Content moderation methods, devices, electronic equipment, and storage media

Abstract

This disclosure relates to a content review method, apparatus, electronic device, and storage medium. The method includes: acquiring review object data; extracting triples to be reviewed based on the review object data, wherein each triple includes a first tag and a second tag with an association relationship; determining, based on the first and second tags of the triples to be reviewed, standard triples corresponding to the triples to be reviewed in a pre-constructed knowledge graph; the knowledge graph includes a coordinate system and multiple standard triples filling the coordinate system; each standard triple includes a first standard tag, a second standard tag, and a connecting line connecting the first and second standard tags, wherein the coordinates of the extreme points of the connecting lines reflect the overall evaluation result of the standard triples; and determining a review conclusion for the review object data based on the coordinates of the extreme points of the connecting lines corresponding to the standard triples to be reviewed, which can improve the efficiency of content review.

Description

Technical Field

[0001] This disclosure relates to the field of content moderation technology, and in particular to a content moderation method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the development of the internet, while online platforms have experienced a massive explosion of content, they have also generated a large amount of harmful information.

[0003] Against this backdrop, how to conduct content review efficiently and quickly is an urgent problem to be solved. Summary of the Invention

[0004] To address, or at least partially address, the aforementioned technical problems, this disclosure provides a content moderation method, apparatus, electronic device, and storage medium.

[0005] Firstly, this disclosure provides a content moderation method, including:

[0006] Obtain the data of the audited entity;

[0007] Based on the data of the audit object, extract the triples to be audited, wherein the triples to be audited include a first tag to be audited and a second tag to be audited that have an association relationship;

[0008] Based on the first label and the second label of the triple to be reviewed, a connecting line is determined in the pre-constructed knowledge graph to correspond to the standard triple to be reviewed; the knowledge graph includes a coordinate system and multiple standard triples filled in the coordinate system; each standard triple includes a first standard label, a second standard label, and a connecting line connecting the first standard label and the second standard label, and the coordinates of the extreme points of the connecting line reflect the overall evaluation result of the standard triple;

[0009] Based on the coordinates of the extreme points of the connecting lines of the standard triples corresponding to the triples to be audited, the audit conclusion for the audit object data is determined.

[0010] Secondly, this disclosure also provides a content moderation device, including:

[0011] The audit object acquisition module is used to acquire audit object data;

[0012] The module for extracting triples to be reviewed is used to extract triples to be reviewed based on the data of the review object. The triples to be reviewed include a first tag to be reviewed and a second tag to be reviewed that have an association relationship.

[0013] The connector determination module is used to determine, based on the first label and the second label of the triplet to be reviewed, a connector line in a pre-constructed knowledge graph corresponding to the triplet to be reviewed; the knowledge graph includes a coordinate system and multiple standard triplets filled in the coordinate system; each standard triplet includes a first standard label, a second standard label, and a connector line connecting the first standard label and the second standard label, and the coordinates of the extreme points of the connector line reflect the overall evaluation result of the standard triplet;

[0014] The review module is used to determine the review conclusion for the data to be reviewed based on the coordinates of the extreme points of the connection line of the standard triplet corresponding to the triplet to be reviewed.

[0015] Thirdly, this disclosure also provides an electronic device, including: a processor and a memory;

[0016] The processor executes the steps of any of the above methods by calling programs or instructions stored in memory.

[0017] Fourthly, this disclosure also provides a computer-readable storage medium that stores a program or instructions that cause a computer to perform the steps of any of the above methods.

[0018] The technical solution provided in this disclosure has the following advantages compared with the prior art:

[0019] The technical solution provided in this disclosure involves setting up a system to extract triples to be reviewed based on the data of the review object; determining the connection line between the standard triples corresponding to the triples to be reviewed in a pre-constructed knowledge graph based on the first and second tags of the triples to be reviewed; and determining the review conclusion for the data of the review object based on the coordinates of the extreme points of the connection line between the standard triples corresponding to the triples to be reviewed. Essentially, this system achieves automated content review based on a knowledge graph. Compared to manual review, it can significantly improve the efficiency of content review and reduce the labor and economic costs of content review.

[0020] The technical solutions provided in this disclosure can achieve ad filtering, prohibited content review, abuse identification, and spam detection by leveraging different types of knowledge graphs.

[0021] The technical solution provided in this disclosure essentially utilizes the extreme points of the connecting lines to reflect the implicit meaning generated by the combination of the first and second standard labels in the standard triplet. This allows the knowledge graph to reflect not only the attributes of each label in the standard triplet but also the implicit meaning of the standard triplet as a whole. This enhances the application value of knowledge graph technology in content moderation. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0023] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart of a content moderation method provided in this embodiment of the disclosure;

[0025] Figure 2 This is a schematic diagram of a pre-constructed knowledge graph provided in this disclosure;

[0026] Figure 3 A flowchart illustrating the steps for constructing a knowledge graph in a content moderation method provided in this disclosure embodiment;

[0027] Figure 4 and Figure 5 A schematic diagram of two coordinate systems provided in the embodiments of this disclosure;

[0028] Figure 6 A schematic diagram of another coordinate system provided in an embodiment of this disclosure;

[0029] Figure 7 A flowchart illustrating the steps for constructing a knowledge graph in another content moderation method provided in this disclosure embodiment;

[0030] Figure 8 This is a structural block diagram of a content moderation device proposed in an embodiment of this disclosure;

[0031] Figure 9 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0032] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0033] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0034] Figure 1 This is a flowchart illustrating a content moderation method provided in an embodiment of this disclosure. The method is executed by an electronic device, specifically a terminal or a server. The method includes:

[0035] S1. Obtain the data of the audit target.

[0036] Optionally, the data to be reviewed may include at least one of text, video, audio, and images.

[0037] S2. Extract the triples to be reviewed based on the data of the review object. The triples to be reviewed include the first tag and the second tag to be reviewed that are related.

[0038] The tentative triple is the basic unit for content review. The three elements that make up the tentative triple are the first tag to be reviewed, the second tag to be reviewed, and the relationship between the first tag to be reviewed and the second tag to be reviewed.

[0039] There are various forms of the triplet to be reviewed, and this disclosure does not impose any restrictions on them.

[0040] For example, if the triple to be reviewed is represented as "entity-relationship-entity", the first tag to be reviewed and the second tag to be reviewed refer to two entities respectively. For example, if a triple to be reviewed can be represented as "BB-parent-AA", where "BB" and "AA" represent names, then in this triple to be reviewed, the first tag to be reviewed is BB, and the second tag to be reviewed is AA.

[0041] If the triple to be reviewed is represented as "entity-attribute-attribute value", the first tag to be reviewed refers to the entity, and the second tag to be reviewed refers to the attribute value. For example, if a triple to be reviewed can be represented as "BB-education level-bachelor", where "BB" represents a person's name, then in this triple to be reviewed, the first tag to be reviewed is BB, and the second tag to be reviewed is bachelor's degree.

[0042] The specific implementation method of this step is to extract information from the object to be reviewed (such as text recognition, image recognition, speech recognition or tag recognition), and perform logical operations based on the recognition results to finally obtain the triplet to be reviewed.

[0043] S3. Based on the first label and the second label of the triple to be audited, determine the connection lines of the standard triples corresponding to the triples to be audited in the pre-constructed knowledge graph.

[0044] Knowledge graphs are a technical approach that uses graph models to describe knowledge and construct the relationships between everything. They consist of nodes and edges, forming a network-like knowledge structure. Nodes can be entities or abstract concepts. Edges can be attributes of entities or relationships between entities.

[0045] For example, Figure 2 This is a schematic diagram of a pre-constructed knowledge graph provided in this disclosure. See also... Figure 2 The knowledge graph includes a coordinate system consisting of an X-axis and a Y-axis, where the X-axis coordinates are determined based on the overall evaluation results of standard triples. The knowledge graph also includes multiple standard triples filling the coordinate system. Each standard triple includes a first standard label, a second standard label, and a connecting line linking the first and second standard labels. The coordinates of the extreme points of the connecting line reflect the overall evaluation results of the standard triples. For example, in... Figure 2 Only two standard triplets are shown. The first standard triplet has the first standard label "XX Medical Device Factory" and the second standard label "Mask". "XX Medical Device Factory" and "Mask" are connected by line C1, with the extreme point of line C1 being M1. The X-axis coordinate of extreme point M1 is "Illegal Daily Necessities Advertisement". This indicates that the overall evaluation result of the first standard triplet is "Illegal Daily Necessities Advertisement". The second standard triplet has the first standard label "YY Hospital" and the second standard label "Medical Insurance Designated Hospital". "YY Hospital" and "Medical Insurance Designated Hospital" are connected by line C2, with the extreme point of line C2 being M2. The X-axis coordinate of extreme point M2 is "Illegal Medical Service Advertisement". This indicates that the overall evaluation result of the second standard triplet is "Illegal Medical Service Advertisement".

[0046] The specific implementation method of this step is as follows: In the knowledge graph, find the first standard label that corresponds to the first label to be reviewed, and find the second standard label that corresponds to the second label to be reviewed. At this point, the standard triplet, including the first standard label corresponding to the first label to be reviewed and the second standard label corresponding to the second label to be reviewed, is the standard triplet corresponding to the triplet to be reviewed. Based on this, determine the connection lines between the standard triplets corresponding to the triplet to be reviewed.

[0047] The "correspondence" in "the first standard label that corresponds to the first label to be reviewed" and "the second standard label that corresponds to the second label to be reviewed" can be specifically understood as being consistent, or that the two are synonyms, near-synonyms, or words with a hierarchical relationship.

[0048] S4. Based on the coordinates of the extreme points of the connecting lines of the standard triples corresponding to the triples to be audited, determine the audit conclusion for the data of the audit object.

[0049] See also Figure 2 If the first label of the pending triplet is "XX Medical Device Factory" and the second label is "Mask", then the connecting line to the standard triplet corresponding to this pending triplet is C1, and the extreme point is M1. The coordinate of the extreme point M1 on the X-axis is "Illegal Daily Necessities Advertisement". This indicates that the review conclusion for this pending triplet is "Illegal Daily Necessities Advertisement".

[0050] The above technical solution extracts triples to be reviewed based on the data of the review object; determines the connection line of the standard triple corresponding to the triple in the pre-constructed knowledge graph based on the first and second tags of the triple to be reviewed; and determines the review conclusion for the data of the review object based on the coordinates of the extreme points of the connection line of the standard triple corresponding to the triple. In essence, it realizes automated content review based on knowledge graph. Compared with manual review, it can significantly improve the efficiency of content review and reduce the labor and economic costs of content review.

[0051] The above technical solutions, by leveraging different types of knowledge graphs, can achieve functions such as ad filtering, prohibited content moderation, abusive language identification, and spam detection. Optionally, following S4, the solution may also include processing the data of the subject being reviewed based on the review conclusions.

[0052] Based on the above technical solution, optionally, before S1, it also includes: constructing a knowledge graph.

[0053] To facilitate understanding of the technical solutions provided in this disclosure, before introducing how to construct a knowledge graph, the standard triple will be explained in detail.

[0054] Standard triples are the basic units for knowledge storage and representation in knowledge graphs. The three elements that constitute a standard triple are the first standard label, the second standard label, and the relationship between the first and second standard labels.

[0055] There are various specific forms of standard triplets, and this disclosure does not limit them.

[0056] For example, if a standard triple is represented as "entity-relationship-entity", the first standard label and the second standard label refer to two entities respectively. For example, if a standard triple can be represented as "BB-parent-AA", where "BB" and "AA" refer to names, then in this standard triple, the first standard label is BB and the second standard label is AA.

[0057] If a standard triple is represented as "entity-attribute-attribute value", the first standard label refers to the entity, and the second standard label refers to the attribute value. For example, if a standard triple can be represented as "BB-education level-Bachelor's degree", then in this standard triple, the first standard label is BB, and the second standard label is Bachelor's degree.

[0058] The essence of building a knowledge graph is to populate the knowledge graph with standard triples in a reasonable way.

[0059] Since the knowledge graph constructed using the knowledge graph construction method provided in this disclosure can be used for content moderation, specifically for ad filtering, prohibited content moderation, abusive language identification, and spam detection, etc., and considering the analysis of usage scenarios, in practice, standard triples mainly include the following two situations.

[0060] Scenario 1: The first or second standard label in the standard triplet has an implicit meaning, and this implicit meaning relates to specific content (such as illegal advertising). For example, suppose a mask produced by XX Medical Device Factory claims in its advertisement that its virus filtration rate reaches 99.9%. However, actual testing shows that the virus filtration rate does not reach the advertised rate. Therefore, the advertisement is suspected of false claiming and is considered illegal daily necessities advertising. Accordingly, a standard triplet is set, with the first standard label being "XX Medical Device Factory" and the second standard label being "mask." If "XX Medical Device Factory" or "mask" appears alone in a piece of content, the content is determined not to be illegal advertising. If both "XX Medical Device Factory" and "mask" appear in a piece of content, the content is determined to have an implicit meaning, and this implicit meaning constitutes illegal daily necessities advertising. In this case, the entire standard triplet needs to be populated into the knowledge graph so that the implicit meaning is displayed.

[0061] Scenario 2: In a standard triplet, the combination of the first or second standard tag does not have an implicit meaning. Or, although the combination of the two tags does have an implicit meaning, that meaning does not involve a specific piece of content (such as belonging to illegal advertising). In this case, simply fill the knowledge graph with the tag that involves a specific piece of content (such as belonging to illegal advertising).

[0062] In the two scenarios above, scenario two focuses on filling in the tags, which is relatively simple. When building a knowledge graph, the focus is on collecting tags.

[0063] Regarding scenario one, how to construct a knowledge graph that can reflect the implicit meaning of standard triples is the key to improving the application value of knowledge graph technology in content moderation, and it is a problem currently facing knowledge graph technology.

[0064] Therefore, this disclosure focuses on providing a detailed description of the knowledge graph construction method related to Case 1. Figure 3 This is a flowchart illustrating the steps involved in constructing a knowledge graph in a content moderation method provided in this embodiment of the disclosure. Figure 3 The provided steps for constructing a knowledge graph are applicable to situations where the implicit meaning of standard triples needs to be displayed. The knowledge graph constructed using the steps provided in this disclosure can be used for ad filtering, prohibited content moderation, abusive language identification, and spam detection, among other applications.

[0065] See Figure 3 The steps for constructing the knowledge graph include:

[0066] S110. Construct a coordinate system.

[0067] In reality, everything is often multifaceted, or has multiple attributes. The purpose of establishing a coordinate system in this step is to use the coordinate system to reflect the multiple attributes of a thing from multiple perspectives, so as to help users fully understand the thing.

[0068] In this step, the coordinate system constructed can be a two-dimensional coordinate system or a three-dimensional coordinate system, and this disclosure does not limit it.

[0069] For example, if the constructed coordinate system is a two-dimensional coordinate system, it can be set to include a first coordinate axis and a second coordinate axis that intersect each other; the coordinate points on the first coordinate axis are obtained based on a first classification criterion, and the coordinate points on the second coordinate axis are obtained based on a second classification criterion; the first classification criterion and the second classification criterion are different. Furthermore, the first classification criterion is the evaluation criterion.

[0070] If the constructed coordinate system is a three-dimensional coordinate system, it can be set to include a first coordinate axis, a second coordinate axis, and a third coordinate axis that intersect each other. The coordinate points on the first coordinate axis are obtained based on a first classification criterion, the coordinate points on the second coordinate axis are obtained based on a second classification criterion, and the coordinate points on the third coordinate axis are obtained based on a third classification criterion. Any two of the first, second, and third classification criteria must be different. Furthermore, the first classification criterion is the evaluation criterion.

[0071] It should be noted that in the above technical solutions, regardless of whether the constructed coordinate system is two-dimensional or three-dimensional, the first classification criterion is the evaluation criterion. In practice, the evaluation criterion is determined based on the scenario in which the knowledge graph is used (or the purpose of the knowledge graph). For example, Figure 4 and Figure 5 This diagram illustrates two coordinate systems provided in embodiments of this disclosure. For example, if the constructed knowledge graph is used for reviewing illegal advertisements, the first coordinate axis (i.e....) Figure 4 and Figure 5 The evaluation criteria (i.e., the first classification criterion) used on the X-axis can be the type of illegal advertisement. In this case, the coordinate points on the first coordinate axis can be set as non-illegal, illegal medical service advertisement, illegal beauty service advertisement, illegal daily necessities advertisement, and illegal investment and financial management advertisement, etc.

[0072] Furthermore, in practice, multiple levels of classification criteria can be set for the same coordinate axis to create a more hierarchical relationship between coordinate points, facilitating the accurate filling of labels from the standard triples into the coordinate system. For example, in... Figure 4 and Figure 5 The second coordinate axis (i.e.) Figure 4 and Figure 5 The coordinates on the Y-axis (in the model) have two levels. Specifically, monosyllabic words, disyllabic words, and polysyllabic words are obtained based on the first-level classification criteria, constituting the first level. Monosyllabic words, disyllabic words, and polysyllabic words are further divided into people, things, and others. People, things, and others are obtained based on the second-level classification criteria, constituting the second level.

[0073] contrast Figure 4 and Figure 5 , Figure 5 It is a three-dimensional coordinate system, relative to Figure 4 It adds a third coordinate axis, and the third classification criterion used for this third coordinate axis is "related to hot topics". In practice, the more coordinate axes a coordinate system has, the more attributes of the tags it reflects.

[0074] S120. Obtain the standard triplet to be filled. The standard triplet includes a first standard label and a second standard label that have an association relationship.

[0075] The standard triplet to be filled mentioned in this step refers to the standard triplet that satisfies condition one in the previous text.

[0076] S130. Fill the coordinate system with the first and second standard labels.

[0077] There are various ways to implement this step, and this disclosure does not limit the specific implementation. For example, firstly, the positions of the first standard label and the second standard label in the coordinate system are determined; then, based on the positions of the first standard label and the second standard label in the coordinate system, the first standard label and the second standard label are filled into the coordinate system.

[0078] Because the coordinate points on different axes in the coordinate system constructed in S110 are obtained based on different classification criteria, while the coordinate points on the same axis are obtained based on the same classification criteria, determining the positions of the first and second standard labels in the coordinate system can be done by determining the genus-species relationship between the first and second standard labels and each coordinate point. Therefore, the essence of this step is to fill the coordinate system with the first and second standard labels based on their attributes.

[0079] Figure 6 This is a schematic diagram of another coordinate system provided in an embodiment of this disclosure. For example, if the first standard label in a standard triplet is "XX Medical Device Factory" and the second standard label is "Mask", then if "XX Medical Device Factory" and "Mask" are filled into... Figure 6 In the coordinate system, these two labels need to be analyzed. Analysis shows that "XX Medical Device Factory" alone is not in violation and should be aligned with the first coordinate axis (i.e., Figure 6 The coordinate point corresponding to "not in violation" on the X-axis should be the same as the coordinate point on the second axis. Additionally, "XX Medical Device Factory" is a polysyllabic word and should correspond to the coordinate point on the second axis. Figure 6 The coordinate point on the Y-axis corresponds to "multisyllabic words - things". Furthermore, since "XX Medical Device Factory" does not involve trending topics, it should correspond to the third coordinate axis (i.e., Figure 6 The coordinate point corresponding to the "non-hotspot" on the Z-axis (middle Z-axis). Based on this, it can be determined that "XX Medical Device Factory" is located in... Figure 6 The coordinates in the coordinate system are (not in violation of regulations, multi-syllable word - thing, non-hot topic). Subsequently, based on (not in violation of regulations, multi-syllable word - thing, non-hot topic), the primary standard label "XX Medical Device Factory" can be filled into... Figure 6 In the coordinate system.

[0080] Similarly, a standalone "mask" is not in violation of regulations; it should be considered in conjunction with the first coordinate axis (i.e., Figure 6 The coordinate point corresponding to "not in violation" on the X-axis should be the same as the coordinate point on the second axis. Additionally, "mask" is a disyllabic word and should correspond to the coordinate point on the second axis. Figure 6 The coordinate point on the Y-axis corresponds to "disyllabic word - thing". Furthermore, "mask" does not involve trending topics and should correspond to the third coordinate axis (i.e., Figure 6 The coordinate point corresponding to the "non-hotspot" on the Z-axis (middle Z-axis). Based on this, it can be determined that the "mask" is located in... Figure 6 The coordinates in the mid-coordinate system are (not against regulations, disyllabic word - thing, non-hot topic). Subsequently, based on (not against regulations, disyllabic word - thing, non-hot topic), the second standard label "mask" can be filled into... Figure 6 In the coordinate system.

[0081] S140. Draw a connecting line in the coordinate system corresponding to the standard triplet, so that the connecting line connects the first standard label and the second standard label, and the extreme point of the connecting line and the standard triplet as a whole correspond to the same coordinate point of the first coordinate axis.

[0082] There are multiple ways to implement this step, and this disclosure does not impose any limitations on it. For example, the specific implementation method of this step includes: First, determining the correspondence between the standard triplet as a whole and the coordinate points in the first coordinate axis. Second, based on the correspondence, determining the position of the extreme point corresponding to the standard triplet in the coordinate system. Finally, based on the position of the extreme point in the coordinate system, drawing a connecting line corresponding to the standard triplet in the coordinate system, so that the connecting line connects the first standard label and the second standard label, and the connecting line passes through the extreme point. Here, the standard triplet as a whole should be understood as the implicit meaning generated by the combination of the first standard label and the second standard label.

[0083] To facilitate understanding, specific examples will be provided below.

[0084] If the rule is that content containing only "XX Medical Device Factory" or only "mask" is not in violation, then the content is deemed not to be in violation. However, if content contains both "XX Medical Device Factory" and "mask," then the content is deemed to be an illegal medical service advertisement. See also... Figure 6 To address this situation, a standard triplet consisting of "XX Medical Device Factory" and "Mask" is determined to correspond to the coordinate point "Illegal Medical Service Advertisement" on the first coordinate axis. Based on this correspondence, the extreme point corresponding to this standard triplet is determined to lie on a plane (hereinafter referred to as the reference plane) that includes the straight line L and is parallel to the plane defined by the Z and Y axes. At this point, any point on the reference plane can be taken as the extreme point M of the connecting line. A connecting line corresponding to this standard triplet is drawn, connecting the labels "XX Medical Device Factory" and "Mask," and passing through the extreme point M.

[0085] See also Figure 6 Since the extreme points of the connecting lines and the overall standard triplet (i.e., the implicit meaning generated by the combination of the first and second standard labels in the standard triplet) correspond to the same coordinate point on the first coordinate axis, the essence of the above technical solution is to use the coordinates of the extreme points of the connecting lines to reflect the evaluation result of the overall standard triplet. Alternatively, it can be described as using the coordinates of the extreme points of the connecting lines to reflect the implicit meaning generated by the combination of the first and second standard labels in the standard triplet. This achieves the goal of reflecting not only the attributes of each label in the standard triplet, but also the attributes of the standard triplet as a whole in the knowledge graph.

[0086] exist Figure 6In the examples, the standard ternary pairs "YY Hospital" and "Medical Insurance Designated Hospital", "Hyaluronic Acid Injection" and "Corn", and "XX Medical Device Factory" and "Mask" are similar. Further details will not be provided here.

[0087] Based on the above technical solution, optionally, for a knowledge graph that needs to display multiple standard triples, the extreme points of the connecting lines of each standard triple can be set to lie on the same plane. For example, see [link to example]. Figure 6 It is possible to set the extreme points of each standard triplet connection line to be located on the first coordinate axis (i.e., Figure 6 The middle X-axis) and the second coordinate axis (i.e. Figure 5 On the plane defined by the Y-axis. This setup allows for a more intuitive display of the overall attributes of standard triples, facilitating user identification and comparison of attributes between different standard triples. It can enhance the application value of knowledge graph technology in content moderation.

[0088] Figure 7 A flowchart illustrating the steps involved in constructing a knowledge graph in another content moderation method provided in this disclosure. See also... Figure 7 The steps for constructing the knowledge graph include:

[0089] S210. Construct a coordinate system, which includes a first coordinate axis and a second coordinate axis that intersect each other; the coordinate points on the first coordinate axis are obtained based on the first classification standard, and the coordinate points on the second coordinate axis are obtained based on the second classification standard; the first classification standard and the second classification standard are different, and the first classification standard is the evaluation standard.

[0090] S220. Obtain sample corpus data.

[0091] Optionally, the sample corpus data includes at least one of text and images.

[0092] There are several ways to implement this step. For example, sample corpus data can be collected based on keywords related to the uses of knowledge graphs.

[0093] When acquiring sample corpus data, this disclosure does not restrict the source of the sample corpus data. For example, the sample corpus data may come from at least one source, such as newspapers, magazines, and the Internet.

[0094] For example, if the constructed knowledge graph is used for reviewing illegal advertisements, the scope of sample corpus data collection can be set to include third-party data such as social media. Optionally, machine learning can be used for data collection.

[0095] S230. Based on the sample corpus data, identify the standard triples to be filled. The standard triples include a first standard label and a second standard label that have an association relationship.

[0096] Optionally, information is extracted from the sample corpus data (such as text recognition or image recognition), and logical operations are performed based on the recognition results to finally obtain the standard triplet to be filled.

[0097] Alternatively, this step can be replaced by: cleaning the sample corpus data; and identifying the standard triples to be filled based on the cleaned sample corpus data.

[0098] The cleaning process has two main objectives: removing irrelevant information and deduplication. When using keywords related to the purpose of the knowledge graph to collect sample corpus data, it's inevitable that some irrelevant information will be crawled. Removing irrelevant information is the primary objective here. Since data collected from different sources may contain at least partial similarities, deduplication involves removing duplicate data from different sources.

[0099] S240. Fill the coordinate system with the first and second standard labels.

[0100] Optionally, see Figure 6 Based on the genus-species relationship between the first and second standard labels of the standard triplet and each coordinate point in the coordinate system, the first and second standard labels are filled into the coordinate system.

[0101] S250. Draw a connecting line in the coordinate system corresponding to the standard triplet, so that the connecting line connects the first standard label and the second standard label, and the extreme point of the connecting line and the standard triplet as a whole correspond to the same coordinate point of the first coordinate axis.

[0102] The essence of the above technical solution is to use the extreme points of the connecting lines to reflect the overall evaluation result of the standard triplet. Alternatively, it can be described as using the extreme points of the connecting lines to reflect the implicit meaning generated by the combination of the first and second standard labels in the standard triplet. This allows the knowledge graph to reflect not only the attributes of each label in the standard triplet, but also the attributes of the standard triplet as a whole. This can enhance the application value of knowledge graph technology in content moderation.

[0103] It should also be noted that the above technical solution focuses on explaining scenario one mentioned earlier. That is, the implicit meaning generated by combining two tags in a standard triple involves a specific content (such as illegal advertising), and the standard triple needs to be filled into the knowledge graph as a whole.

[0104] Regarding scenario two mentioned above, in practice, it includes the following two possibilities:

[0105] Possibility 1: At least one of the two labels in the standard triple belongs to a certain specific content. In this case, only the label related to a certain specific content (such as illegal advertisement, etc.) needs to be filled into the knowledge graph. For example, refer to Figure 6 , if in a certain standard triple, the first standard label is "BB" and the second standard label is "Qian Jinlai". Among them, "BB" is a person's name. Since "BB" is not illegal, it is not filled in the coordinate system of the illegal advertisement knowledge graph. And "Qian Jinlai" belongs to the illegal investment and financial management advertisement. It is determined that the coordinate of "Qian Jinlai" in the coordinate system of the illegal advertisement knowledge graph is (illegal investment and financial management advertisement, multi-syllable word - others, non-hot spot), and "Qian Jinlai" is filled into the illegal advertisement knowledge graph.

[0106] Among them, if both labels in the same standard triple belong to a certain specific content (such as illegal advertisement, etc.), after filling these two labels into the coordinate system. A connection line can be drawn between these two labels, or a connection line can not be drawn. The present disclosure does not limit this.

[0107] Possibility 2: Neither of the two labels in the standard triple belongs to a certain specific content (such as illegal advertisement, etc.). In this case, this standard triple does not need to be filled into the knowledge graph.

[0108] It should also be noted that in practice, it can be executed sequentially in the order of each step in Figure 3 , or the execution order of each step can be adjusted arbitrarily. The present disclosure does not limit this.

[0109] Optionally, based on the above technical solutions, it further includes quality assessment of the constructed knowledge graph to improve the accuracy of the knowledge graph. There are various methods for quality assessment. Exemplarily, it can be achieved by the method of random sampling detection.

[0110] The knowledge graph constructed by using the above technical solutions can be used as a rule engine for screening advertisement content, screening prohibited content, screening abusive content, screening watering content, etc., to improve the recall rate of the audit and make the content audit more intelligent.

[0111] Figure 8 This is the structural block diagram of a content audit device proposed by an embodiment of the present disclosure. Refer to Figure 8 , this content audit device includes:

[0112] An audit object acquisition module 310, configured to acquire audit object data;

[0113] A to-be-audited triple extraction module 320, configured to extract to-be-audited triples based on the audit object data, where the to-be-audited triples include to-be-audited first labels and to-be-audited second labels with an associated relationship;

[0114] The connector determination module 330 is used to determine, based on the first label and the second label of the triplet to be reviewed, a connector line in a pre-constructed knowledge graph corresponding to the triplet to be reviewed; the knowledge graph includes a coordinate system and multiple standard triplets filled in the coordinate system; each standard triplet includes a first standard label, a second standard label, and a connector line connecting the first standard label and the second standard label, and the coordinates of the extreme points of the connector line reflect the overall evaluation result of the standard triplet;

[0115] The audit module 340 is used to determine the audit conclusion for the audit object data based on the coordinates of the extreme points of the connection line of the standard triplet corresponding to the triplet to be audited.

[0116] Furthermore, the device also includes a knowledge graph construction module;

[0117] The knowledge graph construction module is used to construct the knowledge graph before obtaining the data of the audit object.

[0118] Furthermore, the knowledge graph construction module includes:

[0119] A coordinate system construction unit is used for the coordinate system including a first coordinate axis and a second coordinate axis that intersect each other; the coordinate points on the first coordinate axis are obtained based on a first classification standard, and the coordinate points on the second coordinate axis are obtained based on a second classification standard; the first classification standard and the second classification standard are different, and the first classification standard is an evaluation standard;

[0120] A standard triplet acquisition unit is used to acquire standard triplets to be filled, wherein the standard triplets include a first standard label and a second standard label that have an association relationship.

[0121] A label filling unit is used to fill the first standard label and the second standard label into the coordinate system;

[0122] A connecting line drawing unit is used to draw connecting lines corresponding to the standard triplet in the coordinate system, so that the connecting lines connect the first standard label and the second standard label, and the extreme points of the connecting lines and the standard triplet as a whole correspond to the same coordinate point of the first coordinate axis.

[0123] Furthermore, the coordinate system also includes a third coordinate axis, which intersects both the first and second coordinate axes; the coordinate points on the third coordinate axis are obtained based on a third classification standard, and any two of the first, second, and third classification standards are different.

[0124] Furthermore, the standard triplet acquisition unit includes a sample data acquisition subunit and a standard triplet identification subunit;

[0125] The sample data acquisition subunit is used to acquire sample corpus data;

[0126] The standard triplet identification subunit is used to identify the standard triplet to be filled based on the sample corpus data.

[0127] Furthermore, the sample corpus data includes at least one of text and images.

[0128] Furthermore, the standard triplet identification subunit is used to clean the sample corpus data;

[0129] Based on the cleaned sample corpus data, the standard triples to be filled are identified.

[0130] Furthermore, if the standard triple is represented as "entity-relationship-entity", the first standard label and the second standard label refer to two entities respectively;

[0131] If the standard triple is represented as “entity-attribute-attribute value”, the first standard label refers to the entity and the second standard label refers to the attribute value.

[0132] Furthermore, the label filling unit is used for:

[0133] Determine the positions of the first standard label and the second standard label in the coordinate system;

[0134] Based on the positions of the first standard label and the second standard label in the coordinate system, the first standard label and the second standard label are filled into the coordinate system.

[0135] Furthermore, the connector drawing unit is used for:

[0136] Determine the correspondence between the standard triplet as a whole and the coordinate points in the first coordinate axis;

[0137] Based on the correspondence, the position of the extreme point corresponding to the standard triplet in the coordinate system is determined;

[0138] Based on the position of the extreme point in the coordinate system, a connecting line corresponding to the standard triplet is drawn in the coordinate system so that the connecting line connects the first standard label and the second standard label, and the connecting line passes through the extreme point.

[0139] The apparatus disclosed in the above embodiments can implement the process flow of the methods disclosed in the above method embodiments and has the same or corresponding beneficial effects. To avoid repetition, it will not be described again here.

[0140] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this disclosure, such as... Figure 9 As shown, the electronic device may include smart terminals such as mobile phones, tablets, and computers. The electronic device includes:

[0141] One or more processors 301, Figure 9 Taking processor 301 as an example;

[0142] Memory 302;

[0143] The electronic device may also include an input device 303 and an output device 304.

[0144] The processor 301, memory 302, input device 303, and output device 304 in the electronic device can be connected via a bus or other means. Figure 9 Taking the example of a connection between China and Israel via a bus.

[0145] The memory 302, as a non-transitory computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the content moderation method of the application in this embodiment of the present disclosure. The processor 301 executes various functional applications and data processing of the server by running the software programs, instructions, and modules stored in the memory 302, thereby implementing the content moderation method of the above-described method embodiment.

[0146] The memory 302 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 302 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 302 may optionally include memory remotely located relative to the processor 301, and these remote memories can be connected to the terminal device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0147] Input device 303 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device. Output device 304 may include display devices such as a display screen.

[0148] This disclosure also provides a computer-readable storage medium storing a program or instructions that, when executed by a computer, perform a content moderation method, the method comprising:

[0149] Obtain the data of the audited entity;

[0150] Based on the data of the audit object, extract the triples to be audited, wherein the triples to be audited include a first tag to be audited and a second tag to be audited that have an association relationship;

[0151] Based on the first label and the second label of the triple to be reviewed, a connecting line is determined in the pre-constructed knowledge graph to correspond to the standard triple to be reviewed; the knowledge graph includes a coordinate system and multiple standard triples filled in the coordinate system; each standard triple includes a first standard label, a second standard label, and a connecting line connecting the first standard label and the second standard label, and the coordinates of the extreme points of the connecting line reflect the overall evaluation result of the standard triple;

[0152] Based on the coordinates of the extreme points of the connecting lines of the standard triples corresponding to the triples to be audited, the audit conclusion for the audit object data is determined.

[0153] Optionally, when executed by a computer processor, the computer-executable instructions can also be used to execute the technical solutions of the content moderation methods provided in any embodiment of this disclosure.

[0154] Based on the above description of the implementation methods, those skilled in the art can clearly understand that this disclosure can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of this disclosure.

[0155] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0156] The above are merely specific embodiments of this disclosure, enabling those skilled in the art to understand or implement this disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A content moderation method, characterized in that, include: Obtain the data of the audit target; Based on the data of the audit object, extract the triples to be audited, wherein the triples to be audited include a first tag to be audited and a second tag to be audited that have an association relationship; The first tag to be reviewed refers to an entity, and the second tag to be reviewed refers to another entity or attribute value; Based on the pending first label and pending second label of the triplet to be reviewed, a connecting line is determined in a pre-constructed knowledge graph to correspond to the standard triplet to be reviewed. The knowledge graph includes a coordinate system and multiple standard triplets filled in the coordinate system. The coordinate system includes a first coordinate axis and a second coordinate axis that intersect each other. The coordinate points on the first coordinate axis are obtained based on a first classification standard, and the coordinate points on the second coordinate axis are obtained based on a second classification standard. The first classification standard and the second classification standard are different, and the first classification standard is an evaluation standard. Each standard triplet includes a first standard label, a second standard label, and a connecting line connecting the first standard label and the second standard label. The coordinate value of the extreme point of the connecting line on the first coordinate axis reflects the overall evaluation result of the standard triplet. The overall meaning of the standard triplet is the implicit meaning generated by the combination of the first standard label and the second standard label in the standard triplet. Based on the coordinates of the extreme points of the connection lines of the standard triples corresponding to the triples to be audited, the audit conclusion for the audit object data is determined. The method for constructing the knowledge graph includes: Construct a coordinate system; Obtain the standard triplet to be filled, wherein the standard triplet includes a first standard label and a second standard label that have an association relationship; Fill the coordinate system with the first standard label and the second standard label; A connecting line corresponding to the standard triplet is drawn in the coordinate system so that the connecting line connects the first standard label and the second standard label, and the extreme point of the connecting line and the standard triplet as a whole correspond to the same coordinate point of the first coordinate axis; the coordinate system also includes a third coordinate axis, which intersects both the first and second coordinate axes; the coordinate points on the third coordinate axis are obtained based on a third classification standard, and the first, second, and third classification standards are all different.

2. The content review method according to claim 1, characterized in that, The process of obtaining the standard triplet to be filled includes: Obtain sample corpus data; Based on the sample corpus data, the standard triples to be filled are identified.

3. The content moderation method according to claim 2, characterized in that, The sample corpus data includes at least one of text and images.

4. The content review method according to claim 2, characterized in that, The process of identifying the standard triples to be filled based on the sample corpus data includes: The sample corpus data is cleaned; Based on the cleaned sample corpus data, the standard triples to be filled are identified.

5. The content moderation method according to claim 1, characterized in that, If the standard triple is represented as "entity-relationship-entity", the first standard label and the second standard label refer to two entities respectively; If the standard triple is represented as "entity-attribute-attribute value", the first standard label refers to the entity and the second standard label refers to the attribute value.

6. The content moderation method according to claim 1, characterized in that, The step of filling the first standard label and the second standard label into the coordinate system includes: Determine the positions of the first standard label and the second standard label in the coordinate system; Based on the positions of the first standard label and the second standard label in the coordinate system, the first standard label and the second standard label are filled into the coordinate system.

7. The content moderation method according to claim 1, characterized in that, The step of drawing the connecting line corresponding to the standard triplet in the coordinate system includes: Determine the correspondence between the standard triplet as a whole and the coordinate points in the first coordinate axis; Based on the correspondence, the position of the extreme point corresponding to the standard triplet in the coordinate system is determined; Based on the position of the extreme point in the coordinate system, a connecting line corresponding to the standard triplet is drawn in the coordinate system so that the connecting line connects the first standard label and the second standard label, and the connecting line passes through the extreme point.

8. A content moderation device, characterized in that, include: The audit object acquisition module is used to acquire audit object data; The module for extracting triples to be reviewed is used to extract triples to be reviewed based on the data of the review object. The triples to be reviewed include a first tag to be reviewed and a second tag to be reviewed that have an association relationship. The first tag to be reviewed refers to an entity, and the second tag to be reviewed refers to another entity or attribute value. A connector determination module is used to determine, based on the first label and second label of the triplet to be reviewed, a connector line in a pre-constructed knowledge graph corresponding to the triplet to be reviewed. The knowledge graph includes a coordinate system and multiple standard triplets filling the coordinate system. The coordinate system includes intersecting first and second coordinate axes. Coordinate points on the first coordinate axis are obtained based on a first classification standard, and coordinate points on the second axis are obtained based on a second classification standard. The first and second classification standards are different, with the first classification standard being an evaluation standard. Each standard triplet includes a first standard label, a second standard label, and a connector line connecting the first and second standard labels. The coordinate values ​​of the extreme points of the connector line on the first coordinate axis reflect the overall evaluation result of the standard triplet. The overall meaning of the standard triplet is the implicit meaning generated by combining the first and second standard labels within the standard triplet. The review module is used to determine the review conclusion for the data to be reviewed based on the coordinates of the extreme points of the connection line of the standard triplet corresponding to the triplet to be reviewed. A knowledge graph construction module is used to construct a coordinate system; obtain standard triples to be filled, wherein the standard triples include a first standard label and a second standard label with an association relationship; fill the first standard label and the second standard label into the coordinate system; draw connecting lines in the coordinate system corresponding to the standard triples, such that the connecting lines connect the first standard label and the second standard label, and the extreme points of the connecting lines and the standard triples as a whole correspond to the same coordinate point of the first coordinate axis; The coordinate system also includes a third coordinate axis, which intersects both the first and second coordinate axes; the coordinate points on the third coordinate axis are obtained based on a third classification standard, and the first, second, and third classification standards are all different.

9. An electronic device, characterized in that, include: Processor and memory; The processor executes the steps of the method as described in any one of claims 1 to 7 by invoking programs or instructions stored in the memory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program or instructions that cause a computer to perform the steps of the method as described in any one of claims 1 to 7.

Images