

2. The industrial SOP identification method based on the large model of the electronics industry according to claim 1, characterized in that, The entities include a first entity and a second entity, and the relationships include a first relationship and a second relationship; The extraction is from knowledge data in the electronics industry data source, including: Mark the words in the data source that are not the first entity as words to be selected; Calculate the TF_IDF value of each candidate word in the data source; The knowledge data is entity expanded according to the TF_IDF value to obtain multiple second entities.

3. The industrial SOP identification method based on the large model of the electronics industry according to claim 2, characterized in that, The entity expansion of the knowledge data according to the TF_IDF value includes: Compare the TF_IDF value of each candidate word with the preset TF_IDF value threshold; If the TF_IDF value is greater than or equal to the preset TF_IDF threshold, the corresponding candidate word is marked as the second entity; If the TF_IDF value is less than the preset TF_IDF threshold, the corresponding candidate word is marked as non-second entity; All candidate words whose TF_IDF values ​​are greater than or equal to a preset TF_IDF threshold are counted to obtain a plurality of second entities.

4. The industrial SOP identification method based on the large model of the electronics industry according to claim 3 is characterized in that, After obtaining multiple second entities, including: Extract all second entities and first entities; Using the Word2Vec bag-of-words model, the first entity is converted into a first word vector, and the second entity is converted into a second word vector; Calculate the semantic distance between all second word vectors and each first word vector, and compare the semantic distance with a preset semantic distance threshold; wherein the semantic distance is calculated by a cosine similarity algorithm; If the semantic distance is greater than or equal to a preset semantic distance threshold, the second entity corresponding to the second word vector is removed from the multiple second entities; If the semantic distance is less than a preset semantic distance threshold, the second entity corresponding to the second word vector is retained in the plurality of second entities.

5. The industrial SOP identification method based on the large model of the electronics industry according to claim 4 is characterized in that, The constructing of the original knowledge graph according to the knowledge data includes: Construct entity-relationship-attribute triples based on knowledge data; The first and second entities in the triple are taken as original nodes, and the first and second relations are taken as original edges between the original nodes to form an original knowledge graph.

6. The industrial SOP identification method based on the large model of the electronics industry according to claim 5 is characterized in that, The step of calculating the node degree of each original node in the original knowledge graph and filtering out isolated nodes according to the node degree includes: Traverse all the original nodes in the original knowledge graph, and count the number of incoming edges for each original node and the number of outgoing edges ; Number of incoming edges and the number of outgoing edges Perform sum calculation to obtain the node degree of each original node; The node degree is compared with the preset node degree threshold. If the node degree is less than or equal to the preset node degree threshold, the corresponding original node is marked as an isolated node. If the node degree is greater than the preset node degree threshold, the corresponding original node is not marked as an isolated node. Count all original nodes whose node degrees are less than or equal to the node degree threshold and obtain all isolated nodes.

7. The industrial SOP identification method based on the large model of the electronics industry according to claim 6 is characterized in that, The step of acquiring a plurality of associated nodes related to the isolated node includes: Extract the isolated node and all the original nodes, and obtain the text distance between the isolated node and each original node in the data source; Compare the text spacing with a preset text spacing threshold; If the text spacing is less than or equal to the text spacing threshold, it is determined that the isolated node and the corresponding original node co-occur in the data source, and the corresponding original node is marked as an associated node related to the isolated node; If the text spacing is greater than the text spacing threshold, it is determined that the isolated node and the corresponding original node do not co-occur in the data source, and the corresponding original node is not marked as an associated node related to the isolated node.

8. The industrial SOP identification method based on the large model of the electronics industry according to claim 7 is characterized in that, The method of mining the relationship between isolated nodes and associated nodes includes: Locate the target sentence where isolated nodes and associated nodes co-occur in the data source; Construct a dependency tree for the target sentence after word segmentation to obtain dependency trees about isolated nodes and associated nodes; Obtain the shortest dependency path of the isolated node and the associated node in the dependency tree, obtain all predicate verbs in the dependency path, and respectively parse the dependency relationship of each predicate verb with respect to the isolated node and the associated node, and obtain the weight of each dependency relationship to obtain multiple weights; The multiple weights are sorted from large to small according to the numerical values, and the predicate verb corresponding to the first weight is used as the second relationship.

9. An industrial SOP recognition system based on a large model of the electronics industry, characterized in that: The method is implemented based on the industrial SOP identification method based on the large model of the electronics industry according to any one of claims 1 to 8, and the system comprises: A data processing module is used to extract knowledge data from electronic industry data sources and construct an original knowledge graph based on the knowledge data; the knowledge data includes entities, relationships and attributes; The relationship mining module is used to calculate the node degree of each original node in the original knowledge graph, and filter out all isolated nodes according to the node degree size, and after obtaining multiple associated nodes related to the isolated nodes, perform relationship mining on the isolated nodes and the associated nodes to obtain multiple second relationships; A data optimization module is used to optimize the structure of the original knowledge graph according to the second relationship, add new relationship edges to isolated nodes, and obtain a large knowledge graph model applied to the electronics industry; The SOP identification module is used to obtain the query request, input the query request into the knowledge graph model to identify the corresponding SOP data, and provide visual feedback; the query request contains at least keywords of entities or relationships.

Images