Knowledge graph updating method and device, equipment and storage medium

Abstract

The application discloses a knowledge graph updating method and device, equipment and a storage medium. The knowledge graph updating method comprises the following steps: acquiring an evidence storage engine and to-be-updated knowledge, wherein the evidence storage engine comprises initial knowledge, evidence of the initial knowledge and an information source of the evidence of the initial knowledge; judging whether the to-be-updated knowledge satisfies a quality standard indicated by a preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge; and if the to-be-updated knowledge satisfies the quality standard, updating the to-be-updated knowledge to the evidence storage engine and a target knowledge graph. According to the initial knowledge, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge in the evidence storage engine, the quality of to-be-updated knowledge is evaluated, high-quality knowledge is selected to update the knowledge graph, and thus the overall quality of the constructed knowledge graph is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer network security, and particularly relates to a knowledge graph updating method and device, equipment and a storage medium. BACKGROUND

[0002] Network threat intelligence is evidence-based knowledge about existing or potential threats to IT or information assets, including context, mechanism, indicators, inferences and actionable recommendations, which can provide decision-making basis for threat response. In practical application, through information extraction technology, a large amount of unstructured threat intelligence information is converted into structured information, and a network security knowledge graph is constructed, which serves as the basis for attack detection, attack profiling and other tasks.

[0003] However, the current network threat intelligence data has the characteristics of large volume, wide sources and rapid updates, resulting in a large amount of redundancy and inconsistency in the new knowledge extracted by the information extraction system, such as repeated knowledge, non-unique entity labels and uncertain entity relationships, which leads to low overall quality of the constructed knowledge graph. SUMMARY

[0004] The main purpose of the present application is to provide a knowledge graph updating method, device, equipment and storage medium, which aims to solve the technical problem of low overall quality of the knowledge graph in the prior art.

[0005] To achieve the above purpose, the present application provides a knowledge graph updating method, which comprises:

[0006] acquiring an evidence storage engine and to-be-updated knowledge, wherein the evidence storage engine comprises initial knowledge, evidence of the initial knowledge and information source of the evidence of the initial knowledge;

[0007] judging whether the to-be-updated knowledge meets a quality standard indicated by a preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge;

[0008] if the to-be-updated knowledge meets the quality standard, updating the to-be-updated knowledge to the evidence storage engine and a target knowledge graph.

[0009] Optionally, the step of judging whether the to-be-updated knowledge meets the quality standard indicated by the preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge comprises:

[0010] selecting a preset number of reference knowledge of the to-be-updated knowledge from the evidence storage engine based on the initial knowledge;

[0011] determine the evidence of the reference knowledge and the information source of the evidence of the reference knowledge based on the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, and calculate the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge and the information source of the evidence of the reference knowledge;

[0012] determine whether the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range based on the first verification probability.

[0013] Optionally, the step of calculating the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge and the information source of the evidence of the reference knowledge comprises:

[0014] determine the credibility of the information source of the evidence of the reference knowledge;

[0015] perform verification probability calculation on each of the reference knowledge based on the evidence of the reference knowledge and the credibility of the information source of the evidence of the reference knowledge, to obtain a second verification probability of each of the reference knowledge, wherein the verification probability calculation formula is as follows:

[0016]

[0017] wherein t represents the reference knowledge, E t is the evidence set of the reference knowledge, E is the evidence set of the initial knowledge, is the credibility of the information source of the evidence, |·| represents the size of the set, and P(t) represents the second verification probability;

[0018] perform weighted sum calculation on the second verification probability of each of the reference knowledge, to obtain the first verification probability of the knowledge to be updated, wherein the weighted sum calculation formula is as follows:

[0019]

[0020] wherein t c represents the evidence of the knowledge to be updated, and P(t c ) represents the first verification probability of the knowledge to be updated.

[0021] Optionally, the step of selecting the reference knowledge of the preset number of the knowledge to be updated from the evidence storage engine comprises:

[0022] determine the first attribute information of the knowledge to be updated and the second attribute information of the initial knowledge, wherein the attribute information comprises entities, relationships and labels;

[0023] Based on the first attribute information and the second attribute information, similarity calculation is performed on each of the initial knowledge and the knowledge to be updated to obtain a similarity result, wherein a similarity calculation formula is as follows:

[0024]

[0025] wherein t represents knowledge, sim(t i ,t j ) represents similarity of any two knowledge, sub, pre and obj represent entity, relationship and label of knowledge respectively, represents cosine similarity of any two entity vectors, γ1, γ2 and γ3 are coefficients, representing weight of entity, relationship and label respectively, and γ1+γ2+γ3=1;

[0026] The maximum preset number of initial knowledge in the similarity result is determined as reference knowledge of the knowledge to be updated.

[0027] Optionally, the step of determining whether the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range based on the first verification probability comprises:

[0028] Based on the initial knowledge, it is determined whether there is a conflict knowledge of the knowledge to be updated in the evidence storage engine;

[0029] If there is a conflict knowledge of the knowledge to be updated in the evidence storage engine, a third verification probability of the conflict knowledge is calculated;

[0030] The first verification probability and the third verification probability are compared in size, and if the first verification probability is greater than the third verification probability, it is determined that the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range.

[0031] Optionally, after the step of determining whether there is a conflict knowledge of the knowledge to be updated in the evidence storage engine based on the initial knowledge, the method comprises:

[0032] If there is no conflict knowledge of the knowledge to be updated in the evidence storage engine, the first verification probability and a preset probability threshold are compared in size, and if the first verification probability is greater than the probability threshold, it is determined that the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range.

[0033] Optionally, before the step of selecting a preset number of reference knowledge of the knowledge to be updated from the evidence storage engine based on the initial knowledge, the method comprises:

[0034] determine whether duplicate knowledge of the knowledge to be updated exists in the evidence storage engine based on the initial knowledge;

[0035] if the duplicate knowledge of the knowledge to be updated does not exist in the evidence storage engine, select a preset number of reference knowledge of the knowledge to be updated from the evidence storage engine;

[0036] if the duplicate knowledge of the knowledge to be updated exists in the evidence storage engine, determine that the knowledge to be updated meets the quality standard.

[0037] The application further provides a knowledge graph updating device, which comprises:

[0038] an acquisition module configured to acquire an evidence storage engine and knowledge to be updated, wherein the evidence storage engine comprises initial knowledge, evidence of the initial knowledge, and information sources of the evidence of the initial knowledge;

[0039] a judgment module configured to determine whether the knowledge to be updated meets a quality standard indicated by a preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge, and the information sources of the evidence of the initial knowledge;

[0040] an updating module configured to update the knowledge to be updated to the evidence storage engine and a target knowledge graph if the knowledge to be updated meets the quality standard.

[0041] The application further provides a knowledge graph updating device, which comprises a memory, a processor, and a program stored in the memory and used to implement a knowledge graph updating method,

[0042] the memory is configured to store the program used to implement the knowledge graph updating method;

[0043] the processor is configured to execute the program used to implement the knowledge graph updating method, so as to implement the steps of the knowledge graph updating method.

[0044] The application further provides a storage medium, which stores a program used to implement a knowledge graph updating method, and the program is executed by a processor to implement the steps of the knowledge graph updating method.

[0045] The knowledge graph updating method, device, equipment and storage medium provided in the application are different from the related art. In the related art, network threat intelligence data has the characteristics of large volume, wide sources and fast updating, which leads to the problems of a large amount of redundancy and inconsistency of new knowledge extracted by an information extraction system, and the quality of the constructed knowledge graph is low. In the application, an evidence storage engine and to-be-updated knowledge are obtained, wherein the evidence storage engine includes initial knowledge, evidence of the initial knowledge and an information source of the evidence of the initial knowledge. Whether the to-be-updated knowledge meets a quality standard indicated by a preset credibility threshold range is determined based on the initial knowledge, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge. If the to-be-updated knowledge meets the quality standard, the to-be-updated knowledge is updated to the evidence storage engine and a target knowledge graph. That is, in the application, the to-be-updated knowledge is quality evaluated according to the initial knowledge in the evidence storage engine, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, and high-quality knowledge is selected to update the knowledge graph, so as to improve the overall quality of the constructed knowledge graph. BRIEF DESCRIPTION OF DRAWINGS

[0046] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate an embodiment consistent with the present application and, together with the description, serve to explain the principles of the application. In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced as follows. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative labor.

[0047] Figure 1 is a device structure schematic diagram of a hardware running environment related to the embodiment scheme of the application;

[0048] Figure 2 is a flowchart of the first embodiment of the knowledge graph updating method of the application;

[0049] Figure 3 is a module schematic diagram of the knowledge graph updating device of the application;

[0050] Figure 4 is a data association structure schematic diagram of the evidence storage engine in the knowledge graph updating method of the application;

[0051] Figure 5 is a flowchart of the second embodiment of the knowledge graph updating method of the application.

[0052] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0054] As shown in Figure 1 , the terminal structure diagram of the hardware running environment involved in the embodiment scheme of the present application is shown. Figure 1

[0055] The terminal of the embodiment of the present application can be a PC, or a smart phone, a tablet computer, an e-book reader, an MP3 (Moving Picture Experts Group Audio Layer III) player, an MP4 (Moving Picture Experts Group Audio Layer IV) player, a portable computer, or other movable terminal devices with display functions.

[0056] As shown in Figure 1 , the terminal can include a processor 1001, for example, a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the components. The user interface 1003 can include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 can also include a standard wired interface and a wireless interface. The network interface 1004 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 can be a high-speed RAM memory or a stable memory (non-volatile memory) such as a disk memory. Optionally, the memory 1005 can also be a storage device independent of the aforementioned processor 1001.

[0057] ​Optionally, the terminal can further include a camera, RF (Radio Frequency) circuit, sensors, audio circuit, WiFi module, and the like. The sensors can include, for example, a light sensor, a motion sensor, and other sensors. Specifically, the light sensor can include an ambient light sensor and a proximity sensor, wherein the ambient light sensor can adjust the brightness of the display screen according to the brightness of ambient light, and the proximity sensor can turn off the display screen and / or backlight when the mobile terminal is moved to the ear. As one of the motion sensors, the gravity acceleration sensor can detect the size of acceleration in each direction (generally three axes), and when at rest, it can detect the size and direction of gravity, which can be used for identifying the posture of the mobile terminal (such as switching between horizontal and vertical screens, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometers, tapping), and the like. Of course, the mobile terminal can also be configured with a gyroscope, barometer, hygrometer, thermometer, infrared sensor, and other sensors, which will not be described here.

[0058] Those skilled in the art can understand that Figure 1 The terminal structure shown in the above embodiments does not constitute a limitation on the terminal, and can include more or fewer components than shown, or combine certain components, or different component arrangements.

[0059] As shown in Figure 1 The memory 1005 as a computer storage medium can include an operation device, a network communication module, a user interface module, and an update program of the knowledge graph.

[0060] In the terminal shown in Figure 1 The network interface 1004 is mainly used for connecting to a background server and communicating data with the background server; the user interface 1003 is mainly used for connecting to a client (user end) and communicating data with the client; and the processor 1001 can be used to call the update program of the knowledge graph stored in the memory 1005.

[0061] Referring to Figure 2 The embodiments of the present application provide an update method of a knowledge graph, which comprises the following steps:

[0062] In step S100, an evidence storage engine and to-be-updated knowledge are acquired, wherein the evidence storage engine includes initial knowledge, evidence of the initial knowledge, and information source of the evidence of the initial knowledge;

[0063] In step S200, whether the to-be-updated knowledge meets a quality standard indicated by a preset credibility threshold range is judged based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge.

[0064] Step S300, if the knowledge to be updated meets the quality standard, updating the knowledge to be updated to the evidence storage engine and the target knowledge graph.

[0065] In the embodiment, the application scenario is:

[0066] As an example, the scenario of updating the knowledge graph can be updating the knowledge graph of threat intelligence. The current network threat intelligence data has the characteristics of large volume, wide sources and fast updating, which leads to a large amount of redundancy and inconsistency in the new knowledge extracted by the information extraction system, such as repeated knowledge, non-unique entity labels and uncertain entity relationships, resulting in low overall quality of the constructed knowledge graph. For this scenario, the updating method of the knowledge graph in the embodiment performs quality evaluation on the knowledge to be updated according to the initial knowledge in the evidence storage engine, the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, and filters out high-quality knowledge to update the knowledge graph, so as to improve the overall quality of the constructed knowledge graph.

[0067] As an example, the application scenario of updating the knowledge graph not only includes the above-mentioned updating process of the knowledge graph of threat intelligence, but also includes various updating scenarios of the knowledge graph, which are not specifically limited herein.

[0068] The embodiment aims to improve the overall quality of the knowledge graph.

[0069] In the embodiment, the updating method of the knowledge graph is applied to the updating device of the knowledge graph.

[0070] The specific steps are as follows:

[0071] Step S100, obtaining an evidence storage engine and knowledge to be updated, wherein the evidence storage engine includes initial knowledge, evidence of the initial knowledge and information source of the evidence of the initial knowledge;

[0072] In the embodiment, the knowledge to be updated is network threat intelligence that has not been updated to the knowledge graph. The application verifies the quality of the knowledge to be updated to determine whether to update the knowledge to be updated to the knowledge graph.

[0073] In the embodiment, reference is made to Figure 4The data association structure of the evidence storage engine, the evidence storage engine is a preset network engine for assisting knowledge graph update, the evidence storage engine includes initial knowledge, evidence of the initial knowledge and information source of the evidence of the initial knowledge, and further includes association relationship data of the initial knowledge and the evidence of the initial knowledge, and association relationship data of the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, wherein the initial knowledge is knowledge stored in the current evidence storage engine, the information source is a source of information, including but not limited to a website and an institution, and the evidence is a source of an article containing knowledge under the information source, which can be a web link corresponding to the article.

[0074] The association relationship data of the knowledge and the evidence stores a many-to-many relationship of the knowledge ID and the evidence ID by using a relational database, the meaning of which is that one piece of knowledge can be associated with multiple evidences, and one evidence can be associated with multiple pieces of knowledge, the knowledge ID is generated by using a HASH algorithm according to the triple data (SPO) of each piece of knowledge, and the evidence ID is generated by using a HASH algorithm according to the unique features (such as the evidence text content) of the evidence data. In addition, each piece of evidence has three attributes, which are respectively: the total number of correct knowledge, the total number of knowledge, and the evidence credibility. The total number of knowledge represents the total number of all knowledge contained in the evidence; the total number of correct knowledge represents the number of correct knowledge associated; and the attribute value of the evidence credibility is the score of the evidence credibility, and the value interval is [0, 1].

[0075] The association relationship of the information source and the evidence, the information source represents the source of the evidence, such as a certain website. The information source and the evidence are in a one-to-many relationship. In addition, each information source has an attribute named information source credibility, and the attribute value is the score of the information source credibility, and the value interval is [0, 1].

[0076] The data storage of the evidence adopts a non-relational database storage and indexing, and can support storage of massive data of any kind or source, such as graphics, videos, texts and the like.

[0077] The evidence storage engine is established when the knowledge triple is extracted, and is perfected in the knowledge fusion stage. The initial knowledge-evidence-information source association engine is artificially constructed, the knowledge is correct knowledge verified by experts, the credibility score is 1, and the initial score of all information sources is updated in the expert verification stage, the attribute value c of the information source credibility of the information source s s The calculation formula is:

[0078]

[0079] Wherein, E s is all evidences from the information source S, T e and is the set of all knowledge derived from evidence e and the correct set of knowledge derived from evidence e; |·| represents the size of the set. Definitions of some symbols are shown in Table 1. In subsequent steps, the information source credibility attribute value is also calculated in the same way each time the information source is updated.

[0080] Table 1

[0081]

[0082]

[0083] The use of the evidence storage engine can be divided into three types. The first type is to find evidence according to search conditions. Since the evidence data has been indexed, the evidence data that meets the conditions can be quickly matched by fuzzy querying the relevant fields of the evidence storage, and the knowledge data associated with the evidence can also be quickly queried by the evidence ID. The second type is to find evidence according to knowledge. Since each knowledge ID is generated by a triple (SPO) and has uniqueness, the knowledge ID can be quickly generated by the user input knowledge data, the knowledge and evidence association table can be queried by the knowledge ID, a batch of evidence IDs can be obtained, and the evidence data can be quickly obtained by querying the evidence original data storage library. The third type is the most important one, which supports knowledge credibility verification.

[0084] In step S200, it is judged whether the to-be-updated knowledge meets the quality standard indicated by the preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge.

[0085] In this embodiment, the device judges whether the to-be-updated knowledge meets the quality standard indicated by the preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge, that is, the device performs quality evaluation on the to-be-updated knowledge according to the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge in the evidence storage engine, wherein the higher the credibility, the higher the quality of the knowledge, and the high-quality knowledge is screened out to update the knowledge graph, so as to improve the overall quality of the constructed knowledge graph.

[0086] Specifically, the step S200 includes steps S210-S230.

[0087] In step S210, a preset number of reference knowledge of the to-be-updated knowledge is selected from the evidence storage engine based on the initial knowledge.

[0088] In the embodiment, the reference knowledge refers to the knowledge similar to the to-be-updated knowledge in the initial knowledge of the current evidence storage engine, wherein the similarity is calculated according to attributes of the knowledge, and the attributes of the knowledge include entities, relationships and labels. For example, the knowledge is "ATP-18 attack exploits CVE-2009-001 vulnerability", and the entity relationship triple of the knowledge includes (ATP-12, exploits vulnerability, CVE-2009-01), wherein ATP-12 is an entity, exploits vulnerability is a relationship, and CVE-2009-001 is a label. ATP-12 can also be referred to as a main entity, and CVE-2009-001 can also be referred to as a guest entity. That is, the label can also be referred to as a guest entity, and the concept of the main entity is the same as that of the entity.

[0089] Before the step S210 of selecting the reference knowledge of the to-be-updated knowledge in the preset quantity from the evidence storage engine based on the initial knowledge, the method comprises steps A100-A300.

[0090] In the step A100, it is judged whether the duplicate knowledge of the to-be-updated knowledge exists in the evidence storage engine based on the initial knowledge.

[0091] In the embodiment, the device judges whether the duplicate knowledge of the to-be-updated knowledge exists in the evidence storage engine based on the initial knowledge. Specifically, the duplicate knowledge of the to-be-updated knowledge refers to the knowledge that is completely identical to the attributes (entity, relationship and label) of the to-be-updated knowledge. The duplicate knowledge is determined by searching the initial knowledge in the association graph of the evidence storage engine.

[0092] In the step A200, if the duplicate knowledge of the to-be-updated knowledge does not exist in the evidence storage engine, the reference knowledge of the to-be-updated knowledge in the preset quantity is selected from the evidence storage engine.

[0093] In the embodiment, if the duplicate knowledge of the to-be-updated knowledge does not exist in the evidence storage engine, the device selects the reference knowledge of the to-be-updated knowledge in the preset quantity from the evidence storage engine. The device selects the reference knowledge of the to-be-updated knowledge in the preset quantity from the evidence storage engine, which is described in the step S210 and will not be repeated here.

[0094] In the step A300, if the duplicate knowledge of the to-be-updated knowledge exists in the evidence storage engine, it is determined that the to-be-updated knowledge meets the quality standard.

[0095] In the embodiment, if the duplicate knowledge of the knowledge to be updated exists in the evidence storage engine, it is determined that the knowledge to be updated meets the quality standard, that is, the knowledge to be updated is consistent with the initial knowledge stored in the current evidence storage engine. Since the initial knowledge is all high-quality knowledge stored after screening, if the duplicate knowledge of the knowledge to be updated exists in the evidence storage engine, it indicates that the knowledge to be updated is high-quality knowledge, and the device can determine that the knowledge to be updated meets the quality standard. When the device updates the knowledge to be updated to the evidence storage engine, the attribute values of the two attributes of the knowledge corresponding to the evidence in the evidence storage engine, that is, the "correct knowledge total number" and the "knowledge total number", are both increased by 1, the information source credibility attribute value of the knowledge corresponding to the evidence is updated, and then the execution is ended.

[0096] Specifically, the step S210 includes steps S211-S213.

[0097] In the step S211, the first attribute information of the knowledge to be updated and the second attribute information of the initial knowledge are determined, wherein the attribute information includes entities, relations and labels.

[0098] In the embodiment, the attribute information includes entities, relations and labels, the second attribute information of the initial knowledge is the attribute information stored in the current evidence storage engine, and the first attribute information of the knowledge to be updated is determined when the knowledge to be updated is acquired.

[0099] In the step S212, the similarity between each initial knowledge and the knowledge to be updated is calculated based on the first attribute information and the second attribute information, and a similarity result is obtained, wherein the similarity calculation formula is as follows:

[0100]

[0101] Wherein, t represents knowledge, sim(t i ,t j ) represents the similarity between any two knowledge, sub, pre and obj represent the entity, relation and label of the knowledge respectively, cosim(a, b) represents the cosine similarity between any two entity vectors, γ1, γ2 and γ3 are coefficients, representing the weight of entity, relation and label respectively, and γ1+γ2+γ3=1.

[0102] In the embodiment, cosim(a, b) calculates the cosine similarity between two vectors, and the specific calculation is as follows:

[0103]

[0104] In this embodiment, the device calculates the similarity between each initial knowledge in the current evidence storage engine and the knowledge to be updated, and obtains the similarity result between each initial knowledge and the knowledge to be updated.

[0105] Step S213: The initial knowledge with the largest preset number in the similarity results is determined as the reference knowledge of the knowledge to be updated.

[0106] In this embodiment, the device determines the largest preset number of initial knowledge in the similarity results as the reference knowledge of the knowledge to be updated. Specifically, the similarity results of each initial knowledge and the knowledge to be updated are sorted from largest to smallest, and the preset number of initial knowledge at the top of the sort is determined as the reference knowledge of the knowledge to be updated. For example, if the current initial knowledge includes three A, B, and C, and A1, C1, and B1 are the similarities of A, B, and C respectively, and are sorted from largest to smallest as A1, C1, and B1, and the preset number is 2, then A and B are determined as the reference knowledge of the knowledge to be updated.

[0107] Step S220: Based on the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, determine the evidence of the reference knowledge and the information source of the evidence of the reference knowledge, and calculate the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge and the information source of the evidence of the reference knowledge;

[0108] In this embodiment, since the reference knowledge is selected from the set of initial knowledge, the device determines the evidence of the reference knowledge and the information source of the evidence of the initial knowledge based on the evidence of the initial knowledge and the information source of the evidence of the initial knowledge.

[0109] In this embodiment, the device calculates a first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge, and the information source of the evidence of the reference knowledge. The verification probability is data that reflects the reliability (or credibility) of the knowledge. The higher the verification probability, the higher the reliability of the knowledge to be updated. This quantifies the reliability of the knowledge and facilitates the selection of high-quality knowledge.

[0110] Specifically, step S220 includes the following steps S221-S223:

[0111] Step S221: Determine the credibility of the information source of the evidence for the reference knowledge;

[0112] In this embodiment, the device determines the credibility of the information source of the evidence of the reference knowledge, wherein the credibility of the information source is the information stored in the current evidence storage engine, which is updated according to the knowledge to be updated.

[0113] Step S222, based on the evidence of the reference knowledge and the credibility of the information source of the evidence of the reference knowledge, a verification probability calculation is performed on each of the reference knowledge to obtain a second verification probability of each of the reference knowledge, wherein the verification probability calculation formula is as follows:

[0114]

[0115] wherein t represents the reference knowledge, E t is the evidence set of the reference knowledge, E is the evidence set of the initial knowledge, is the credibility of the information source of the evidence, |·| represents the size of the set, P(t) represents the second verification probability;

[0116] In this embodiment, the device performs a verification probability calculation on each of the reference knowledge based on the evidence of the reference knowledge and the credibility of the information source of the evidence of the reference knowledge to obtain a second verification probability of each of the reference knowledge, wherein the second verification probability is calculated according to the evidence of the reference knowledge and the credibility information of the evidence.

[0117] Step S223, a weighted sum calculation is performed on the second verification probability of each of the reference knowledge to obtain a first verification probability of the knowledge to be updated, wherein the weighted sum calculation formula is as follows:

[0118]

[0119] wherein t c represents the evidence of the knowledge to be updated, P(t c ) represents the first verification probability of the knowledge to be updated.

[0120] In this embodiment, the device performs a weighted sum calculation on the second verification probability of each of the reference knowledge to obtain a first verification probability of the knowledge to be updated, that is, the device performs a total weighted sum on the second verification probability of the reference knowledge of the knowledge to be updated to obtain the first verification probability of the knowledge to be updated.

[0121] Step S230, based on the first verification probability, it is judged whether the knowledge to be updated satisfies the quality standard indicated by the preset credibility threshold range.

[0122] In this embodiment, based on the first verification probability, it is judged whether the knowledge to be updated satisfies the quality standard indicated by the preset credibility threshold range, specifically, the first verification probability is compared with a preset probability threshold in size, if the first verification probability is greater than the probability threshold, it is determined that the knowledge to be updated satisfies the quality standard indicated by the preset credibility threshold range, wherein the probability threshold can be adjusted by itself.

[0123] Specifically, the step S230 includes steps S231-S234.

[0124] The step S231 judges whether there is the conflicting knowledge of the knowledge to be updated in the evidence storage engine based on the initial knowledge.

[0125] In this embodiment, the conflicting knowledge is the knowledge consistent with the entity and the relationship of the initial knowledge but inconsistent in the label. The device judges whether there is the conflicting knowledge of the knowledge to be updated in the evidence storage engine based on the first attribute information of the initial knowledge and the second attribute information of the knowledge to be updated. It should be noted that there is only one conflicting knowledge. After the main entity and the relationship are determined, the label is a unique value. The evidence storage engine has screened this, and ensured that there is only one conflicting knowledge in the knowledge database.

[0126] The step S232 calculates the third verification probability of the conflicting knowledge if there is the conflicting knowledge of the knowledge to be updated in the evidence storage engine.

[0127] In this embodiment, the third verification probability of the conflicting knowledge is calculated if there is the conflicting knowledge of the knowledge to be updated in the evidence storage engine. Specifically, the verification probability of the conflicting knowledge is calculated, and the conflicting knowledge is represented as t a The verification probability of t a is:

[0128]

[0129] wherein, is the evidence set of the conflicting knowledge t a , E is the set of all evidences, is the credibility of the information source of the evidence, |·| represents the size of the set, and P(t a ) represents the third verification probability.

[0130] The step S233 compares the first verification probability with the third verification probability. If the first verification probability is greater than the third verification probability, it is determined that the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range.

[0131] In this embodiment, the device compares the first verification probability with the third verification probability. If the first verification probability is greater than the third verification probability, it is determined that the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range. Specifically, the device compares the first verification probability of the knowledge to be updated with the third verification probability of the conflicting knowledge, and judges whether the knowledge to be updated or the conflicting knowledge is correct knowledge.

[0132] If the first verification probability is greater than the third verification probability, that is, P(tci )>P(t a If the knowledge to be updated is t, then it is pending. c If the knowledge is correct (high-quality), add it to the knowledge database and establish a connection between tc and ec in the association graph. Delete conflicting knowledge ta and all evidence corresponding to ta. The "Total Correct Knowledge" attribute value was reduced by 1, and the update was performed. The execution process ends when the credibility attribute value of all evidence sources is determined.

[0133] If the first verification probability is less than or equal to the third verification probability, i.e., P(t) ci )≤P(t a If the knowledge to be updated is t, then it is pending. ci If the knowledge is not correct, it will not be added to the knowledge database, but it needs to be updated. c Corresponding evidence e c The attribute value, e c Total knowledge +1, update e c Information source s c The execution process ends when the information source credibility attribute value is determined.

[0134] Step S234: If there is no conflicting knowledge of the knowledge to be updated in the evidence storage engine, the first verification probability is compared with a preset probability threshold. If the first verification probability is greater than the probability threshold, the knowledge to be updated is determined to meet the quality standard indicated by the preset credibility threshold range.

[0135] In this embodiment, if there is no conflicting knowledge of the knowledge to be updated in the evidence storage engine, the first verification probability is compared with a preset probability threshold. If the first verification probability is greater than the probability threshold, i.e., P(t) c If )>α, then the knowledge to be updated t ci If it is correct knowledge, add it to the knowledge database and update the evidence storage engine. c The "Total Knowledge" and "Total Correct Knowledge" attributes are both incremented by 1, and e is updated. c Information source s c The information source credibility attribute value. α is a preset probability threshold, a hyperparameter that can be manually adjusted; it is recommended to set it in the range [0.5, 0.8]. Then the execution process ends.

[0136] If the first verification probability is less than or equal to the probability threshold, i.e., P(t) ci If t ≤ α, then the candidate knowledge t c If the knowledge is not correct, it will not be added to the knowledge database, but it needs to be updated. c Corresponding evidence e cithe attribute value of the total number of knowledge of e c the attribute value of the total number of knowledge of e is increased by 1, and the attribute value of the correct total number of knowledge is unchanged c the information source s c the attribute value of the credibility of the information source, and the execution flow is ended.

[0137] In step S300, if the knowledge to be updated meets the quality standard, the knowledge to be updated is updated to the evidence storage engine and the target knowledge graph.

[0138] In this embodiment, if the knowledge to be updated meets the quality standard, the device updates the knowledge to be updated to the evidence storage engine and the target knowledge graph, that is, the knowledge to be updated is updated to the evidence storage engine for assisting the update of the knowledge graph and the target knowledge graph, so as to provide a high-precision quality detection standard and a high-quality knowledge graph.

[0139] The knowledge graph updating method provided in the present application is compared with the related art in which network threat intelligence data has the characteristics of large volume, wide source, and fast update, resulting in a large amount of redundancy and inconsistency of new knowledge extracted by an information extraction system, and low overall quality of a constructed knowledge graph. In the present application, an evidence storage engine and knowledge to be updated are obtained, wherein the evidence storage engine includes initial knowledge, evidence of the initial knowledge, and an information source of the evidence of the initial knowledge; whether the knowledge to be updated meets a quality standard indicated by a preset credibility threshold range is determined based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge; and if the knowledge to be updated meets the quality standard, the knowledge to be updated is updated to the evidence storage engine and a target knowledge graph. That is, in the present application, the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge in the evidence storage engine are used to evaluate the quality of the knowledge to be updated, and high-quality knowledge is selected to update the knowledge graph, so as to improve the overall quality of the constructed knowledge graph.

[0140] Based on the above first embodiment, the present application further provides another embodiment, which is described with reference to Figure 5 applied in a knowledge updating stage, the knowledge graph updating method includes the following steps B100-B900:

[0141] In step B100, candidate knowledge and repeated knowledge of the candidate knowledge are obtained.

[0142] The candidate knowledge is new knowledge, which is equivalent to the knowledge to be updated in the above first embodiment, and is replaced by the candidate knowledge hereinafter; the repeated knowledge is knowledge with consistent entities, relationships, and labels selected from existing knowledge.

[0143] Step B200, judge whether the duplicate knowledge is empty, if the duplicate set is not empty, execute step B300; otherwise execute step B400;

[0144] Step B300, store the evidence e ci corresponding to the candidate knowledge into the evidence engine, and update the link relationship of e ci and the duplicate knowledge. The labels of the two attributes of the correct knowledge total number and the knowledge total number of e ci are both +1. Update the link relationship of e ci and the information source, if the information source exists, link the relationship between e ci and the information source ID; if the information source does not exist, add a new information source and information source ID, and link e ci , end execution.

[0145] Step B400, construct the reference knowledge set of the candidate knowledge, and obtain the evidence set of all reference knowledge:

[0146] Since there is no knowledge consistent with the candidate knowledge entity and relationship in the existing knowledge, we construct the reference set of the candidate knowledge. We define the candidate knowledge as t c , and use sub c , pre c and obj c to represent the entity, relationship and label of t c . The similarity between any two knowledge is defined as:

[0147]

[0148] Where γ1, γ2 and γ3 are coefficients, representing the contribution of entity, relationship and label to the final similarity, and γ1+γ2+γ3=1. cosim(a, b) calculates the cosine similarity of two vectors:

[0149]

[0150] Calculate the similarity of all existing knowledge and the candidate knowledge, and select the k most similar ones as reference knowledge, denoted as set T c ;

[0151] Step B500, calculate the verification probability of the candidate knowledge. Since the evidence of the candidate knowledge at this time is only one, and the information source is not necessarily stored in the evidence storage engine, we estimate the verification probability of the candidate knowledge through the reference knowledge set. For any reference knowledge t∈T c , the verification probability calculation formula is:

[0152]

[0153] Where Et is the evidence set of knowledge t t is the information source credibility of e, |·| represents the size of the set. The verification probability of candidate knowledge t ci is the weighted sum of all reference knowledge verification probabilities, denoted as:

[0154]

[0155] Step B600, get the conflict knowledge of the candidate knowledge;

[0156] The conflict knowledge is selected from the existing knowledge that the entity and the relationship are consistent but the label is inconsistent.

[0157] Step B700, judge whether the conflict knowledge is empty. If the conflict knowledge is not empty, execute step B800; otherwise, execute B900:

[0158] Step B800, calculate the verification probability of the sum of the conflict knowledge. Compare the verification probability of the candidate knowledge and the conflict knowledge to determine whether the candidate knowledge or the conflict knowledge is the correct knowledge. The verification probability of the conflict knowledge t a is:

[0159]

[0160] where is the evidence set of knowledge t a , E is the set of all evidence, is the information source credibility of e, |·| represents the size of the set.

[0161] If P(t ci )>P(t a ), the candidate knowledge t ci is the correct knowledge, which is added to the knowledge database, and the evidence storage engine is updated. Delete knowledge t a , t a corresponding to the evidence e a “correct knowledge total number” attribute value-1, e a and e a information source credibility update;

[0162] If P(t ci )<P(t a ), the candidate knowledge t ci is not the correct knowledge, which is not added to the knowledge database, but the attribute value of e ci corresponding to the evidence e ci needs to be updated, e ci knowledge total number+1, e ci and e ci ​the information source of the evidence e

[0163] End the execution flow.

[0164] Step B900, judge whether the set candidate knowledge is correct knowledge or not;

[0165] If P(t ci )>a, the candidate knowledge t ci is correct knowledge, which is added to the knowledge database, and the evidence storage engine is updated. Delete the knowledge t a , t a corresponding evidence e a The attribute value of the "total number of correct knowledge" of e a and e a The trust score of the information source is updated. Alpha is a human-adjusted hyperparameter, and the recommended setting interval is [0.5, 0.8];

[0166] End the execution flow.

[0167] In this embodiment, the present application provides an evidence storage engine, establishes the relationship among knowledge, evidence and information source, and uses the evidence and information source in the evidence storage engine to reflect the reliability of knowledge, so as to quickly evaluate the credibility of new knowledge, filter high-quality knowledge, and update the knowledge graph, so as to improve the overall quality of the constructed knowledge graph.

[0168] The present application also provides a knowledge graph updating device, which comprises:

[0169] An acquisition module 10 is configured to acquire an evidence storage engine and to-be-updated knowledge, wherein the evidence storage engine comprises initial knowledge, evidence of the initial knowledge, and information source of the evidence of the initial knowledge;

[0170] A judgment module 20 is configured to judge whether the to-be-updated knowledge meets a quality standard indicated by a preset credibility threshold range based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge;

[0171] An updating module 30 is configured to update the to-be-updated knowledge to the evidence storage engine if the to-be-updated knowledge meets the quality standard.

[0172] Optionally, the judgment module 20 comprises:

[0173] A selection module is configured to select a preset number of reference knowledge of the to-be-updated knowledge from the evidence storage engine based on the initial knowledge;

[0174] The first calculation module is configured to determine the evidence of the reference knowledge and the information source of the evidence of the reference knowledge based on the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, and calculate the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge and the information source of the evidence of the reference knowledge.

[0175] The verification module is configured to determine whether the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range based on the first verification probability.

[0176] Optionally, the first calculation module comprises:

[0177] The credibility determination module is configured to determine the credibility of the information source of the evidence of the reference knowledge.

[0178] The second calculation module is configured to perform verification probability calculation on each reference knowledge based on the evidence of the reference knowledge and the credibility of the information source of the evidence of the reference knowledge, to obtain the second verification probability of each reference knowledge, wherein the verification probability calculation formula is as follows:

[0179]

[0180] wherein t represents the reference knowledge, E t is the evidence set of the reference knowledge, E is the evidence set of the initial knowledge, is the credibility of the information source of the evidence, |·| represents the size of the set, and P(t) represents the second verification probability.

[0181] The weighted sum calculation module is configured to perform weighted sum calculation on the second verification probability of each reference knowledge to obtain the first verification probability of the knowledge to be updated, wherein the weighted sum calculation formula is as follows:

[0182]

[0183] wherein t c represents the evidence of the knowledge to be updated, P(t c ) represents the first verification probability of the knowledge to be updated.

[0184] Optionally, the selection module comprises:

[0185] The attribute information determination module is configured to determine the first attribute information of the knowledge to be updated and the second attribute information of the initial knowledge, wherein the attribute information comprises entities, relationships and labels.

[0186] The similarity calculation module is configured to calculate the similarity between each of the initial knowledge and the knowledge to be updated based on the first attribute information and the second attribute information, and obtain a similarity result, where a similarity calculation formula is as follows:

[0187]

[0188] where t represents knowledge, sim(t i , t j ) represents the similarity between any two knowledge, sub, pre and obj represent the entity, the relationship and the label of the knowledge respectively, sim (sub, sub') = cos (sub, sub') represents the cosine similarity between any two entity vectors, γ1, γ2 and γ3 are coefficients, representing the weight of the entity, the relationship and the label respectively, and γ1+γ2+γ3=1;

[0189] The reference knowledge determination module is configured to determine the initial knowledge with the maximum similarity result as the reference knowledge of the knowledge to be updated.

[0190] Optionally, the verification module comprises:

[0191] The conflict knowledge judgment module is configured to judge whether there is the conflict knowledge of the knowledge to be updated in the evidence storage engine based on the initial knowledge.

[0192] The third calculation module is configured to calculate the third verification probability of the conflict knowledge if there is the conflict knowledge of the knowledge to be updated in the evidence storage engine.

[0193] The first comparison verification module is configured to compare the first verification probability and the third verification probability in size, and determine that the knowledge to be updated meets the quality standard indicated by the preset confidence threshold range if the first verification probability is greater than the third verification probability.

[0194] Optionally, the verification module further comprises:

[0195] The second comparison verification module is configured to compare the first verification probability and a preset probability threshold in size if there is no conflict knowledge of the knowledge to be updated in the evidence storage engine, and determine that the knowledge to be updated meets the quality standard indicated by the preset confidence threshold range if the first verification probability is greater than the probability threshold.

[0196] The judgment module 20 further comprises:

[0197] The repeated knowledge judgment module is configured to judge whether there is the repeated knowledge of the knowledge to be updated in the evidence storage engine based on the initial knowledge.

[0198] The reference knowledge selection module is configured to select reference knowledge of the to-be-updated knowledge from the evidence storage engine in a preset number if the duplicate knowledge of the to-be-updated knowledge does not exist in the evidence storage engine.

[0199] The standard meeting determination module is configured to determine that the to-be-updated knowledge meets the quality standard if the duplicate knowledge of the to-be-updated knowledge exists in the evidence storage engine.

[0200] The updating device for the knowledge graph of the present application has basically the same implementation as the above-mentioned embodiments of the method for updating the knowledge graph, and will not be described here.

[0201] Referring to Figure 1 , Figure 1 is a terminal structure diagram of a hardware running environment involved in the embodiment of the present application.

[0202] As Figure 1 shown, the terminal can include a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the components. The user interface 1003 can include a display screen (Display) and an input unit such as a keyboard (Keyboard). The optional user interface 1003 can also include a standard wired interface and a wireless interface. The network interface 1004 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 can be a high-speed RAM memory or a stable memory (non-volatile memory) such as a magnetic disk memory. The memory 1005 can also be a storage device independent of the aforementioned processor 1001.

[0203] Optionally, the updating device for the knowledge graph can also include a rectangular user interface, a network interface, a camera, an RF (Radio Frequency, Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, etc. The rectangular user interface can include a display screen (Display) and an input sub-module such as a keyboard (Keyboard). The optional rectangular user interface can also include a standard wired interface and a wireless interface. The network interface can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0204] Those skilled in the art can understand that Figure 1 the structure of the updating device for the knowledge graph shown in the above-mentioned embodiments does not constitute a limitation on the updating device for the knowledge graph, and can include more or fewer components than the diagram, or combine certain components, or different component arrangements.

[0205] As Figure 1As shown, the memory 1005 as a storage medium can include an operating system, a network communication module, and a knowledge graph updating program. The operating system is a program for managing and controlling hardware and software resources of the knowledge graph updating device, supporting the running of the knowledge graph updating program and other software and / or programs. The network communication module is used to realize the communication between the components in the memory 1005, and the communication with other hardware and software in the knowledge graph updating system.

[0206] In Figure 1 In the knowledge graph updating device shown, the processor 1001 is used to execute the knowledge graph updating program stored in the memory 1005, and realize the steps of the knowledge graph updating method described in any of the above embodiments.

[0207] The specific implementation of the knowledge graph updating device of the present application is basically the same as that of the above-mentioned knowledge graph updating method, and will not be repeated here.

[0208] The present application also provides a storage medium, which stores a program for implementing a knowledge graph updating method, and the program is executed by a processor to implement the knowledge graph updating method as follows:

[0209] The specific implementation of the storage medium of the present application is basically the same as that of the above-mentioned knowledge graph updating method, and will not be repeated here.

[0210] The present application also provides a computer program product, which includes a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned knowledge graph updating method.

[0211] The specific implementation of the computer program product of the present application is basically the same as that of the above-mentioned knowledge graph updating method, and will not be repeated here.

[0212] It should be noted that in this paper, the term "including", "containing" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the sentence "including a…" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0213] The above-mentioned embodiment number of the present application is only for description, not representing the advantages and disadvantages of the embodiments.

[0214] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, can also be through hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of the prior art contribution can be embodied in the form of software products, the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disc, optical disc) as described above, including a number of instructions to make a terminal device (may be a mobile phone, computer, server, air conditioner, or network equipment, etc.) executes the method described in various embodiments of the present application.

[0215] The above is only the preferred embodiment of the present application, not therefore limit the patent scope of the present application, all the equivalent structure or equivalent process transformation, or directly or indirectly applied in other related technical fields, using the contents of the present application specification and drawings, are also included in the patent protection scope of the present application.

Claims

1. A method for updating a knowledge graph, characterized in that, The methods for updating the knowledge graph include: The evidence storage engine and the knowledge to be updated are obtained, wherein the evidence storage engine includes initial knowledge, evidence of the initial knowledge, and information sources of the evidence of the initial knowledge; Based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge, determine whether the knowledge to be updated meets the quality standard indicated by the preset credibility threshold range; The step of determining whether the knowledge to be updated meets the quality standard indicated by a preset credibility threshold range based on the initial knowledge, evidence of the initial knowledge, and information sources of the evidence of the initial knowledge includes: Based on the initial knowledge, a preset number of reference knowledge for the knowledge to be updated are selected from the evidence storage engine; Based on the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, determine the evidence of the reference knowledge and the information source of the evidence of the reference knowledge, and calculate the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge and the information source of the evidence of the reference knowledge; The step of calculating the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge, and the information source of the evidence of the reference knowledge includes: Determine the credibility of the information source of the evidence for the reference knowledge; Based on the evidence of the reference knowledge and the credibility of the information source of the evidence of the reference knowledge, a verification probability calculation is performed on each of the reference knowledge to obtain a second verification probability for each of the reference knowledge, wherein the verification probability calculation formula is as follows: Where t represents reference knowledge. It is a collection of evidence referencing knowledge. It is a set of evidence for initial knowledge. It refers to the credibility of the information source of the evidence. Indicates the size of the set. Indicates the second verification probability; The weighted sum of the second verification probabilities of each of the reference knowledge pieces is used to obtain the first verification probability of the knowledge to be updated, wherein the weighted sum calculation formula is as follows: in, Evidence indicating knowledge that needs to be updated This represents the first verification probability of the knowledge to be updated. Based on the first verification probability, determine whether the knowledge to be updated meets the quality standard indicated by the preset confidence threshold range; If the knowledge to be updated meets the quality standard, then the knowledge to be updated will be updated in the evidence storage engine and the target knowledge graph.

2. The knowledge graph updating method as described in claim 1, characterized in that, The step of selecting a preset number of reference knowledge for the knowledge to be updated from the evidence storage engine includes: Determine the first attribute information of the knowledge to be updated and the second attribute information of the initial knowledge, wherein the attribute information includes entities, relationships and tags; Based on the first attribute information and the second attribute information, a similarity calculation is performed between each initial knowledge and the knowledge to be updated to obtain a similarity result. The similarity calculation formula is as follows: in, Representing knowledge, This represents the similarity between any two pieces of knowledge. , and These represent entities, relations, and tags of knowledge, respectively. Let cosine similarity represent the similarity between any two entity vectors. , and These are coefficients, representing the weights of entities, relations, and labels, respectively. ; The initial knowledge with the largest preset number in the similarity results is determined as the reference knowledge for the knowledge to be updated.

3. The knowledge graph updating method as described in claim 1, characterized in that, The step of determining whether the knowledge to be updated meets the quality standard indicated by the preset confidence threshold range based on the first verification probability includes: Based on the initial knowledge, determine whether there is conflicting knowledge in the evidence storage engine that is to be updated; If conflicting knowledge exists in the evidence storage engine, then the third verification probability of the conflicting knowledge is calculated. The first verification probability is compared with the third verification probability. If the first verification probability is greater than the third verification probability, then the knowledge to be updated is determined to meet the quality standard indicated by the preset confidence threshold range.

4. The knowledge graph updating method as described in claim 3, characterized in that, After the step of determining whether conflicting knowledge exists in the evidence storage engine based on the initial knowledge, the method includes: If there is no conflicting knowledge of the knowledge to be updated in the evidence storage engine, the first verification probability is compared with a preset probability threshold. If the first verification probability is greater than the probability threshold, the knowledge to be updated is determined to meet the quality standard indicated by the preset credibility threshold range.

5. The knowledge graph updating method as described in claim 1, characterized in that, Before the step of selecting a preset number of reference knowledge points for the knowledge to be updated from the evidence storage engine based on the initial knowledge, the method includes: Based on the initial knowledge, determine whether there is duplicate knowledge of the knowledge to be updated in the evidence storage engine; If there is no duplicate knowledge of the knowledge to be updated in the evidence storage engine, then a preset number of reference knowledge of the knowledge to be updated are selected from the evidence storage engine. If duplicate knowledge of the knowledge to be updated exists in the evidence storage engine, then the knowledge to be updated is determined to meet the quality standard.

6. A knowledge graph updating device, characterized in that, The knowledge graph updating device includes: An acquisition module is used to acquire an evidence storage engine and knowledge to be updated, wherein the evidence storage engine includes initial knowledge, evidence of the initial knowledge, and information sources of the evidence of the initial knowledge; The judgment module is used to determine, based on the initial knowledge, the evidence of the initial knowledge, and the information source of the evidence of the initial knowledge, whether the knowledge to be updated meets the quality standard indicated by a preset credibility threshold range. The step of determining whether the knowledge to be updated meets the quality standard indicated by a preset credibility threshold range based on the initial knowledge, evidence of the initial knowledge, and information sources of the evidence of the initial knowledge includes: Based on the initial knowledge, a preset number of reference knowledge for the knowledge to be updated are selected from the evidence storage engine; Based on the evidence of the initial knowledge and the information source of the evidence of the initial knowledge, determine the evidence of the reference knowledge and the information source of the evidence of the reference knowledge, and calculate the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge and the information source of the evidence of the reference knowledge; The step of calculating the first verification probability of the knowledge to be updated based on the reference knowledge, the evidence of the reference knowledge, and the information source of the evidence of the reference knowledge includes: Determine the credibility of the information source of the evidence for the reference knowledge; Based on the evidence of the reference knowledge and the credibility of the information source of the evidence of the reference knowledge, a verification probability calculation is performed on each of the reference knowledge to obtain a second verification probability for each of the reference knowledge, wherein the verification probability calculation formula is as follows: Where t represents reference knowledge. It is a collection of evidence referencing knowledge. It is a set of evidence for initial knowledge. It refers to the credibility of the information source of the evidence. Indicates the size of the set. Indicates the second verification probability; The weighted sum of the second verification probabilities of each of the reference knowledge pieces is used to obtain the first verification probability of the knowledge to be updated, wherein the weighted sum calculation formula is as follows: in, Evidence indicating knowledge that needs to be updated This represents the first verification probability of the knowledge to be updated. Based on the first verification probability, determine whether the knowledge to be updated meets the quality standard indicated by the preset confidence threshold range; An update module is used to update the knowledge to be updated to the evidence storage engine and the target knowledge graph if the knowledge to be updated meets the quality standard.

7. A knowledge graph updating device, characterized in that, The knowledge graph updating device includes: a memory, a processor, and a program stored in the memory for implementing the knowledge graph updating method. The memory is used to store programs that implement the knowledge graph update method; The processor is configured to execute a program that implements the knowledge graph update method to implement the steps of the knowledge graph update method as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores a program that implements a method for updating a knowledge graph, which is executed by a processor to implement the steps of the method for updating a knowledge graph as described in any one of claims 1 to 5.

Images