Focus on the methods, devices, equipment, and storage media for constructing graph models.

Abstract

This disclosure provides a method, apparatus, device, and storage medium for constructing an attention graph model, relating to the field of data processing technology, particularly big data and information flow technologies. The specific implementation is as follows: In response to a first object exhibiting attention behavior towards a second object within a current preset period, a first attention relationship is determined between the first and second objects, and this first attention relationship is added to the attention relationship set; in response to a third object exhibiting reading behavior towards a fourth object's resources within the current period, a second attention relationship is determined between the third and fourth objects, and this second attention relationship is added to the attention relationship set, wherein the third object had already paid attention to the fourth object before the current period; based on the attention relationship set, an attention graph model is constructed. New attention relationships are obtained using the first attention relationship, while existing effective attention relationships are activated using the second attention relationship, thereby broadening the coverage of the attention graph model.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, particularly to the fields of big data and information flow, and specifically to a method, apparatus, device, and storage medium for constructing an interest graph model. Background Technology

[0002] Attention graph models utilize the attention relationships between users and authors to construct a graph model, and then train it by traversing the graph to obtain graph embedding vectors for users and authors. Graph embedding vectors can be used in stages such as recall and ranking to improve attention metrics. Summary of the Invention

[0003] This disclosure provides a method, apparatus, device, and storage medium for constructing an attention graph model.

[0004] According to one aspect of this disclosure, a method for constructing an attention graph model is provided, comprising:

[0005] In response to the first object showing interest in the second object within the current preset period, it is determined that there is a first interest relationship between the first object and the second object, and the first interest relationship is added to the interest relationship set;

[0006] In response to a third object reading a resource of a fourth object during the current period, a second interest relationship is determined between the third object and the fourth object, and the second interest relationship is added to the interest relationship set, wherein the third object had already paid attention to the fourth object before the current period;

[0007] Construct an attention graph model based on the set of attention relationships.

[0008] According to another aspect of this disclosure, an apparatus for constructing an interest graph model is provided, comprising:

[0009] The first determining unit is configured to, in response to the first object exhibiting a attention behavior towards the second object within the current preset period, determine that there is a first attention relationship between the first object and the second object, and add the first attention relationship to the attention relationship set;

[0010] The second determining unit is used to determine that there is a second attention relationship between the third object and the fourth object in response to the third object's reading behavior on the resources of the fourth object in the current period, and to add the second attention relationship to the attention relationship set, wherein the third object has already paid attention to the fourth object before the current period;

[0011] The building unit is used to construct an attention graph model based on a set of attention relationships.

[0012] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0013] At least one processor; and

[0014] The memory is communicatively connected to the at least one processor; wherein,

[0015] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described in the present disclosure.

[0016] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform any of the methods according to embodiments of this disclosure.

[0017] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the methods according to embodiments of this disclosure.

[0018] According to the method, apparatus, device, and storage medium for constructing an attention graph model provided in this disclosure, in response to a first object exhibiting an attention behavior towards a second object within a current preset period, a first attention relationship is determined between the first object and the second object, and this first attention relationship is added to an attention relationship set. In response to a third object exhibiting a reading behavior towards a fourth object's resources within the current period, a second attention relationship is determined between the third object and the fourth object, and this second attention relationship is added to the attention relationship set, wherein the third object had previously paid attention to the fourth object before the current period. Based on the attention relationship set, an attention graph model is constructed. New attention relationships can be obtained using the first attention relationship, while existing effective attention relationships can be activated using the second attention relationship, thereby broadening the coverage of the attention graph model.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0020] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0021] Figure 1 A schematic diagram of the system structure for applying the interest graph model construction method of the embodiments of this disclosure;

[0022] Figure 2 This is a schematic diagram of a method for constructing an interest graph model according to an embodiment of this disclosure;

[0023] Figure 3This is a schematic diagram of an interest graph model corresponding to a first interest relationship provided according to an embodiment of this disclosure;

[0024] Figure 4 This is a schematic diagram of a method for constructing an interest graph model according to another embodiment of this disclosure;

[0025] Figure 5 A schematic diagram of an apparatus for constructing a concern graph model according to an embodiment of this disclosure;

[0026] Figure 6 This is a block diagram of an electronic device used to implement the interest graph model construction method of the embodiments of this disclosure. Detailed Implementation

[0027] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0028] The disclosed embodiments provide a method, apparatus, electronic device, and storage medium for constructing an attention graph model. Specifically, the method for constructing the attention graph model in this disclosure embodiment can be executed by an electronic device, which can be a terminal or a server, etc. The terminal can be a smartphone, tablet computer, laptop computer, smart voice interaction device, smart home appliance, wearable smart device, aircraft, smart vehicle terminal, etc. The terminal can also include a client, which can be an audio client, video client, browser client, instant messaging client, or mini-program, etc. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN), and big data and artificial intelligence platforms.

[0029] Attention graph models utilize the attention relationships between users and authors to construct a graph model, and then train it by traversing the graph to obtain graph embedding vectors for users and authors. Graph embedding vectors can be used in recall, ranking, and other stages to improve attention metrics.

[0030] In related technologies, attention graph models employ a daily incremental training method, meaning new attention relationships are acquired every day. However, this leads to existing attention relationships being gradually forgotten. Additionally, platforms may have low-attention users (users with few or even zero followers). The limited number of low-attention user samples results in insufficient training of the attention graph model, leading to poor accuracy in the generated graph embedding vectors for these users.

[0031] To address at least one of the aforementioned problems, embodiments of this disclosure provide a method, apparatus, device, and storage medium for constructing an attention graph model. The method involves: in response to a first object exhibiting an attention behavior towards a second object within a current preset period, determining that a first attention relationship exists between the first and second objects, and adding the first attention relationship to an attention relationship set; in response to a third object exhibiting a reading behavior towards a fourth object's resources within the current period, determining that a second attention relationship exists between the third and fourth objects, and adding the second attention relationship to the attention relationship set, wherein the third object had previously paid attention to the fourth object before the current period; and constructing an attention graph model based on the attention relationship set. The first attention relationship can be used to acquire new attention relationships, while the second attention relationship can be used to activate previously existing effective attention relationships, thereby broadening the coverage of the attention graph model.

[0032] The embodiments of this disclosure will be described in detail below with reference to the accompanying drawings.

[0033] Figure 1 This is a schematic diagram of the system structure for applying the interest graph model construction method of this disclosure embodiment. Please refer to... Figure 1 The system includes a terminal 110 and a server 120, etc.; the terminal 110 and the server 120 are connected via a network, such as a wired or wireless network.

[0034] The terminal 110 can be used to display a graphical user interface. This terminal interacts with the user through the graphical user interface, for example, by downloading and installing a client, running a mini-program, or accessing a website. In this embodiment, the user can view content provided by the platform through the terminal 110. The server 120, in response to a first object showing interest in a second object within a current preset period, determines a first interest relationship between the first and second objects and adds this relationship to the interest relationship set. Similarly, in response to a third object reading resources of a fourth object within the current period, the server 120 determines a second interest relationship between the third and fourth objects and adds this relationship to the interest relationship set, wherein the third object had previously shown interest in the fourth object before the current period. Based on the interest relationship set, an interest graph model is constructed.

[0035] It should be noted that the platform's applications can be applications installed on desktop computers, applications installed on mobile devices, or mini-programs embedded in applications.

[0036] It should be noted that the above application scenarios are shown only to facilitate understanding of the spirit and principles of this disclosure, and the embodiments of this disclosure are not limited in any way. On the contrary, the embodiments of this disclosure can be applied to any applicable scenario.

[0037] The following is a detailed description. It should be noted that the order of description of the following embodiments is not intended to limit the priority of the embodiments.

[0038] Figure 2 This is a schematic diagram of a method for constructing an interest graph model according to an embodiment of this disclosure; Figure 3 This is a schematic diagram of an interest graph model corresponding to a first interest relationship provided according to an embodiment of this disclosure. Please refer to... Figure 1 and Figure 2 This disclosure provides a method 200 for constructing an interest graph model, including the following steps S201 to S203.

[0039] Step S201: In response to the first object showing interest in the second object within the current preset period, determine that there is a first interest relationship between the first object and the second object, and add the first interest relationship to the interest relationship set.

[0040] Step S202: In response to the third object reading the resources of the fourth object in the current period, determine that there is a second attention relationship between the third object and the fourth object, and add the second attention relationship to the attention relationship set, wherein the third object had already paid attention to the fourth object before the current period.

[0041] Step S203: Construct an attention graph model based on the set of attention relationships.

[0042] It's understandable that Method 200 can be used on various platforms such as social media platforms, information platforms, and shopping platforms. The attention graph model can be a graph model generated based on the attention relationships between users on the platform.

[0043] like Figure 3 As shown, both the first object 301 and the second object 302 are users of the platform. Similarly, the third and fourth objects are also users of the platform. The preset period can be the construction and training period of the attention graph model, which can be set as needed, such as one day, one week, or one month. Taking a preset period of one day as an example, the platform data from each day can be used to construct the attention graph model. The current preset period can be the preset period corresponding to the attention graph model to be constructed, such as the attention graph model constructed on the current day.

[0044] Taking the current preset period as the day as an example, in step S201, each first follow relationship can be determined through each newly added follow behavior within the day. Follow behavior can be, for example, clicking the follow button.

[0045] For example, by iterating through all users on the platform, for each user, it can be determined whether that user (the first object) has engaged in any follow behavior towards other users (the second object) within the current preset period. If so, it indicates that a first follow relationship exists between the first object and the second object, and this first follow relationship can be included in the follow relationship set. It can be understood that at the beginning of each preset period, i.e., before step S201, the follow relationship set can be an empty set.

[0046] For example, new follower behaviors can be directly obtained through platform logs. Then, for each new follower relationship, the follower (first object) and the followed object (second object) can be identified, thus determining that there is a first follower relationship between the first object and the second object. It can be understood that the first object did not follow the second object before the current preset period.

[0047] Step S201 allows you to add all newly added attention relationships within the current preset period to the attention relationship set.

[0048] Step S202 can target users on the platform who already have a following relationship within the current preset period. For example, if a third object has followed a fourth object before the current preset period, and the third object reads the resources of the fourth object that it has followed in the current preset period, then it is determined that there is a second following relationship between the third object and the fourth object.

[0049] In this context, the fourth entity is the author of the resource, and the third entity's reading behavior towards the fourth entity's resource can be understood as the third entity reading, watching, or viewing the resource. Furthermore, each time a user reads a resource by an author they follow, they are activating the follow relationship between the two.

[0050] This step can be achieved by iterating through all users on the platform and determining for each user (the third object) whether they have read the resources of other users they are following (the fourth object they were following before the current preset period) within the current preset period. If so, it indicates that a second following relationship exists between the third object and the fourth object. Alternatively, step S202 can also obtain the reading behavior of each user from the platform logs and then determine the second following relationship from the reading behavior.

[0051] It can be understood that step S202 obtains all second-level attention relationships within the current preset period on the platform and adds them to the attention relationship set; that is, the second-level attention relationship is also an attention relationship in the attention relationship set. It can be understood that after executing steps S201 and S202, the attention relationships in the attention relationship set are either the first-level attention relationship or the second-level attention relationship.

[0052] Step S203 can construct an attention graph model based on the set of attention relationships.

[0053] In some embodiments, an attention graph model is constructed based on a set of attention relationships, including the following steps:

[0054] For the target interest relationship in the interest relationship set, obtain the objects being followed and the objects being followed that have the target interest relationship;

[0055] Based on the object of interest and the object of interest, a first target node is constructed to represent the object of interest and a second target node is constructed to represent the object of interest. Both the first target node and the second target node are nodes in the interest graph model.

[0056] Based on the target attention relationship, a target edge is constructed to connect the first target node and the second target node, and the target edge is an edge in the attention graph model.

[0057] It can be understood that a target attention relationship can be one of the attention relationships in a set of attention relationships; it can be a first attention relationship or a second attention relationship. Specifically, a target attention relationship can include the object of attention, the object being followed, and the pointing relationship between the object of attention and the object being followed.

[0058] Taking the target attention relationship as the first attention relationship as an example, the object of attention is the first object, the object being followed is the second object, and the pointing relationship between the object of attention and the object being followed is that the second object points to the first object.

[0059] Taking the target attention relationship as the second attention relationship as an example, the attention object is the third object, the object being followed is the fourth object, and the pointing relationship between the attention object and the object being followed is from the fourth object to the third object.

[0060] The attention graph model can be a directed graph model, which can include nodes and edges. Nodes can represent objects of interest or objects being followed, and edges can point from one node to another, thus representing the pointing relationship between objects of interest and objects being followed.

[0061] When constructing an attention graph model, for each attention relationship in the attention relationship set, taking the target attention relationship as an example, we can first obtain the attention object and the attention object involved in the target attention relationship. The attention object and the attention object can be constructed as nodes in the attention graph model. For example, the attention object can be constructed as the first target node, and the second attention object can be constructed as the second target node.

[0062] Then, based on the pointing relationship between the object of interest and the object being interested, a target edge connecting the first target node and the second target node can be constructed.

[0063] like Figure 3 Taking the construction of a concern graph model based on the first concern relationship as an example, the concern object is the first object, which can be constructed as the first target node 301, and the concerned object is the second object, which can be constructed as the second target node 302. The pointing relationship between the concern object and the concerned object is that the second object points to the first object, that is, it is constructed as a target edge 303 with a pointing.

[0064] In this embodiment, the pointing relationship is represented by an arrow pointing from the object being watched to the object being watched. In other embodiments, the pointing relationship can also be represented by an arrow pointing from the object being watched to the object being watched.

[0065] It is understandable that by constructing the nodes and edges corresponding to all the attention relationships in the attention relationship set, the attention graph model within the current preset period can be quickly and conveniently constructed.

[0066] After obtaining the attention graph model, it can be trained through methods such as walk training to obtain graph embedding vectors. These graph embedding vectors can be used in the recall or ranking stages to improve attention metrics.

[0067] After steps S201 and S202, the set of attention relationships includes both first and second attention relationships. That is, the set of attention relationships used to construct the attention graph model includes not only the first attention relationships corresponding to newly added attention behaviors within the current preset period, but also the second attention relationships between users with existing attention relationships. It is understood that in related technologies, existing attention relationships are ignored when constructing the attention graph model for the current preset period. This embodiment, based on the reading behavior between users with existing attention relationships, can transform the reading behavior between users with existing attention relationships into second attention relationships for constructing the attention graph model. This activates previously effective attention relationships, improves the coverage of the attention graph model, and avoids forgetting previous attention relationships. This is beneficial to the accuracy of the graph embedding vector and the accuracy of the results in subsequent recall and ranking stages.

[0068] In some embodiments, step S202, in response to a third object performing a reading action on the resources of a fourth object during the current period, determining that there is a second interest relationship between the third object and the fourth object, may include:

[0069] In response to a third object engaging in reading behavior on a fourth object's resources within the current period, and the duration of the reading behavior being greater than or equal to a first preset duration, a second interest relationship is determined between the third object and the fourth object.

[0070] When determining the second interest relationship, in addition to considering whether the third object has read the resources of the fourth object, the duration of the reading behavior can also be considered, that is, the duration of the reading behavior.

[0071] The first preset duration can be 3 seconds, 5 seconds, etc. If there is reading behavior and the duration of the reading behavior is greater than or equal to the first preset duration, it is determined that there is a second attention relationship between the third object and the fourth object.

[0072] This embodiment can activate previously effective follow relationships by filtering the second follow relationships between followed users whose reading behavior duration is greater than or equal to a first preset duration, thereby improving the accuracy of the second follow relationships.

[0073] Figure 4 This is a schematic diagram of a method for constructing an interest graph model according to another embodiment of this disclosure. Please refer to... Figure 4 This disclosure also provides a method 400 for constructing an interest graph model, including the following steps S401 to S405.

[0074] Step S401: In response to the first object showing interest in the second object within the current preset period, determine that there is a first interest relationship between the first object and the second object, and add the first interest relationship to the interest relationship set.

[0075] Step S402: In response to the third object reading the resources of the fourth object in the current period, determine that there is a second attention relationship between the third object and the fourth object, and add the second attention relationship to the attention relationship set, wherein the third object had already paid attention to the fourth object before the current period.

[0076] Step S403: In response to the fact that the number of objects followed by the fifth object is less than the preset number of objects followed and the attention rate of the fifth object to the first distribution resource distributed to the fifth object is higher than the preset attention rate, it is determined that there is a third attention relationship between the fifth object and the sixth object, and the third attention relationship is added to the attention relationship set; wherein, the first distribution resource is the resource of the sixth object, and the attention rate is used to characterize the probability that the fifth object follows the sixth object through the first distribution resource.

[0077] Step S404: In response to the fact that the number of objects followed by the seventh object is less than the preset number of objects followed and the seventh object interacts with the second distribution resource distributed to the seventh object, it is determined that there is a fourth follow relationship between the seventh object and the eighth object, and the fourth follow relationship is added to the follow relationship set; wherein, the second distribution resource is the resource of the eighth object.

[0078] Step S405: Construct an attention graph model based on the set of attention relationships.

[0079] The implementation of steps S401 and S402 is the same as steps S201 and S202 in the above embodiments, and can be referred to the above embodiments for details, which will not be repeated here.

[0080] The fifth and sixth objects are also users within the platform. This means the platform can distribute resources to each user, essentially recommending resources that the user might be interested in.

[0081] The number of objects followed by the fifth user can be the number of objects that the fifth user follows, that is, the number of objects followed in the user's following list. A preset number of objects followed is used as a standard for judging low-following users; for example, it could be 15, 20, etc. It can be understood that if a user's number of objects followed is lower than the preset number, then that user is considered a low-following user.

[0082] We can first filter out users with low engagement on the platform. For each user with low engagement, taking the fifth user as an example, the multiple resources distributed to them by the platform can be multiple first-level distribution resources. Taking one of the first-level distribution resources as an example, if the fifth user's engagement rate for the first-level distribution resource is higher than the preset engagement rate, then it is determined that the fifth user and the author of that first-level distribution resource (the sixth user) have a third engagement relationship.

[0083] It's understandable that a user's attention rate to a resource represents the probability that the user follows the resource's author through that resource. When distributing resources to users, for each resource, the platform (e.g., in logs) records its attention score for that user, i.e., the attention rate. Preset attention rates can be 0.01, 0.015, etc.

[0084] For each low-attention user, such as the fifth object, the attention rate of the first distribution resource distributed to them can be obtained. If the attention rate is higher than the preset attention rate, the first distribution resource can be determined as a high-attention distribution sample. At the same time, it is determined that there is a third attention relationship between the fifth object and the sixth object, so the third attention relationship between the fifth object and the sixth object can be added to the attention relationship set, that is, the third attention relationship is also a attention relationship in the attention relationship set.

[0085] In addition, when the preset attention rate is 0.01, the attention rate of the distribution sample (first distribution resource) that determines the third attention relationship is 33 times that of the average attention rate, which is significantly higher than the average attention rate. Therefore, the confidence of the distribution sample can be guaranteed.

[0086] Taking the third follow relationship in the follow relationship set as an example, the follower is the fifth object, the followed object is the sixth object, and the pointer relationship between the follower and the followed object is from the sixth object to the fifth object. The third follow relationship can increase the number of follow relationships in the follow relationship set.

[0087] It is understandable that for low-follower users, the number of authors they follow is relatively small. If only the methods in the relevant technologies are used to construct the attention graph model, then the number of edges in the attention graph model corresponding to the low-follower user will be small. During walk training, the number of times the user is walked and traversed will be very small, that is, the training samples are few, the training results are insufficient, and the accuracy of the obtained graph embedding vector is low.

[0088] In this embodiment, high-attention-rate distribution samples are introduced for low-attention users. Specifically, the first candidate resource with high attention rate is determined based on its attention rate to the first candidate resource distributed to the user. This increases the third-party attention relationship between low-attention users and resource authors, thereby increasing the number of distribution samples for low-attention users. Furthermore, the increased distribution samples have a certain level of attention confidence. Using these confident distribution samples increases the number of edges for the low-attention user in the attention graph model, leading to more thorough training and improved graph embedding vector accuracy. Simultaneously, the coverage of the attention graph model is enhanced.

[0089] In some embodiments, method 400 may further include: inputting the fifth object and the first distribution resource into a ranking model to obtain the attention rate of the first distribution resource.

[0090] It is understandable that the ranking model can be a recommendation model used to recommend resources to users. By inputting the fifth object and multiple first-distribution resources into the ranking model, the attention rate of each first-distribution resource can be obtained, that is, the attention score (fine-ranked attention score, etc.) of the resource author followed by the fifth object through the first-distribution resources. The ranking model can obtain a high-accuracy attention rate.

[0091] Continue to refer to Figure 4 In step S404, the seventh and eighth objects can also be users within the platform. The number of objects followed by the seventh object is also less than the preset number of objects followed, meaning that the seventh object is also a low-follower user within the platform.

[0092] We can first filter out users with low engagement on the platform. For each user with low engagement, taking the seventh user as an example, the multiple resources distributed to them by the platform can be multiple second-level distribution resources. Taking one of the second-level distribution resources as an example, if the seventh user interacts with the second-level distribution resource, then it is determined that the seventh user and the author of that second-level distribution resource (the eighth user) have a fourth engagement relationship.

[0093] Interactive behaviors include at least one of the following: commenting, liking, sharing, saving, and clicking on the author's homepage.

[0094] It can be understood that the interaction between the seventh object and the second distributed resource can be as follows: the seventh object comments on the second distributed resource, or the seventh object likes the second distributed resource, or the seventh object shares the second distributed resource, or the seventh object saves the second distributed resource, or the seventh object clicks on the author's homepage through the second distributed resource, etc.

[0095] Furthermore, the attention rates of distribution samples (secondary distribution resources) determined through interactive behavior were all higher than the average attention rate. For example, the attention rate of distribution samples determined by commenting behavior was 22 times that of the average attention rate; the attention rate of distribution samples determined by collecting behavior was 86 times that of the average attention rate; the attention rate of distribution samples determined by sharing behavior was 38 times that of the average attention rate; the attention rate of distribution samples determined by liking behavior was 59 times that of the average attention rate; and the attention rate of distribution samples determined by author homepage click behavior was 15 times that of the average attention rate.

[0096] If the seventh object exhibits one of the aforementioned interactive behaviors, it is determined that there is a fourth attention relationship between the seventh object and the eighth object. This helps to further increase the number of distribution samples from low-attention users. Furthermore, since the attention rate of the distribution samples determined by the interactive behaviors is higher than the average attention rate, the confidence level of the distribution samples is also guaranteed.

[0097] If there is a fourth concern relationship between the seventh object and the eighth object, the fourth concern relationship can be added to the concern relationship set, that is, the fourth concern relationship is also a concern relationship in the concern relationship set.

[0098] Taking the fourth follow relationship in the follow relationship set as an example, the follower is the seventh object, the followed object is the eighth object, and the pointer relationship between the follower and the followed object is from the eighth object to the seventh object. The fourth follow relationship can increase the number of follow relationships in the follow relationship set.

[0099] It is understandable that for low-follower users, the number of authors they follow is relatively small. If only the methods in the relevant technologies are used to construct the attention graph model, then the number of edges in the attention graph model corresponding to the low-follower user will be small. During walk training, the number of times the user is walked and traversed will be very small, that is, the training samples are few, the training results are insufficient, and the accuracy of the obtained graph embedding vector is low.

[0100] In this embodiment, by introducing distribution samples with interactive behavior to low-attention users—that is, filtering second-candidate resources based on whether they generate interactive behavior—a fourth attention relationship between low-attention users and resource authors is increased. This increases the number of distribution samples for low-attention users, and the increased distribution samples have a certain attention confidence. Using distribution samples with attention confidence increases the number of edges for low-attention users in the attention graph model, resulting in more thorough training and improved accuracy of graph embedding vectors. Simultaneously, it enhances the coverage of the attention graph model.

[0101] Step S405 can construct an attention graph model using the attention relationships in the attention relationship set. The attention relationships can be the first attention relationship, the second attention relationship, the third attention relationship, or the fourth attention relationship. The method for constructing the attention graph model is the same as or similar to step S203. For details, please refer to the embodiment of step S203, which will not be repeated here.

[0102] It is understood that in other embodiments, method 400 may also include only steps S401, S402, S403 and S405, or method 400 may also include only steps S401, S402, S404 and S405.

[0103] Based on the above embodiments, before constructing the attention graph model based on the attention relationship set in step S203, method 200 further includes: removing duplicate attention relationships from the attention relationship set.

[0104] It is understandable that the attention relationships in the attention relationship set may be the first attention relationship, the second attention relationship, the third attention relationship, or the fourth attention relationship, and there may be duplicate attention relationships among the first attention relationship, the second attention relationship, the third attention relationship, or the fourth attention relationship.

[0105] A duplicate attention relationship is one in which the attention object, the attention object, and the pointing relationship between the attention object and the attention object are all consistent.

[0106] This embodiment can remove duplicate attention relationships from the attention relationship set, simplify the construction of the attention graph model, and avoid duplicate nodes and edges in the attention graph model.

[0107] In one specific embodiment, a method for constructing an attention graph model is provided, which improves the coverage of the attention graph model. The method may include the following three points.

[0108] 1. Effective distribution of content, improving coverage. This improvement addresses the issue of forgotten existing relationships of interest in related technologies.

[0109] Effective follower distribution is defined as: a user reading (watching) resources from an author they already follow for 5 seconds or more. Since the user has already followed the author, follower distribution does not generate a follow-up action. We can think of follower distribution this way: each time a user reads resources from a followed author, they are activating that follow relationship. Therefore, by adding this action, previously effective follow relationships (secondary follow relationships) can be extracted and preserved.

[0110] 2. For users with low engagement, introduce distribution samples with high engagement rates (engagement q). This is an improvement addressing the insufficient training of low-engagement users in related technologies.

[0111] The "attention q" refers to a user's precise attention score for a resource, representing the probability that a user will follow the author of that resource after seeing it. This value is calculated by the ranking model. We consider attention q values ​​greater than or equal to 0.01 to be high attention q. This portion of the sample has an attention rate 33 times higher than the average attention rate, significantly higher than the average. Selecting this portion of the sample can expand the training sample of low-attention users while also ensuring a certain level of confidence in the attention level.

[0112] 3. For low-attention users, introduce interactive distribution samples. This is an improvement to address the insufficient training of low-attention users in related technologies.

[0113] Interactions primarily refer to comments, favorites, shares, likes, and clicks on the author's homepage. The attention rates of these interaction distribution samples are significantly higher than the average attention rate. Specifically, the comment distribution sample has a attention rate 22 times higher than the average; the favorite distribution sample has a attention rate 86 times higher; the share distribution sample has a attention rate 38 times higher; the like distribution sample has a attention rate 59 times higher; and the author's homepage click sample has a attention rate 15 times higher. Selecting these samples expands the training sample of low-attendance users while also ensuring a certain level of confidence in the attention level.

[0114] The attention graph model construction method provided in this embodiment improves coverage, activates previously effective attention relationships, expands the sample of low-attention users, and also ensures the attention confidence of these samples.

[0115] Figure 5A schematic diagram of the apparatus for constructing a focus graph model according to embodiments of this disclosure; please refer to... Figure 5 This disclosure provides an apparatus 500 for constructing an interest graph model, which includes the following units.

[0116] The first determining unit 501 is used to determine that there is a first attention relationship between the first object and the second object in response to the first object showing attention to the second object within the current preset period, and to add the first attention relationship to the attention relationship set.

[0117] The second determining unit 502 is used to determine that there is a second attention relationship between the third object and the fourth object in response to the third object's reading behavior on the resources of the fourth object in the current period, and to add the second attention relationship to the attention relationship set, wherein the third object has already paid attention to the fourth object before the current period;

[0118] Building unit 503 is used to construct an attention graph model based on a set of attention relationships.

[0119] In some embodiments, the second determining unit 502 is further configured to:

[0120] In response to a third object engaging in reading behavior on a fourth object's resources within the current period, and the duration of the reading behavior being greater than or equal to a first preset duration, a second interest relationship is determined between the third object and the fourth object.

[0121] In some embodiments, the device 500 further includes:

[0122] The third determining unit is used to determine that there is a third attention relationship between the fifth object and the sixth object in response to the fact that the number of objects of interest of the fifth object is less than the preset number of objects of interest and the attention rate of the fifth object to the first distribution resource distributed to the fifth object is higher than the preset attention rate, and to add the third attention relationship to the attention relationship set.

[0123] The first distribution resource is the resource of the sixth object, and the attention rate is used to characterize the probability that the fifth object pays attention to the sixth object through the first distribution resource.

[0124] In some embodiments, the device 500 further includes:

[0125] The distribution unit is used to input the fifth object and the first distribution resource into the ranking model to obtain the attention rate of the first distribution resource.

[0126] In some embodiments, the device 500 further includes:

[0127] The fourth determining unit is used to determine that there is a fourth attention relationship between the seventh object and the eighth object in response to the seventh object having fewer than the preset number of attention objects and the seventh object interacting with the second distribution resource distributed to the seventh object, and to add the fourth attention relationship to the attention relationship set.

[0128] The second distribution resource is the resource of the eighth object.

[0129] In some embodiments, the interactive behavior includes at least one of the following: commenting, liking, sharing, saving, and clicking on the author's homepage.

[0130] In some embodiments, the device 500 further includes:

[0131] The removal unit is used to remove duplicate concerns from the concern set.

[0132] In some embodiments, the building unit 503 is further configured to:

[0133] For the target interest relationship in the interest relationship set, obtain the objects being followed and the objects being followed that have the target interest relationship;

[0134] Based on the object of interest and the object of interest, a first target node is constructed to represent the object of interest and a second target node is constructed to represent the object of interest. Both the first target node and the second target node are nodes in the interest graph model.

[0135] Based on the target attention relationship, a target edge is constructed to connect the first target node and the second target node, and the target edge is an edge in the attention graph model.

[0136] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0137] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0138] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0139] This disclosure provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform any of the methods described above.

[0140] This disclosure provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the above-described method.

[0141] This disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0142] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0143] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0144] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0145] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the method for constructing an interest graph model. For example, in some embodiments, the method for constructing an interest graph model can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the method for constructing an interest graph model described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the method for constructing an interest graph model by any other suitable means (e.g., by means of firmware).

[0146] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0147] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0148] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0149] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0150] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0151] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0152] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0153] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for constructing an attention graph model, comprising: In response to the first object showing interest in the second object within the current preset period, it is determined that there is a first interest relationship between the first object and the second object, and the first interest relationship is added to the interest relationship set; In response to a third object reading a resource of a fourth object within the current preset period, it is determined that there is a second attention relationship between the third object and the fourth object, and the second attention relationship is added to the attention relationship set, wherein the third object had already paid attention to the fourth object before the current preset period; Based on the set of attention relationships, an attention graph model is constructed; Also includes: Input the fifth object and the first distribution resource into the ranking model to obtain the attention rate of the first distribution resource; In response to the fact that the number of objects followed by the fifth object is less than the preset number of objects followed and the attention rate of the fifth object to the first distribution resource distributed to the fifth object is higher than the preset attention rate, it is determined that there is a third attention relationship between the fifth object and the sixth object, and the third attention relationship is added to the attention relationship set; Wherein, the first distribution resource is the resource of the sixth object, and the attention rate is used to characterize the probability that the fifth object pays attention to the sixth object through the first distribution resource.

2. The method according to claim 1, wherein, In response to a third object reading resources of a fourth object within the current preset period, determining that a second interest relationship exists between the third object and the fourth object includes: In response to the third object performing the reading behavior on the resources of the fourth object within the current preset period and the duration of the reading behavior being greater than or equal to a first preset duration, it is determined that the third object and the fourth object have a second attention relationship.

3. The method according to claim 1, further comprising: In response to the fact that the number of objects followed by the seventh object is less than the preset number of objects followed and the seventh object interacts with the second distribution resource distributed to the seventh object, it is determined that there is a fourth follow relationship between the seventh object and the eighth object, and the fourth follow relationship is added to the follow relationship set; The second distribution resource is the resource of the eighth object.

4. The method according to claim 3, wherein, The interactive behaviors include at least one of the following: commenting, liking, sharing, saving, and clicking on the author's homepage.

5. The method according to any one of claims 1-4, wherein, Before constructing the attention graph model based on the aforementioned set of attention relationships, the method further includes: Remove duplicate attention relationships from the set of attention relationships.

6. The method according to any one of claims 1-4, wherein, Based on the aforementioned set of attention relationships, an attention graph model is constructed, including: For the target attention relationship in the set of attention relationships, obtain the attention objects and the objects being followed that have the target attention relationship; Based on the object of interest and the object of interest, a first target node for representing the object of interest and a second target node for representing the object of interest are constructed, wherein the first target node and the second target node are nodes in the attention graph model; Based on the target attention relationship, a target edge is constructed to connect the first target node and the second target node, and the target edge is an edge in the attention graph model.

7. An apparatus for constructing an attention graph model, comprising: The first determining unit is configured to, in response to the first object exhibiting a attention behavior towards the second object within the current preset period, determine that there is a first attention relationship between the first object and the second object, and add the first attention relationship to the attention relationship set; The second determining unit is configured to, in response to a third object performing a reading action on the resources of a fourth object within the current preset period, determine that there is a second attention relationship between the third object and the fourth object, and add the second attention relationship to the attention relationship set, wherein the third object had already paid attention to the fourth object before the current preset period; A construction unit is used to construct an attention graph model based on the set of attention relationships; Also includes: The distribution unit is used to input the fifth object and the first distribution resource into the ranking model to obtain the attention rate of the first distribution resource. The third determining unit is configured to determine that there is a third attention relationship between the fifth object and the sixth object in response to the fact that the number of objects followed by the fifth object is less than the preset number of objects followed and the attention rate of the fifth object to the first distribution resource distributed to the fifth object is higher than the preset attention rate, and to add the third attention relationship to the attention relationship set. Wherein, the first distribution resource is the resource of the sixth object, and the attention rate is used to characterize the probability that the fifth object pays attention to the sixth object through the first distribution resource.

8. The apparatus according to claim 7, wherein, The second determining unit is further configured to: In response to the third object performing the reading behavior on the resources of the fourth object within the current preset period and the duration of the reading behavior being greater than or equal to a first preset duration, it is determined that the third object and the fourth object have a second attention relationship.

9. The apparatus according to claim 7, further comprising: The fourth determining unit is configured to, in response to the fact that the number of objects followed by the seventh object is less than the preset number of objects followed and the seventh object interacts with the second distribution resource distributed to the seventh object, determine that there is a fourth following relationship between the seventh object and the eighth object, and add the fourth following relationship to the following relationship set. The second distribution resource is the resource of the eighth object.

10. The apparatus according to claim 9, wherein, The interactive behaviors include at least one of the following: commenting, liking, sharing, saving, and clicking on the author's homepage.

11. The apparatus according to any one of claims 7-10, further comprising: The removal unit is used to remove duplicate attention relationships from the attention relationship set.

12. The apparatus according to any one of claims 7-10, wherein, The building unit is also used for: For the target attention relationship in the set of attention relationships, obtain the attention objects and the objects being followed that have the target attention relationship; Based on the object of interest and the object of interest, a first target node for representing the object of interest and a second target node for representing the object of interest are constructed, wherein the first target node and the second target node are nodes in the attention graph model; Based on the target attention relationship, a target edge is constructed to connect the first target node and the second target node, and the target edge is an edge in the attention graph model.

13. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

15. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images