An automated construction method for a knowledge graph of illegal behaviors at live working sites centered on behavior nodes

Abstract

An automated construction method for a knowledge graph of illegal behaviors at live working sites centered on behavior nodes, which belongs to the technical field of monitoring illegal behaviors at live working sites. The present invention solves the problems that the knowledge graph constructed by the existing method has one-sided content, forms knowledge islands, and contains unknown inferences and cognitive holes. Without human intervention and additional input of association relationships, the present invention can establish a knowledge graph of illegal behaviors at live working sites only by using the input case data of illegal behaviors at live working sites; the behaviors, cases, risks, and their association relationships can be obtained from the knowledge graph, solving the problems of unknown inferences and cognitive holes in the knowledge graph constructed by the existing method, and providing effective support for the automated analysis of live working sites. The method of the present invention can be applied to the automated construction of a knowledge graph of illegal behaviors at live working sites.

Description

Technical Field

[0001] The present invention belongs to the technical field of monitoring illegal behaviors at live working sites, and particularly relates to an automated construction method of a knowledge graph of illegal behaviors at live working sites centered on behavior nodes. Background Art

[0002] Live working is an important means to ensure power supply reliability. However, the operation risk is high, and illegal behaviors endanger the safety of operators and the stability of the power grid. Constructing a knowledge graph of illegal behaviors at live working sites can assist artificial intelligence programs to discover on-site violations and infer possible risks, which is of great significance for safe construction, ensuring power supply safety and the personal safety of workers.

[0003] Currently, traditional methods for constructing a knowledge graph of illegal behaviors at live working sites include: 1) Manual construction, where domain experts analyze illegal behaviors at live working sites and input them node by node manually. This method consumes a large amount of manpower and time, is difficult to scale up, and is difficult to update. At the same time, due to the variety of illegal behaviors, some illegal behaviors are not thought of by experts, so the knowledge graph constructed by this method is often one-sided and cannot effectively provide support for intelligent recognition. 2) Providing clear concept definitions and relationship structures and using ontology for automated construction. Although the efficiency is improved, a key problem with this type of method is that the illegal behaviors at live working sites are independent of each other, and the correlation in the original dataset is low (unless the same person violates the regulations multiple times). Therefore, the knowledge graph established by this method will form relatively isolated knowledge islands and is difficult to form effective reasoning support for artificial intelligence programs. 3) Natural language processing knowledge matching. To solve the isolation problem brought by ontology, a vector database is used to perform knowledge matching and association on the case description content. This method can find the correlation between different illegal behaviors semantically. However, the natural language processing vectorization process (or model) is usually established for general application scenarios, and this processing program often ignores key content and words, resulting in many unknown reasoning and cognitive holes in the knowledge graph, leading to missed and misjudged illegal behaviors.

[0004] It can be seen that it is very important to quickly and effectively establish a knowledge graph of illegal behaviors at live working sites and maintain effective graph nodes, attributes, and relationships. Summary of the Invention

[0005] The object of the present invention is to solve the problems that the knowledge graph constructed by the existing method has one-sided content, forms knowledge islands, and contains unknown reasoning and cognitive holes, and proposes an automated construction method of a knowledge graph of illegal behaviors at live working sites centered on behavior nodes. Based on the case data of illegal behaviors at live working sites and the behaviors and resulting hazards of the case data, a corresponding knowledge graph is automatically established with the behavior node as the center of association.

[0006] The technical solution adopted by the present invention to solve the above technical problems is: an automated construction method of a knowledge graph of illegal behaviors at live working sites centered on behavior nodes, and the method specifically includes the following steps:

[0007] Step S1, input the list of illegal behavior case data DDDataSet at the live working site, input the list of dangerous behavior words WXDict, and input the list of risk point words FXDict;

[0008] Establish a behavior alternative list DDKeyDict for the list of illegal behavior data at the live working site; establish a risk point alternative list DDFXDict for the illegal behavior at the live working site;

[0009] Step S2, establish a knowledge graph KG of illegal behaviors at the live working site, add behavior nodes to KG based on DDKeyDict, and add risk point nodes to KG based on DDFXDict;

[0010] Step S3, based on DDDataSet and the existing behavior nodes, establish nodes and links of knowledge in the knowledge graph KG.

[0011] Furthermore, the specific process of step S1 is as follows:

[0012] S101, input the list of illegal behavior case data DDDataSet at the live working site, and each element of this list corresponds to a case data of an illegal behavior at the live working site, including 4 fields;

[0013] DDID: The number of a case of an illegal behavior at the live working site;

[0014] DDImage: Picture data of the illegal behavior at the live working site;

[0015] DDXWDesc: Description of the illegal behavior at the live working site, using natural language to describe the illegal behavior that appears at the live working site;

[0016] DDFXDesc: Description of the risk point behavior, using natural language to describe the risk points existing at the live working site and the harms and damages caused;

[0017] S102, input the list of dangerous behavior words WXDict, and each element of this list corresponds to a dangerous behavior entry during live working;

[0018] S103, input the list of risk point words FXDict, and each element of this list corresponds to a description entry of the result caused by a dangerous behavior during live working;

[0019] S104, The behavior counter DDCounter = 1; The behavior alternative list DDKeyDict of the illegal behavior data list at the live working site = an empty list; Establish the risk point alternative list DDFXDict for illegal behaviors at the live working site = an empty list;

[0020] S105, The behavior temporary data DDTemp = retrieve the DDCounter-th data from DDDataSet;

[0021] S106, Retrieve the content of the DDXWDesc field of DDTemp, search for the words that appear in WXDict in the retrieved content, and put all the found words into the single-case behavior list DDXWCT; Retrieve the content of the DDFXDesc field of DDTemp, search for the words that appear in FXDict in the retrieved content, and put all the found words into the single-case behavior risk list DDFXCT;

[0022] S107, DDKeyDict = obtain the union of DDKeyDict and DDXWCT; DDFXDict = obtain the union of DDFXDict and DDFXCT;

[0023] S108, Let DDCounter = DDCounter + 1;

[0024] S109, If DDCounter is less than or equal to the number of elements in DDDataSet, go to S105, otherwise go to S110;

[0025] S110, Step S1 ends, obtaining the final behavior alternative list DDKeyDict and the risk point alternative list DDFXDict.

[0026] Furthermore, the specific process of the said step S2 is as follows:

[0027] S201, Establish the knowledge graph KG for illegal behaviors at the live working site = an empty knowledge graph;

[0028] S202, The behavior node counter KGCCounter for knowledge graph establishment = 1;

[0029] S203, Establish a new knowledge graph behavior node NXWNode;

[0030] S204, Set the name NNodeName attribute of NXWNode to the KGCCounter-th element of DDKeyDict;

[0031] S205, Create a new behavior description vector KGDescVector = a vector with the same number of dimensions as the number of elements in DDKeyDict and all elements being 0;

[0032] S206, Set the KGCCounter-th element of KGDescVector to 1;

[0033] S207, Set the behavior attribute NnodeAction attribute of NXWNode to KGDescVector;

[0034] S208, Add NXWNode to KG;

[0035] S209, KGCCounter = KGCCounter + 1;

[0036] S210, If KGCCounter is less than or equal to the number of elements in DDKeyDict, go to S203; otherwise go to S211;

[0037] S211, The risk point counter KGFXCounter for knowledge graph establishment = 1;

[0038] S212, Create a new knowledge graph risk node NFXNode;

[0039] S213, The risk name NFXNodeName attribute of NFXNode is the KGFXCounter-th element of DDFXDict;

[0040] S214, Add NFXNode to KG;

[0041] S215, Let KGFXCounter = KGFXCounter + 1;

[0042] S216, If KGFXCounter is less than or equal to the number of elements in DDFXDict, go to S212, otherwise go to S217;

[0043] S217, End.

[0044] Furthermore, the specific process of step S3 is as follows:

[0045] S301, Create a process counter DDCCCounter = 1;

[0046] S302, Create a new knowledge graph case node NANNode;

[0047] S303. Establish the process temporary case DDCCTemp = retrieve the DDCCCounter-th element of DDDataSet;

[0048] S304. Retrieve the content of the DDXWDesc field of DDCCTemp, search for the words that appear in WXDict in the retrieved content, and put all the found words into the process establishment single case behavior list DDCCXW; retrieve the content of the DDFXDesc field of DDCCTemp, search for the words that appear in FXDict in the retrieved content, and put all the found words into the process establishment single behavior case risk list DDCCFX;

[0049] S305. Establish the process to find the behavior node list DDCCFindXWList = search in the behavior nodes of KG and find all the nodes whose name NNodeName attribute appears in DDCCXW;

[0050] S306. Establish links from all nodes in DDCCFindXWList to NANNode in KG;

[0051] S307. Establish the process to find the risk node list DDCCFindFXList = search in the behavior nodes of KG and find all the nodes whose name NFXNodeName attribute appears in DDFXDict;

[0052] S308. Establish links from NANNode to all nodes in DDCCFindFXList in KG;

[0053] S309. Establish the process behavior description vector DDCCXWVector = a vector with the same dimension as the number of elements in DDKeyDict, and all elements are 0;

[0054] S310. Establish the process behavior entry position DDCCXWPos = find the entries in DDKeyDict that appear in DDCCXW and obtain the position list of the entries;

[0055] S311. Set the corresponding elements at the positions recorded by DDCCXWPos in DDCCXWVector to 1;

[0056] S312. Establish the process risk description vector DDCCFXVector = a vector with the same dimension as the number of elements in DDFXDict, and all elements are 0;

[0057] S313. Establish the process risk entry position DDCCFXPos = find the entries in DDFXDict that appear in DDCCFX and obtain the position list of the entries;

[0058] S314, set the element corresponding to the position recorded by DDCCFXPos in DDCCFXVector to 1;

[0059] S315, the case behavior description attribute ANXW of NANNode = DDCCXWVector;

[0060] S316, the case risk description attribute ANFX of NANNode = DDCCFXVector;

[0061] S317, the source case attribute LYAN of NANNode = the DDID attribute of DDCCTemp;

[0062] S318, the source picture attribute LYTP of NANNode = the DDImage attribute of DDCCTemp;

[0063] S319, the source specific description LYContet of NANNode = the DDXWDesc attribute of DDCCTemp + the DDFXDesc attribute of DDCCTemp, where the + sign indicates string combination;

[0064] S320, add NANNode to KG;

[0065] S321, store KG to complete the construction process of the knowledge graph.

[0066] The beneficial effects of the present invention are:

[0067] The present invention can establish a knowledge graph of illegal operation behaviors at the live working site only by using the input case data of illegal operation behaviors at the live working site without manual intervention and additional input of association relationships; from this knowledge graph, behaviors, cases, risks and their association relationships can be obtained, solving the problems of unknown reasoning and cognitive holes in the knowledge graph constructed by the existing methods, providing effective support for the automated analysis of the live working site, and at the same time solving the problems of one-sided content and knowledge islands in the knowledge graph constructed by the existing methods. Brief Description of the Drawings

[0068] Figure 1 is a flowchart of an automated construction method of a knowledge graph of illegal operation behaviors at the live working site centered on behavior nodes of the present invention. Detailed Embodiments

[0069] Detailed Embodiment 1: Combine Figure 1Describe this embodiment. A method for automatically constructing a knowledge graph of illegal behaviors at live working sites centered on behavior nodes according to this embodiment specifically includes the following steps:

[0070] Step S1, input the list of illegal behavior case data DDDataSet at live working sites, input the list of dangerous behavior words WXDict, and input the list of risk point words FXDict;

[0071] Establish a behavior alternative list DDKeyDict for the list of illegal behavior data at live working sites; establish a risk point alternative list DDFXDict for illegal behaviors at live working sites;

[0072] Step S2, establish a knowledge graph KG of illegal behaviors at live working sites, add behavior nodes to KG based on DDKeyDict, and add risk point nodes to KG based on DDFXDict;

[0073] Step S3, based on DDDataSet and the existing behavior nodes, establish nodes and links of knowledge in the knowledge graph KG.

[0074] Specific Embodiment 2: The difference between this embodiment and Specific Embodiment 1 is that the specific process of step S1 is as follows:

[0075] S101, input the list of illegal behavior case data DDDataSet at live working sites. Each element of this list corresponds to a case data of an illegal behavior at a live working site and contains 4 fields;

[0076] DDID: The number of a case of an illegal behavior at a live working site;

[0077] DDImage: Picture data of an illegal behavior at a live working site, which is the entire illegal behavior at the live working site;

[0078] DDXWDesc: Description of an illegal behavior at a live working site, using natural language to describe the illegal behavior that occurs at the live working site;

[0079] DDFXDesc: Description of risk point behaviors, using natural language to describe the main risk points existing at the live working site and the possible hazards and damages that may be caused;

[0080] S102, input the list of dangerous behavior words WXDict. Each element of this list corresponds to a dangerous behavior entry during live working, such as: not wearing a safety helmet, not grounding;

[0081] S103, input the list of risk point words FXDict. Each element of this list corresponds to a description entry of the result caused by a dangerous behavior during live working, such as: electric shock, breakdown, explosion;

[0082] S104, the behavior counter DDCounter = 1; the behavior alternative list DDKKeyDict of the live working site violation behavior data list = an empty list; establish the risk point alternative list DDFXDict of the live working site violation behavior = an empty list;

[0083] S105, the behavior temporary data DDTemp = retrieve the DDCounter-th data from DDDataSet;

[0084] S106, retrieve the content of the DDXWDesc field of DDTemp, search for the words that appear in WXDict in the retrieved content, and put all the found words into the single case behavior list DDXWCT; retrieve the content of the DDFXDesc field of DDTemp, search for the words that appear in FXDict in the retrieved content, and put all the found words into the single case risk list DDFXCT;

[0085] S107, DDKKeyDict = obtain the union of DDKKeyDict and DDXWCT; DDFXDict = obtain the union of DDFXDict and DDFXCT;

[0086] S108, let DDCounter = DDCounter + 1;

[0087] S109, if DDCounter is less than or equal to the number of elements in DDDataSet, go to S105, otherwise go to S110;

[0088] S110, step S1 ends, and obtain the final behavior alternative list DDKKeyDict and risk point alternative list DDFXDict.

[0089] Other steps and parameters are the same as those in the first specific implementation manner.

[0090] The third specific implementation manner: The difference between this implementation manner and the first or second specific implementation manner is that the specific process of step S2 is as follows:

[0091] S201, establish the knowledge graph KG of the live working site violation behavior = an empty knowledge graph;

[0092] S202, the behavior node counter KGCCounter for knowledge graph establishment = 1;

[0093] S203, establish a new knowledge graph behavior node NXWNode;

[0094] S204, set the name NNodeName attribute of NXWNode to the KGCCounter-th element of DDKeyDict;

[0095] S205, create a new behavior description vector KGDescVector = a vector with the same number of dimensions as the number of elements in DDKeyDict and all elements being 0;

[0096] S206, set the KGCCounter-th element of KGDescVector to 1;

[0097] S207, set the behavior attribute NnodeAction attribute of NXWNode to KGDescVector;

[0098] S208, add NXWNode to KG;

[0099] S209, KGCCounter = KGCCounter + 1;

[0100] S210, if KGCCounter is less than or equal to the number of elements in DDKeyDict, go to S203; otherwise go to S211;

[0101] S211, the risk point counter KGFXCounter for knowledge graph establishment = 1;

[0102] S212, create a new knowledge graph risk node NFXNode;

[0103] S213, the risk name NFXNodeName attribute of NFXNode is the KGFXCounter-th element of DDFXDict;

[0104] S214, add NFXNode to KG;

[0105] S215, let KGFXCounter = KGFXCounter + 1;

[0106] S216, if KGFXCounter is less than or equal to the number of elements in DDFXDict, go to S212, otherwise go to S217;

[0107] S217, end.

[0108] Other steps and parameters are the same as those in the first or second specific implementation manner.

[0109] Specific implementation manner four: The difference between this implementation manner and one of the first to third specific implementation manners is that the specific process of step S3 is as follows:

[0110] S301, Establish a process counter DDCCCounter = 1;

[0111] S302, Establish a new knowledge graph case node NANNode;

[0112] S303, Establish a process temporary case DDCCTemp = Retrieve the DDCCCounter-th element of DDDataSet;

[0113] S304, Retrieve the content of the DDXWDesc field of DDCCTemp, search for words that appear in WXDict in the retrieved content, and put all the found words into the process single case behavior list DDCCXW; Retrieve the content of the DDFXDesc field of DDCCTemp, search for words that appear in FXDict in the retrieved content, and put all the found words into the process single behavior case risk list DDCCFX;

[0114] S305, Establish a process found behavior node list DDCCFindXWList = Search in the behavior nodes of KG and find all nodes whose NNodeName attribute of the name appears in DDCCXW;

[0115] S306, Establish links from all nodes in DDCCFindXWList to NANNode in KG;

[0116] S307, Establish a process found risk node list DDCCFindFXList = Search in the behavior nodes of KG and find all nodes whose NFXNodeName attribute of the name appears in DDFXDict;

[0117] S308, Establish links from NANNode to all nodes in DDCCFindFXList in KG;

[0118] S309, Establish a process behavior description vector DDCCXWVector = A vector with the same number of dimensions as the number of elements in DDKeyDict, and all elements are 0;

[0119] S310, Establish a process behavior entry position DDCCXWPos = Find the entries in DDKeyDict that appear in DDCCXW and obtain the position list of the entries;

[0120] S311, Set the corresponding elements at the positions recorded by DDCCXWPos in DDCCXWVector to 1;

[0121] S312, establish a process risk description vector DDCCFXVector = a vector with the same number of elements as the dimensions of DDFXDict, and all elements are 0;

[0122] S313, establish the process risk entry positions DDCCFXPos = find the entries in DDFXDict that appear in DDCCFX and obtain a list of the positions of the entries;

[0123] S314, set the elements corresponding to the positions recorded by DDCCFXPos in DDCCFXVector to 1;

[0124] S315, the case behavior description attribute ANXW of NANNode = DDCCXWVector;

[0125] S316, the case risk description attribute ANFX of NANNode = DDCCFXVector;

[0126] S317, the source case attribute LYAN of NANNode = the DDID attribute of DDCCTemp;

[0127] S318, the source picture attribute LYTP of NANNode = the DDImage attribute of DDCCTemp;

[0128] S319, the source specific description LYContet of NANNode = the DDXWDesc attribute of DDCCTemp + the DDFXDesc attribute of DDCCTemp, where the + sign indicates string combination;

[0129] S320, add NANNode to KG;

[0130] S321, store KG to complete the construction process of the knowledge graph.

[0131] Other steps and parameters are the same as one of the specific embodiments one to three.

[0132] The above examples of the present invention are only for explaining in detail the calculation model and calculation process of the present invention, rather than limiting the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is impossible to list all the implementation manners here. Any obvious changes or modifications derived from the technical solutions of the present invention still fall within the protection scope of the present invention.

Claims

1. An automated construction method for an illegal behavior knowledge graph of live working sites centered on behavior nodes, characterized in that, The method specifically includes the following steps: Step S1, input the list of violation behavior case data DDDataSet at the live working site, input the list of dangerous behavior words WXDict, and input the list of risk point words FXDict; Establish the behavior alternative list DDKeyDict of the violation behavior data list at the live working site; establish the risk point alternative list DDFXDict of the violation behavior at the live working site; The specific process of the said Step S1 is as follows: S101, input the list of violation behavior case data DDDataSet at the live working site. Each element of this list corresponds to the case data of a violation behavior at the live working site and contains 4 fields; DDID: The number of a violation behavior case at the live working site; DDImage: The picture data of the violation behavior at the live working site; DDXWDesc: The description of the violation behavior at the live working site, using natural language to describe the violation behavior that appears at the live working site; DDFXDesc: The description of the risk point behavior, using natural language to describe the risk points existing at the live working site and the harms and damages caused; S102, input the list of dangerous behavior words WXDict. Each element of this list corresponds to a dangerous behavior entry during live working; S103, input the list of risk point words FXDict. Each element of this list corresponds to a description entry of the result caused by a dangerous behavior during live working; S104, behavior counter DDCounter = 1; the behavior alternative list DDKeyDict of the violation behavior data list at the live working site = an empty list; establish the risk point alternative list DDFXDict of the violation behavior at the live working site = an empty list; S105, behavior temporary data DDTemp = retrieve the DDCounter-th data of DDDataSet; S106, retrieve the content of the DDXWDesc field of DDTemp, find the words that appear in WXDict in the retrieved content, and put all the found words into the single case behavior list DDXWCT; retrieve the content of the DDFXDesc field of DDTemp, find the words that appear in FXDict in the retrieved content, and put all the found words into the single case risk list DDFXCT; S107, DDKeyDict = obtain the union of DDKeyDict and DDXWCT; DDFXDict = obtain the union of DDFXDict and DDFXCT; S108, let DDCounter = DDCounter + 1; S109, if DDCounter is less than or equal to the number of elements of DDDataSet, then go to S105, otherwise go to S110; S110, Step S1 ends, and the final behavior alternative list DDKeyDict and risk point alternative list DDFXDict are obtained; Step S2, establish the knowledge graph KG of the violation behavior at the live working site, add behavior nodes to KG based on DDKeyDict, and add risk point nodes to KG based on DDFXDict; The specific process of step S2 is as follows: S201, establish a knowledge graph KG of illegal behaviors at the live working site = an empty knowledge graph; S202, the behavior node counter KGCCounter for knowledge graph establishment = 1; S203, establish a new knowledge graph behavior node NXWNode; S204, set the name NNodeName attribute of NXWNode to the KGCCounter-th element of DDKeyDict; S205, establish a new behavior description vector KGDescVector = a vector with the same dimension as the number of elements in DDKeyDict and all elements being 0; S206, set the KGCCounter-th element of KGDescVector to 1; S207, set the behavior attribute NnodeAction attribute of NXWNode to KGDescVector; S208, add NXWNode to KG; S209, KGCCounter = KGCCounter + 1; S210, if KGCCounter is less than or equal to the number of elements in DDKeyDict, go to S203; otherwise go to S211; S211, the risk point counter KGFXCounter for knowledge graph establishment = 1; S212, establish a new knowledge graph risk node NFXNode; S213, the risk name NFXNodeName attribute of NFXNode is the KGFXCounter-th element of DDFXDict; S214, add NFXNode to KG; S215, let KGFXCounter = KGFXCounter + 1; S216, if KGFXCounter is less than or equal to the number of elements in DDFXDict, go to S212, otherwise go to S217; S217, end; Step S3, based on DDDataSet and the existing behavior nodes, establish the nodes and links of knowledge in the knowledge graph KG; The specific process of step S3 is as follows: S301, establish a process counter DDCCCounter = 1; S302, establish a new knowledge graph case node NANNode; S303, establish a process temporary case DDCCTemp = retrieve the DDCCCounter-th element of DDDataSet; S304, retrieve the content of the DDXWDesc field of DDCCTemp, find the words that appear in WXDict in the retrieved content, and put all the found words into the single-case behavior list DDCCXW for the establishment process; retrieve the content of the DDFXDesc field of DDCCTemp, find the words that appear in FXDict in the retrieved content, and put all the found words into the single-behavior case risk list DDCCFX for the establishment process; S305. The establishment process finds the list of behavior nodes DDCCFindXWList = searches among the behavior nodes of KG and finds all nodes whose name NNodeName attribute appears in DDCCXW; S306. Establish links from all nodes in DDCCFindXWList to NANNode in KG; S307. The establishment process finds the list of risk nodes DDCCFindFXList = searches among the behavior nodes of KG and finds all nodes whose name NFXNodeName attribute appears in DDFXDict; S308. Establish links from NANNode to all nodes in DDCCFindFXList in KG; S309. The establishment process creates the behavior description vector DDCCXWVector = a vector with the same number of dimensions as the number of elements in DDKeyDict, and all elements are 0; S310. The establishment process determines the behavior entry positions DDCCXWPos = finds the entries in DDKeyDict that appear in DDCCXW and obtains the list of entry positions; S311. Set the corresponding elements at the positions recorded by DDCCXWPos in DDCCXWVector to 1; S312. The establishment process creates the risk description vector DDCCFXVector = a vector with the same number of dimensions as the number of elements in DDFXDict, and all elements are 0; S313. The establishment process determines the risk entry positions DDCCFXPos = finds the entries in DDFXDict that appear in DDCCFX and obtains the list of entry positions; S314. Set the corresponding elements at the positions recorded by DDCCFXPos in DDCCFXVector to 1; S315. The case behavior description attribute ANXW of NANNode = DDCCXWVector; S316. The case risk description attribute ANFX of NANNode = DDCCFXVector; S317. The source case attribute LYAN of NANNode = the DDID attribute of DDCCTemp; S318. The source picture attribute LYTP of NANNode = the DDImage attribute of DDCCTemp; S319. The source specific description LYContet of NANNode = the DDXWDesc attribute of DDCCTemp + the DDFXDesc attribute of DDCCTemp, where the + sign indicates string combination; S320. Add NANNode to KG; S321. Store KG to complete the knowledge graph construction process.

Images