Object processing method and apparatus, resource processing method and apparatus, product, and device

By dividing and merging relationship communities based on social connections and combining consumption data to predict the interests and preferences of different user groups, the problem of inaccurate user group interest preferences has been solved, and efficient and accurate content resource recommendations have been achieved.

WO2026237588A1PCT designated stage Publication Date: 2026-11-19TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2026/087909
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-15
Filing Date
2026-04-01
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

In large-scale social networks, content resources that are frequently used within a user's social circle may not reflect their characteristics across the entire user base, leading to inaccurate acquisition of user interests and preferences.

Method used

By acquiring the social connections of objects through computer devices, dividing and merging relationship communities, expanding object circles, and predicting circle preferences based on related consumption data, the interest and preference of object circles for content resources can be obtained.

Benefits of technology

It improves the efficiency and accuracy of target audience segmentation, enhances the accuracy of content resource recommendations, and improves user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2026087909_19112026_PF_FP_ABST
    Figure CN2026087909_19112026_PF_FP_ABST
Patent Text Reader

Abstract

An object processing method, performed by a computer device, the method comprising: selecting a first object from an object set, and acquiring a second object having a social association relationship with the first object (S401); acquiring a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs, the first relationship community being used for describing the first object and another object having a social association relationship with the first object, the second object belonging to the other object described in the first relationship community, and the second relationship community being used for describing the second object and another object having a social association relationship with the second object (S402); selecting a homogeneous relationship community of the first relationship community from the at least one second relationship community, wherein the first object belongs to another object described by the homogeneous relationship community (S403); and merging the first relationship community with the homogeneous relationship community to obtain an object cohort diffused on the basis of the first object (S404).
Need to check novelty before this filing date? Find Prior Art

Description

Object processing methods, resource processing methods, devices, products and equipment

[0001] Related applications

[0002] This application claims priority to Chinese Patent Application No. 202510631714.0, filed on May 15, 2025, entitled "Method, Apparatus, Product and Equipment for Object Processing", and also claims priority to Chinese Patent Application No. 202510632474.6, filed on May 15, 2025, entitled "Resource Processing Method, Apparatus, Product and Equipment", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the technical field of resource processing, and more particularly to an object processing method, resource processing method, apparatus, product, and equipment. Background Technology

[0004] User circles refer to user groups with certain commonalities formed from a global user base based on users' social attributes (such as social connections, geographic location, and interests). Building accurate social circles is of great value for optimizing social products within social applications and improving user experience. However, in the complex scenarios of large-scale user social networks, the effective mining and efficient computation of social circles remains a technical challenge for both academia and industry.

[0005] Users within a user community can interact with content resources (such as videos, articles, and products). Therefore, the content resources most frequently interacted with within a user community can be considered as content resources reflecting the community's interests and preferences. However, content resources frequently interacted with within a user community are also likely frequently interacted with globally. Consequently, the degree of preference for these content resources within the user community is less pronounced compared to the global user community. This means that the frequently interacted content resources within a user community may not exhibit the specific characteristics (i.e., preferences) of that user community, leading to an inaccurate assessment of the content resources reflecting the user community's interests and preferences. Summary of the Invention

[0006] This application provides an object processing method, a resource processing method, an object processing apparatus, a product, and an equipment.

[0007] This application provides an object processing method, executed by a computer device, the method comprising:

[0008] Select a first object from the object set, and obtain a second object that has a social relationship with the first object;

[0009] Obtain the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object;

[0010] Select a homogeneous relation community from at least one second relation community; the first object belongs to another object described by the homogeneous relation community.

[0011] By merging the first relationship community with the homogeneous relationship community, we obtain the object circle that spreads based on the first object.

[0012] This application provides an object processing apparatus, which includes:

[0013] The object selection module is used to select a first object from the object set and obtain a second object that has a social relationship with the first object;

[0014] The community acquisition module is used to acquire the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object;

[0015] The community selection module is used to select a homogeneous relation community of the first relation community from at least one second relation community; the first object belongs to other objects described by the homogeneous relation community.

[0016] The merging module is used to merge the first relationship community with the homogeneous relationship community to obtain the object circle based on the first object.

[0017] This application provides a resource processing method, executed by a computer device, the method comprising:

[0018] Obtain the object layer, which includes multiple objects obtained by aggregating objects in the global object set into layers;

[0019] The first content resource is selected from the resource set, which contains multiple content resources. The first content resource is the content resource that the target audience has historically consumed.

[0020] Obtain the associated consumption data of the first content resource. This associated consumption data includes first consumption data generated by objects within the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource; and

[0021] Based on the associated consumption data of the first content resource, we perform circle preference prediction processing on the first content resource to obtain the interest and preference of the target circle for the first content resource.

[0022] This application provides a resource processing apparatus, which includes:

[0023] The first acquisition module is used to acquire object layers, which include multiple objects obtained by aggregating objects in the global object set.

[0024] The selection module is used to select the first content resource from the resource set. The resource set contains multiple content resources, and the first content resource is the content resource that the target audience has historically consumed.

[0025] The second acquisition module is used to acquire the associated consumption data of the first content resource. The associated consumption data includes first consumption data generated by objects in the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource.

[0026] The prediction module is used to perform circle preference prediction processing on the first content resource based on the associated consumption data of the first content resource, so as to obtain the interest and preference of the target circle for the first content resource.

[0027] This application provides a computer device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the method of this application.

[0028] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the method described in the above-mentioned aspect.

[0029] This application provides a computer program product comprising a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the methods provided in the various alternative embodiments described above.

[0030] This application can select a first object from an object set and obtain a second object that has a social relationship with the first object; it can also obtain a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have a social relationship with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have a social relationship with the second object; this application can select a homogeneous relationship community of the first relationship community from the at least one second relationship community; the first object belongs to other objects described by the homogeneous relationship community; thus, the first relationship community and the homogeneous relationship community can be merged to obtain an object circle based on the first object. Therefore, the method proposed in this application can merge the relational communities of the first object (e.g., the first relational community) and the relational communities of the second object (e.g., selected homogeneous relational communities) through the social connections between objects (e.g., the first object and the second object), to efficiently obtain the object circle based on the first object, thus improving the efficiency of obtaining the object circle. Furthermore, the multiple relational communities used to merge the object circle (e.g., the first relational community and the homogeneous relational community) also contain the objects to which the other belongs, making the multiple relational communities closely related in social relationships. This results in the object circle obtained by merging the multiple relational communities also having closely related social relationships, thus improving the accuracy of the obtained object circle.

[0031] This application can obtain an object circle, which includes multiple objects obtained by aggregating objects in a global object set; it can select a first content resource from a resource set, which may contain multiple content resources, the first content resource being the content resource historically consumed by the object circle; and it can obtain associated consumption data of the first content resource, including first consumption data generated by objects in the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource; thus, based on the associated consumption data of the first content resource, circle preference prediction processing can be performed on the first content resource to obtain the object circle's interest preference degree for the first content resource. Therefore, the method proposed in this application can obtain the first content resources historically consumed by the target audience's social circle. By combining the first consumption data of objects in the target audience's social circle and the second consumption data of objects in the global object set for the first content resource, the method can comprehensively perform circle preference prediction processing on the first content resource to obtain the accurate interest preference of the target audience's social circle for the first content resource. This interest preference can reflect the target audience's interest preference for the first content resource based on the global object set, thereby improving the accuracy of obtaining the target audience's interest preference for content resources. Furthermore, it can further improve the accuracy of recommending content resources to objects in the target audience's social circle based on the target audience's interest preference for content resources. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 is a schematic diagram of a data push network and a network architecture for consuming content resources provided in an embodiment of this application;

[0034] Figure 2 is a schematic diagram of a scenario where objects in an object set are divided into object layers according to an embodiment of this application;

[0035] Figure 3 is a schematic diagram of a scenario for obtaining the interest and preference of an object circle for a first content resource according to an embodiment of this application;

[0036] Figure 4 is a flowchart illustrating an object processing method provided in an embodiment of this application;

[0037] Figure 5 is a schematic diagram illustrating the effect of dividing objects into relational communities according to an embodiment of this application;

[0038] Figure 6 is a schematic diagram of a scenario for merging relational communities provided in an embodiment of this application;

[0039] Figure 7 is a schematic diagram illustrating the effect of obtaining a homogeneous relationship community of a first relationship community according to an embodiment of this application;

[0040] Figure 8 is a schematic diagram of a data push process to an object provided in an embodiment of this application;

[0041] Figure 9 is a schematic diagram illustrating the principle of obtaining the relational weight of an object in an object hierarchy according to an embodiment of this application;

[0042] Figure 10 is a schematic diagram of the application interface of a social application provided in an embodiment of this application;

[0043] Figure 11 is a schematic diagram of an interface for gift-giving on an e-commerce platform provided in an embodiment of this application;

[0044] Figure 12 is a schematic diagram of another interface for gift-giving on an e-commerce platform provided in an embodiment of this application;

[0045] Figure 13 is a flowchart illustrating a resource processing method provided in an embodiment of this application;

[0046] Figure 14 is a schematic diagram illustrating the principle of predicting the interest preference of target groups for a first content resource according to an embodiment of this application;

[0047] Figure 15 is a flowchart illustrating a method for obtaining a sequence of preferred resources for an object layer, as provided in an embodiment of this application.

[0048] Figure 16 is a schematic diagram of the effect of an effective concentric circle provided in an embodiment of this application;

[0049] Figure 17 is a schematic diagram of the effect of an invalid layer provided in an embodiment of this application;

[0050] Figure 18 is a schematic diagram of a scenario for recommending content resources to objects in an object circle according to an embodiment of this application;

[0051] Figure 19 is a flowchart illustrating a method for constructing object layers according to an embodiment of this application;

[0052] Figure 20 is a schematic diagram of the structure of an object processing device provided in an embodiment of this application;

[0053] Figure 21 is a schematic diagram of a resource processing device provided in an embodiment of this application;

[0054] Figure 22 is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0055] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] All data collected in this application (such as target audience demographics, content resources, and data on the consumption of content resources by the target audience) are collected with the consent and authorization of the data owner (such as a user, organization, or enterprise), and the collection, use, and processing of the relevant data must comply with the relevant laws, regulations, and standards of the relevant regions.

[0057] 1. Social Network: A social network is a graph structure composed of nodes (usually representing users) and edges (representing social relationships between users; in this application, these relationships are referred to as direct social relationships, meaning that two nodes (which can represent two users) connected by the same edge in the social network have a direct social relationship). It is used to describe social interactions and relationships between users, such as friend relationships, follow relationships, etc. Social network analysis aims to uncover the structural characteristics of user groups, such as community divisions, information dissemination paths, etc. For example, this social network may include the global relationship network and object relationship network (belonging to personal relationship networks) of this application.

[0058] 2. Social Community: A social community refers to a tightly connected subgraph structure within a social network, typically composed of user groups with similar attributes or close relationships. The connection density within a community is significantly higher than the connection density between communities, such as colleague communities, classmate communities, and interest communities. In this application, the social community can be a relational community segmented from a user's personal relationship network.

[0059] 3. Ego-network: A personal social network is a local social network centered on a specific user, constructing first-degree (e.g., direct friends) and second-degree (e.g., friends of direct friends) relationships. It is a subset of a global social network (e.g., a global relationship network) and is typically used to analyze the social circle structure of an individual user. This personal social network can be the object relationship network of the object in this application.

[0060] 4. Social Circle: A social circle refers to a user group formed based on a user's social attributes (such as location, interests, etc.) or close social relationships. A social circle can be a collection of multiple social communities, reflecting the multidimensional identity characteristics of a user in real-world social scenarios. For example, a user may belong to both an interest circle and a geographical circle simultaneously. In this application, the social circle can be an object-oriented circle.

[0061] 5. Likelihood Probability: This is a core concept in statistics for measuring the probability of a parameter space based on observed data. Essentially, it is a measure of the likelihood of generating observed data. In this application, it primarily refers to the representational effect of observed object behavior (such as an object's consumption behavior of content resources) on object strata (e.g., the representation of the preferred resource sequence constructed by an object strata based on the object's interest and preference for content resources within that strata).

[0062] 6. Multi-degree Relationship: A multi-degree relationship refers to an indirect connection established between users through multiple levels of social relationships (such as third-degree or fourth-degree relationships). For example, if user A and user B are friends, this is a first-degree relationship; if user B and user C are friends, this is a second-degree relationship; and if user C and user D are friends, this is a third-degree relationship. Multi-degree relationships can be used to capture broader social connections.

[0063] Please refer to Figure 1. Figure 1 is a schematic diagram of a network architecture provided in an embodiment of this application. It can be used as a network architecture for data push networks and also as a network architecture for consuming content resources. As shown in Figure 1, the network architecture may include a server 200 and a cluster of terminal devices. The cluster of terminal devices may include one or more terminal devices; the number of terminal devices is not limited here. As shown in Figure 1, the multiple terminal devices may specifically include terminal device 1, terminal device 2, terminal device 3, ..., terminal device n. As shown in Figure 1, terminal device 1, terminal device 2, terminal device 3, ..., terminal device n can all connect to the server 200 via the network, so that each terminal device can interact with the server 200 through the network connection.

[0064] As shown in Figure 1, server 200 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing 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, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Terminal devices can be smart terminals such as smartphones, tablets, laptops, desktop computers, smart TVs, in-vehicle terminals, and smart home devices. The following description uses the communication between terminal device 1 and server 200 as an example to illustrate the specific implementation of this application.

[0065] For a data push network, terminal device 1 can have a social application. Objects (i.e., users) within this social application can have social connections. All objects in the social application can form an object set. Terminal device 1 can be any user's device. Server 200 can divide the objects in this object set into object circles. These object circles are formed by groups of homogeneous objects, such as colleagues, classmates, or those sharing similar interests (e.g., data browsing). Therefore, server 200 can push data to users belonging to terminal device 1 based on these object circles. For example, it can recommend trending data from the user's object circle, such as data frequently accessed by users in that object circle. Server 200 can then push the recommended data to the user on terminal device 1, i.e., to the social application on terminal device 1, allowing the user to view the data (such as articles or short videos) pushed by server 200 on the terminal device 1's interface.

[0066] For the network architecture that consumes content resources, terminal devices 1 through n can all be user terminals, and each terminal device can have an application program. This application program can be any program capable of consuming content resources (such as videos, articles, audio, news, goods, etc.) (which can involve operations such as browsing, liking, and placing orders). Users can consume content resources through their respective terminal devices within this application. Server 200 can be the backend server for this application. Therefore, server 200 can obtain the content resource consumption status (such as consumption behavior) of each user within their respective terminal device's application. Server 200 can obtain the interest and preference of the target audience for content resources based on the consumption status of each user. This target audience can be obtained by aggregating objects from a global object set. This global object set can contain a large number of objects, such as a large number of users registered in the application. These users can include all users belonging to terminal devices 1 through n. Furthermore, the content resources consumed by users in the application can be content resources from a resource set, which can contain multiple content resources.

[0067] Please also refer to Figure 2, which is a schematic diagram of a scenario for dividing objects in an object set into object layers according to an embodiment of this application. As shown in Figure 2, the object set can contain a large number of objects. The server 200 in Figure 1 can select a first object from the object set. The selected first object can be a seed object used for subsequent diffusion of object layers. The server 200 can also obtain a second object that has a social relationship with the first object. The second object can be an object used for diffusion of object layers based on the first object.

[0068] Server 200 can obtain the first relational community to which the first object belongs (i.e., where it is located), and at least one (or more) second relational communities to which the second object belongs. The first relational community can be used to describe the first object and other objects that have social relationships with the first object, and these other objects may include the second object; and the second relational community can be used to describe the second object and other objects that have social relationships with the second object.

[0069] Server 200 can select a homogeneous relation community of the first object's first relation community from at least one second relation community of the second object, wherein other objects described by the homogeneous relation community contain the first object. That is, objects described by the second object that contain the first object's second relation community can be considered as homogeneous relation communities of the first object's first relation community.

[0070] Server 200 can merge the first relational community and its homogeneous relational communities to obtain an object circle based on the first object. This object circle can simultaneously include the objects described by the first relational community of the first object and the objects described by its homogeneous relational communities. Subsequently, server 200 can push data to the objects contained in the object circle through the obtained object circle.

[0071] Using the object processing method provided in this application, the homogeneous relationship community of the relationship community (e.g., the first relationship community) of an object (e.g., the first object) can be selected through the social relationship between objects (e.g., between the first object and the second object). Thus, the corresponding object circle can be quickly and accurately diffused through the relationship community and its homogeneous relationship community. The object circle obtained through diffusion can then be used to push more accurate data to users.

[0072] Please refer to Figure 3, which is a schematic diagram of a scenario for obtaining the interest preference of an object circle for a first content resource according to an embodiment of this application. As shown in Figure 3, the server 200 in Figure 1 can obtain the object circle and select the first content resource historically consumed by the object circle (i.e., historically consumed by objects aggregated in the object circle) from the resource set. The server 200 can obtain the associated consumption data of the first content resource. The associated consumption data may include the consumption data generated by objects in the object circle consuming the first content resource (which can be called the first consumption data) and the consumption data generated by objects in the global object set consuming the first content resource (which can be called the second consumption data). The server 200 performs circle preference prediction processing on the first content resource through the associated consumption data to obtain the interest preference of the object circle for the first content resource. This interest preference can be used to reflect the degree of interest preference of the object circle for the first content resource. In practical applications, the server 200 may acquire numerous object circles. Through the aforementioned process, the server 200 can determine the interest and preference levels of each object circle for its respective primary content resources. Subsequently, this interest and preference level can be used to perform related application tasks such as merging and filtering of various object circles and recommending resources within those circles. The specific process can also be found in the relevant descriptions in the following embodiments.

[0073] Using the method provided in this application, the first content resource can be comprehensively processed by combining the first consumption data of objects in the object circle for the first content resource and the second consumption data of objects in the global object set for the first content resource, so as to obtain the accurate interest preference degree of the object circle for the first content resource. Subsequently, the obtained interest preference degree can be used to realize various diverse association processing of the object circle.

[0074] Please refer to Figure 4, which is a flowchart illustrating an object processing method provided in an embodiment of this application. The executing entity in this embodiment can be an object processing device (hereinafter referred to as the processing device). This object processing device can be a computer device or a cluster of multiple computer devices. The computer device can be a server or other devices; this application does not impose any limitations on this. As shown in Figure 4, the method may include:

[0075] Step S401: Select a first object from the object set and obtain a second object that has a social relationship with the first object.

[0076] Specifically, the processing device can select a first object from the object set. This first object can be a seed object used for subsequent object layer diffusion. There can be one or more first objects, i.e., there can be one or more seed objects. Since the principle of object layer diffusion through each first object is the same, the following explanation can be based on the process of object layer diffusion through a single first object.

[0077] In this application, the first object can be selected in one or more rounds. After each round of diffusion of the first object selected, if the diffusion of each object layer does not cover all or most of the objects in the object set (such as the target proportion set for the objects in the object set, such as 90%), the first object of the next round can be selected to continue to diffuse more object layers through the selected first object, so that the diffusion of the object layers can cover (i.e. include) more objects.

[0078] The aforementioned object set can contain all objects, and the number of objects in the set can be large. An object can be a user, which can be a user within a social application. This social application can be any application that allows users to have social relationships, such as a web application, a mini-program, or an app (application). Users within this social application can have direct social relationships, such as being friends or following each other. For example, users A and B are friends, meaning user B is in user A's friend list, and user A is also in user B's friend list. Similarly, users A and B are related, meaning user B is in user A's following list, and user A is also in user B's following list. The type of direct social relationship (such as the type of friend relationship or the type of following relationship) can be determined based on the actual application scenario; this application does not limit the type of direct social relationship. The processing device can be the backend device of the social application (such as a backend server).

[0079] Therefore, if this is the selection of the first object in the first round, the first object can be randomly selected from the object set by setting a selection quantity. The number of selected first objects is equal to this selection quantity, which can be the set number of first objects selected in one round. Therefore, this selection quantity can also be called the unit quantity. Alternatively, to ensure the effectiveness of subsequent diffusion of the object circle, this application can also select objects with a large number of direct social objects (objects with direct social relationships, such as friends) as objects used for diffusion of the object circle. Therefore, this application can also randomly select objects from the object set whose number of social objects is greater than or equal to a set second quantity threshold as the first object by setting the selection quantity. The second quantity threshold can be set according to the actual application scenario, such as setting it to 40, etc.

[0080] If the first object is selected in a round after the first round, the processing device can obtain the currently diffused object layer, which can include the various object layers that have been diffused through the first object selected in the previous rounds.

[0081] The processing device can filter at least one candidate object from the object set using the already diffused object layers. This process can include: there can be G object layers, where G is a non-negative integer. The processing device can obtain the number of object layers each object in the object set belongs to within the G object layers that have been diffused, that is, count the number of object layers in the G object layers that contain each object. An object can have a count of one layer, which is the number of object layers in the G object layers that contain that object.

[0082] In real-world social scenarios, the more hits a user has within the expanded target audience (i.e., the number of target audience circles they belong to), the higher their social relevance across those circles. Conversely, the fewer hits a user has within the expanded target audience (i.e., the number of target audience circles they belong to), the lower their social relevance across those circles. Therefore, the processing device can use objects with fewer than H target audience circles as intermediate targets for filtering. These intermediate targets are those with relatively low social relevance within the currently expanded target audience circles. H is a positive integer, and H is less than G.

[0083] Therefore, the processing device can select intermediate objects as candidate objects if the number of objects with direct social relationships is greater than or equal to the aforementioned second threshold. There can be one or more such candidate objects. The direct social relationships of an intermediate object are those objects with which the intermediate object has a direct social relationship.

[0084] The processing device can randomly select a first object from the at least one candidate object according to the set selection quantity, and the number of selected first objects can be equal to the selection quantity. If the number of candidate objects is less than the selection quantity, the number of selected first objects can be equal to the number of candidate objects, that is, each candidate object can be treated as a first object for object layer diffusion.

[0085] In one implementation, if this is selecting the first object in the first round, the first object can also be selected using the method described above for selecting candidate objects through the already diffused object layers. However, in this case, there can be zero already diffused object layers, meaning there are no already diffused object layers. In this case, all objects in the object set can be considered as candidate objects before selecting the first object.

[0086] The aforementioned diffusion of object circles can refer to continuously expanding to more objects, thereby creating an object circle composed of multiple objects. This application can continuously expand and strengthen the object circle through the relational communities of objects, as detailed in the descriptions of the following embodiments.

[0087] Furthermore, the processing device can also acquire a second object that has a social relationship with the first object. The specific definition of this social relationship, and which objects the second object specifically includes that have a social relationship with the first object, can be found in the relevant description in step S402 below. This second object can be an object that originates from the first object and undergoes diffusion and merging of relationship communities to obtain object circles.

[0088] There may be one or more second objects. Since the principle of processing each second object is the same based on the first object, the following can be explained using one second object as an example, as described below.

[0089] Step S402: Obtain the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object.

[0090] Specifically, the processing device can acquire a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs. The first relationship community can be used to describe the first object and other objects with social relationships to the first object. The second object can belong to one of the other objects described by the first relationship community; that is, the second object can be an object within the other objects described by the first relationship community, meaning the first relationship community contains the second object. Similarly, the second relationship community can be used to describe the second object and other objects with social relationships to the second object. In other words, the first relationship community can be used to describe the objects it contains; the objects described by the first relationship community are the objects contained within the first relationship community, and the first relationship community can be used to describe the social relationships between the objects it contains. The second relationship community is similar; it can also be used to describe the social relationships between the objects it contains.

[0091] The process by which the processing device obtains the first relational community to which the first object belongs may include: the processing device obtaining the object relational network of the first object, which may contain various objects that have social relationships with the first object, centered on the first object. The object relational network of the first object may also be referred to as the first object's personal social network (ego-network).

[0092] This object relationship network can contain nodes and edges; that is, it can be composed of nodes and edges. A node represents an object, and an edge represents a direct social relationship between objects. Specifically, if two objects have a direct social relationship, then there can be an edge connecting the nodes of those two objects. The object relationship network of the first object can contain the nodes of the first object and the nodes of objects that have social relationships with the first object. These social relationships can include, but are not limited to, direct social relationships.

[0093] The processing device can partition (or split) the object relationship network centered on the first object by analyzing the connections between nodes in the first object's object relationship network. This partitioning yields at least one relationship community for the first object. Nodes with closer connections are more likely to be assigned to the same relationship community of the first object. Each relationship community of the first object contains nodes of the first object, but different relationship communities of the first object may not contain overlapping nodes of other objects. In other words, an object (not the first object) with a social relationship with the first object will not exist in multiple relationship communities of the first object simultaneously, but rather in only one of the first object's relationship communities.

[0094] In one implementation, the processing device can employ the fast-unfolding (Louvain) method (an unsupervised community discovery algorithm based on modularity optimization) to partition the object relation network of a first object into communities, thereby obtaining at least one relation community for the first object. This fast-unfolding method can be used to mine tightly connected groups of nodes (i.e., "communities," referred to as relation communities in this application) from complex social networks. Its core idea is to maximize the optimization of the network's grouping structure through modularity, achieving linear time complexity and suitability for processing large-scale social networks. The modularity can be calculated based on the connections between nodes, reflecting the tightness between the searched node groups. This allows for the merging of highly connected node groups to discover and partition relation communities composed of closely related nodes.

[0095] Through the above process, at least one relational community of the first object can be obtained by dividing the object relational network of the first object. The first relational community can be any one of the at least one relational communities divided for the first object. It should be noted that during the diffusion process of object layers, object layers are diffused through each relational community of the first object to obtain one or more object layers. Since the principle of object layer diffusion through each relational community of the first object is the same, this application uses the process of object layer diffusion through any one relational community of the first object (i.e., the first relational community) as an example to explain the relevant process in detail.

[0096] Please refer to Figure 5, which is a schematic diagram illustrating the effect of dividing an object into relational communities according to an embodiment of this application. As shown in Figure 5, object u0 can refer to the first object. Here, the object relational network of the first object can include objects with a social hierarchy of 2 levels or less with the first object, including objects a1 to a10. The fast-unfolding method described above can be used to divide the object relational network of object u0 into communities, thereby obtaining three relational communities for object u0, including relational community 1, relational community 2, and relational community 3.

[0097] The relationship community 1 can contain object u0 and objects with social relationships with object u0 (including objects a4, a5, a6 and a7); the relationship community 2 can contain object u0 and objects with social relationships with object u0 (including objects a1, a2 and a3); the relationship community 3 can contain object u0 and objects with social relationships with object u0 (including objects a8, a9 and a10).

[0098] In one embodiment, the processing device can also acquire the object relationship network of the second object, and can divide the object relationship network of the second object with the second object as the center, according to the same principle as the division of the relationship community of the first object, to obtain at least one relationship community of the second object. This at least one relationship community can be referred to as at least one second relationship community of the second object. That is, the at least one second relationship community includes all the relationship communities obtained by dividing the object relationship network of the second object.

[0099] In this context, the object relationship network of an object (such as the first object or the second object) can be a local relationship network within the global relationship network of all objects in the object set; that is, the object relationship network of an object can be a part of the global relationship network. This global relationship network can be constructed through edges connecting the nodes of each object in the object set and the nodes of objects with direct social relationships. This global relationship network can be used to represent the social connections among all users in the object set.

[0100] In this application, the definition of the object relation network for each object is the same, that is, the principle of obtaining the object relation network of each object from the global relation network is the same. Therefore, the definition of the object relation network is explained here by taking the object relation network of the first object as an example, as described below.

[0101] The nodes in the object relationship network of the first object include the nodes of the first object and the nodes of objects that have social relationships with the first object. The objects that have social relationships with the first object may include: objects whose social hierarchy with the first object is within the Nth social hierarchy, that is, objects whose social hierarchy with the first object is from the 1st social hierarchy to the Nth social hierarchy.

[0102] Here, an object with a social hierarchy of level i to the first object refers to an object in the global relationship network whose shortest link to the first object contains i edges, where i is less than or equal to N and is a positive integer. An object with a social hierarchy of level 1 to the first object refers to an object with a direct social relationship to the first object, that is, an object represented by a node in the global relationship network that is directly connected to a node of the first object.

[0103] In a global relational network, the shortest link from one object to another is the shortest path from the nodes of one object to the other object based on the edges between the nodes. This shortest link can be formed by the nodes and edges that are traversed from the nodes of one object to the nodes of the other object with the fewest steps.

[0104] Based on the above description, for example, if the direct social relationship is a friend relationship, then the relationship between the first object's friend (i.e., the object with a direct social relationship with the first object) and the first object is the first level of social hierarchy; the relationship between the first object's friend's friend and the first object is the second level of social hierarchy; the relationship between the first object's friend's friend's friend and the first object is the third level of social hierarchy, and so on. The object with a social hierarchy of the i-th level with the first object can also be called the first object's i-th degree friend.

[0105] Therefore, the second object can include: objects within the first relationship community that have a social hierarchy of M levels with the first object, where M is a positive integer and less than or equal to N. That is, M can be any positive integer within the range of N. If M equals 1, it indicates that the second object includes objects within the first relationship community that have a direct social relationship with the first object; if M equals 2, it indicates that the second object includes objects within the first relationship community that have a first-level social hierarchy with the first object, and objects that have a second-level social hierarchy with the first object; and so on.

[0106] In this context, a smaller social hierarchy level between the user and the first object indicates a closer social connection, while a larger social hierarchy level indicates a more distant social connection. Therefore, a smaller M results in higher accuracy in subsequent object-based diffusion based on the first object, closer social connections between objects within the resulting object layer, and potentially lower coverage (fewer objects diffused within the object layer). Conversely, a larger M results in lower accuracy in subsequent object-based diffusion based on the first object, weaker social connections between objects within the resulting object layer, but potentially higher coverage (more objects diffused within the object layer) and higher diffusion efficiency. Therefore, the value of M can be flexibly set according to actual application needs. M can be any positive integer less than or equal to N, and is an adjustable value.

[0107] Step S403: Select a homogeneous relation community of the first relation community from at least one second relation community; the first object belongs to other objects described by the homogeneous relation community.

[0108] Specifically, the processing device can select a homogeneous relation community of the first relation community from at least one second relation community of the second object. The homogeneous relation community can be a second relation community that contains the first object. That is, the first object can be another object described by the homogeneous relation community. In other words, the first object is an object among the other objects described by the homogeneous relation community. The homogeneous relation community contains the first object (i.e., a node that contains the first object).

[0109] In practical applications, if the first object and the second object are colleagues, the first relationship community of the first object that includes the second object can be the colleague community of the first object, and the second relationship community of the second object that includes the first object can also be the colleague community of the second object. Therefore, merging the first relationship community that includes the second object and the second relationship community that includes the first object can yield a larger colleague circle (an object circle belonging to the colleague type) of the first object and the second object.

[0110] As can be seen from the above, the first relational community of the first object contains the second object, and the homogeneous relational community of the first relational community selected from one or more second relational communities of the second object also contains the first object. In other words, the relational communities of different objects that mutually contain each other are likely to have the same community attributes (i.e., communities of the same nature, such as colleagues, or hometown, classmates, etc.). These relational communities belong to each other's homogeneous relational communities. This application can select the homogeneous relational communities of a certain relational community (such as the first relational community) through the social association between objects.

[0111] Step S404: Merge the first relationship community with the homogeneous relationship community to obtain the object circle based on the first object.

[0112] Specifically, the processing device can merge the first relation community of the first object and the homogeneous relation community selected by the second object for the first relation community to obtain an object circle based on the first object, which may include the first relation community and the homogeneous relation community.

[0113] There can be many second objects within the first relation community. Therefore, the processing device can select each second object's homogeneous relation community from the second relation community of each second object according to the above principle. The first relation community can be merged with each homogeneous relation community of the first relation community (including the homogeneous relation communities of the first relation community selected from the second relation community of each second object) to obtain an object circle that is finally diffused through the first relation community of the first object.

[0114] For example, the second objects within the first relational community include objects 1, 2, and 3. The processing device can select homogeneous relational communities of the first relational community from at least one relational community of object 1, from at least one relational community of object 2, and from at least one relational community of object 3. Thus, the processing device can merge the first relational community, the homogeneous relational communities of the first relational community selected from the relational communities of object 1, the homogeneous relational communities of the first relational community selected from the relational communities of object 2, and the homogeneous relational communities of the first relational community selected from the relational communities of object 3, to obtain an object layer diffused through the first relational community of the first object.

[0115] Furthermore, since the first object can have multiple relational communities, the processing device can, according to the principle described above (i.e., the principle of object layer diffusion through the first relational community), diffuse object layers through each relational community of the first object separately, thereby obtaining object layers diffused through each relational community of the first object separately. One object layer can be obtained through diffusion from one relational community of the first object. Therefore, the number of object layers that can ultimately be obtained based on the number of relational communities of the first object is directly proportional to the number of relational communities it has.

[0116] Please refer to Figure 6, which is a schematic diagram of a scenario for merging relationship communities provided by an embodiment of this application. As shown in Figure 6, object u1 can be the first object, and object u1 can have three relationship communities, including relationship community c1, relationship community c2, and relationship community c3. Relationship community c1 can contain object u1 and objects with social relationships with object u1 (including objects b3, b4, b5, and b6); relationship community c2 can contain object u1 and objects with social relationships with object u1 (including objects b1 and b2); relationship community c3 can contain object u1 and objects with social relationships with object u1 (including objects b7, b8, and b9). Here, relationship community c2 of object u1 can be regarded as the first relationship community.

[0117] Furthermore, object b2 can be a second object, and object b2 can have two second relationship communities, including relationship community c4 and relationship community c5. Relationship community c4 can contain object b2 and objects with social relationships with object b2 (including object b12 and object b13); relationship community c5 can contain object b2 and objects with social relationships with object b2 (including object u1, object b1, object b10, and object b11).

[0118] Since the relation community c5 of object b2 contains object u1, relation community c5 can be a homogeneous relation community of relation community c2 of object u1. Relation community c5 can be merged with relation community c2. Furthermore, object b1 in relation community c2 can also be considered a second object; therefore, homogeneous relation communities of relation community c2 can be selected from at least one relation community of object b1. Thus, relation community c2 and all its homogeneous relation communities can be merged to obtain an object hierarchy based on relation community c2.

[0119] Please refer to Figure 7, which is a schematic diagram illustrating the effect of obtaining a homogeneous relation community of a first relation community according to an embodiment of this application. As shown in Figure 7, object u2 can be the first object, and the relation community e1 of object u2 can be the first relation community. The second objects in the first relation community include object d2 and object d5. Of course, the second objects can also include other objects; here, object d2 and object d5 are used as examples for illustration.

[0120] The relation community e2 of object d2 contains object u2; therefore, relation community e2 can be considered a homogeneous relation community of relation community e1. Similarly, the relation community e3 of object d5 also contains object u2; therefore, relation community e3 can also be considered a homogeneous relation community of relation community e1. The object hierarchy ultimately derived through relation community e1 can simultaneously include relation communities e1, e2, and e3.

[0121] Furthermore, the processing device can acquire at least one object circle derived from the first object, which may include various object circles derived from the various relational communities of the first object. The processing device can further filter the at least one object circle to obtain a filtered target object circle. This target object circle may be an object circle derived from the first object and ultimately put into use (i.e., adopted). There may be one or more target object circles. The process of filtering the target object circle can be described as follows.

[0122] The processing device can optimize each of the at least one object layer to obtain at least one optimized object layer, wherein one of the at least one object layers can be optimized to obtain a corresponding optimized object layer.

[0123] Any one of the at least one object layer can be called a candidate object layer. Since the principle of optimizing each object layer to obtain the corresponding optimized object layer is the same, the process of optimizing the candidate object layer to obtain the corresponding optimized object layer will be used as an example for specific explanation.

[0124] The candidate object layer can be obtained by merging L relation communities. These L relation communities can include one relation community of the first object and homogeneous relation communities of the selected relation community. L is an integer greater than 1, and the specific value of L can be determined according to the actual application scenario. The processing device can count the number of relation communities that each node in the candidate object layer belongs to in the L relation communities. The number of communities corresponding to each node can be counted, which is the number of relation communities containing that node in the L relation communities.

[0125] The more times a node (i.e., an object) in a candidate object circle hits (i.e., belongs to the number of communities) in the L relational communities used to merge and obtain the candidate object circle, the higher the social closeness of that object in the candidate object circle; conversely, the fewer times a node (i.e., an object) in a candidate object circle hits (i.e., belongs to the number of communities) in the L relational communities used to merge and obtain the candidate object circle, the lower the social closeness of that object in the candidate object circle. Therefore, the processing device can remove nodes in the candidate object circle whose corresponding number of communities (i.e., the number of communities they belong to) is less than K, to obtain the optimized object circle corresponding to the candidate object circle. This ensures a high confidence that the users retained in the optimized object circle belong to the same object circle. K is a positive integer, and K is less than L. The value of K can be set according to the actual application scenario, such as L / 7. The objects represented by the nodes whose corresponding number of communities is less than K can be considered as objects with weak social connections in the candidate object circle.

[0126] The processing device can obtain the optimized object layer corresponding to each object layer obtained based on the diffusion of the first object, i.e., at least one optimized object layer, according to the above principle.

[0127] The processing device can select the final target object layer from the at least one optimized object layer. This process may include: the processing device can obtain the number of nodes contained in each optimized object layer. An optimized object layer can obtain a number of nodes, which is the number of nodes contained in the optimized object layer.

[0128] In this application, excessively small object layers obtained from diffusion can also be removed. Therefore, the processing device can eliminate (i.e., remove) optimized object layers whose corresponding number of nodes is less than or equal to a set first quantity threshold, to obtain the target object layer diffused based on the first object. The number of nodes contained in the target object layer can be greater than the first quantity threshold. The first quantity threshold can be set according to the actual application scenario, such as setting it to 50.

[0129] The processing device can follow the principle described above (i.e., the principle of obtaining the target object circle based on the first object) to finally diffuse, optimize and filter the target object circle based on each seed object (i.e. each first object). This target object circle is the final object circle in which the objects in the object set are located.

[0130] Using the method of this application, a wide range of target object circles can be obtained by spreading through the relationship community of each object based on the social connections between objects. Furthermore, the same object can be in multiple different target object circles to satisfy the multi-dimensional circle identity of the same object, such as the same object being in a colleague circle, classmate circle, and interest circle at the same time.

[0131] Subsequently, data can be pushed to objects through the various target object layers obtained from the diffusion. The specific process can be seen in the description of the corresponding embodiment in Figure 8 below.

[0132] This application proposes a merging algorithm for homogeneous relational communities, which can progressively merge multi-source communities (multi-level, multi-object relational communities) that meet the homogeneity condition (such as each containing the other's relational communities) to ultimately form a globally optimized object concentric structure.

[0133] In this application, there can be multiple first objects, which can be selected in multiple rounds, with multiple first objects selected in each round. Therefore, the efficient diffusion of object layers can be achieved in parallel and in a distributed manner using multiple selected first objects. Thus, by selecting a small number of seed objects from the entire set of objects, rapid diffusion to the object layers of all objects can be achieved, greatly improving the efficiency of acquiring object layers and reducing the time and resource consumption (such as device resources) required for acquisition. Using the method provided in this application, the discovery of object layers in ultra-large-scale social networks can be supported, enabling the feasible construction of multi-degree relationship object layers, such as a maximum of 2M degrees of social relationships (i.e., 2M social levels) within an object layer. Furthermore, it supports a single object belonging to multiple object layers simultaneously, satisfying the implementation of multiple identity attributes for a single object.

[0134] In this application, by selecting a subset of seed objects for object layer diffusion, the computational complexity is reduced from O(n log n) to O(k log k), compared to existing solutions that traverse all objects to find object layers. Here, n is the total number of objects, and k is the number of seed objects selected in this application. This significantly reduces the computational complexity of object layer discovery and improves the efficiency and computational power of object layer diffusion.

[0135] This application can select a first object from an object set and obtain a second object that has a social relationship with the first object; it can also obtain a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have a social relationship with the first object, the second object is one of the other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have a social relationship with the second object; this application can select a homogeneous relationship community of the first relationship community from the at least one second relationship community; the first object is one of the other objects described by the homogeneous relationship community; thus, the first relationship community and the homogeneous relationship community can be merged to obtain an object circle based on the first object. Therefore, the method proposed in this application can merge the relational communities of the first object (e.g., the first relational community) and the relational communities of the second object (e.g., selected homogeneous relational communities) through the social connections between objects (e.g., the first object and the second object), to efficiently obtain the object circle based on the first object, thus improving the efficiency of obtaining the object circle. Furthermore, the multiple relational communities used to merge the object circle (e.g., the first relational community and the homogeneous relational community) also contain the objects to which the other belongs, making the multiple relational communities closely related in social relationships. This results in the object circle obtained by merging the multiple relational communities also having closely related social relationships, thus improving the accuracy of the obtained object circle.

[0136] Please refer to Figure 8, which is a schematic diagram of a data push process to an object according to an embodiment of this application. As shown in Figure 8, the process may include:

[0137] Step S801: Obtain the relationship weight of the target object in each of the Q target object layers.

[0138] Specifically, any object in the aforementioned object set can be referred to as a target object. Since the principle of data push for each object is the same, the process of pushing data to the target object will be used as an example for specific explanation. The target object can exist within Q target object circles obtained through diffusion. These Q target object circles can be optimized from the various object circles obtained through diffusion (such as object circles obtained through diffusion from various seed objects). That is, the Q target object circles can be the target object circles in which the target object is located among the various target object circles obtained through diffusion from various seed objects. Q is a positive integer, and the specific value of Q can be determined according to the actual application scenario. The process of obtaining the corresponding target object circles based on the diffusion from various seed objects can be referred to the relevant description in the corresponding embodiment of Figure 4 above.

[0139] The processing device can obtain the relationship weight of the target object in each of the Q target object circles. Each target object can have one relationship weight within its respective target object circle. The relationship weight of the target object within a target object circle reflects the strength of its social connections with other objects within that circle. A smaller relationship weight indicates a weaker social connection between the target object and other objects in that circle, and vice versa.

[0140] Here, the first object is the seed object in the target object circle obtained by diffusion based on the first object; that is, any object can be a seed object in the target object circle obtained by diffusion based on this object. Therefore, the processing device can obtain the link number between the target object and the seed object in its first target object circle, and can use this link number as the relationship weight of the target object in any target object circle. The first target object circle can be any of the Q target object circles mentioned above where the seed object is not the target object; that is, the first target object circle is not an object circle obtained by diffusion based on the target object, and the target object is not a seed object in the first target object circle.

[0141] In other words, the relational weight of a target object in any target object circle that is not a seed object can be the number of links (i.e., the number of links, or the number of paths) between the target object and the seed object in that target object circle.

[0142] For example, if the seed object in a target object circle is object A, then the relationship weight of the target object in that target object circle can be the number of links from the target object to object A. For example, if there are 5 links, then the relationship weight is 5.

[0143] If the target object is a seed object within a target object circle, its relation weight within that circle can be considered very high, perhaps even infinite. The target object circles derived from the target object can be called the second target object circle, and the target object is a seed object within that second circle. In other words, the relation weight of an object within a target object circle derived from it can be considered very high, perhaps even infinite, or set to a very high value (e.g., 1000). Alternatively, the relation weight of a seed object within a target object circle can be set to a fixed weight, which can be a relatively high value, perhaps higher than the relation weight of most objects within that target object circle (e.g., exceeding a certain set percentage of objects).

[0144] The processing device can calculate and obtain the relation weights of the target object in each target object layer according to the same principle described above.

[0145] Please refer to Figure 9, which is a schematic diagram illustrating the principle of obtaining the relational weight of an object in an object hierarchy according to an embodiment of this application. As shown in Figure 9, the exemplary object hierarchy here can include object u3 (i.e., node u3), object f1 (i.e., node f1), object f2 (i.e., node f2), object f3 (i.e., node f3), object f4 (i.e., node f4), and object f5 (i.e., node f5). Object u3 can be a seed object in this object hierarchy, that is, this object hierarchy can be obtained by spreading through object u3.

[0146] Taking object f2 as an example, there can be three links between object f2 and seed object u3: link L1, link L2, and link L3. Link L1 can be a path from node f2 to node f1 and then to node u3, and it includes the edges connecting node f2 and node f1, as well as the edges connecting node f1 and node u3. Link L2 can be a path from node f2 to node u3, and it includes the edges connecting node f2 and node u3. Finally, link L3 can be a path from node f2 to node f3 and then to node u3, and it includes the edges connecting node f2 and node f3, as well as the edges connecting node f3 and node u3.

[0147] Since there are 3 links between object f2 and the seed object in this object layer, the relation weight of object f2 in this object layer can be 3.

[0148] Step S802: Push data to the target object based on the relationship weight of the target object in each target object circle.

[0149] Specifically, the processing device can push data to the target object, i.e., recommend data, based on the relationship weight of the target object within each target object circle. For example, the higher the relationship weight of a target object within a target object circle, the more data from that circle can be recommended to the target object; conversely, the lower the relationship weight of a target object within a target object circle, the less data from that circle can be recommended to the target object. Data within a target object circle can include data generated by actions performed by objects within that circle (such as viewing, browsing, or purchasing). This data could be electronic data such as articles, songs, products, or videos; this data can also be referred to as the content data of the actions performed.

[0150] In one exemplary implementation, the processing device can acquire multiple hotspot data points and a popularity score for each hotspot data point within each target object circle to which the target object belongs. For example, it can acquire 10 hotspot data points for a single target object circle. The hotspot data for a target object circle can be data related to the hotspot operations of objects within that circle, such as data with high pageviews, high purchase volumes, or high play counts.

[0151] A trending data point can have a popularity score within its target audience segment. This popularity score can be calculated using set scoring rules. For example, the scoring rules can be based on user actions performed on the trending data within the target audience segment. These actions could include the number of times the trending data was viewed and the duration of those views. The popularity score can be calculated by weighting the number of views and the duration of those views. Therefore, the more times a data point is viewed and / or the longer it is viewed, the higher its popularity score can be.

[0152] In one implementation, each data point within a target audience segment can have a popularity score, and the multiple trending data points within that segment can be the top few (e.g., the top 10 with the highest popularity scores) ranked from highest to lowest within that segment. As the name suggests, a data point's popularity score within its target audience segment reflects its popularity within that segment; a higher popularity score indicates greater popularity within that segment, and vice versa.

[0153] The processing device can weight (or multiply) the popularity scores of hot data in each target object's social circle based on the relation weights within that circle, thus obtaining a weighted popularity score for each hot data point. By weighting the popularity score of a hot data point using relation weights, the weighted popularity score for that hot data point can be obtained.

[0154] The processing device can push data to the target object based on the weighted heat score of each hot data point (including hot data points in each target object circle where the target object is located). This process may include: For example, the processing device can sort each hot data point in descending order of its weighted heat score to obtain sorted hot data points. In the sorted hot data points, the hot data points with higher weighted heat scores can be arranged at the front.

[0155] The processing device can push the top T trending data points from the sorted trending data to the target object, where T is a positive integer. T can be less than the total number of trending data points across all target object circles, such as T equaling 20. When pushing (i.e., recommending) these top T trending data points to the target object, the processing device can push a list of these T trending data points sorted in descending order of their weighted popularity scores. Pushing these T trending data points to the target object can refer to the target device, such as a social application logged into the target object's account.

[0156] Please refer to Figure 10, which is a schematic diagram of the application interface of a social application provided in an embodiment of this application. As shown in Figure 10, the application interface g1 can be an interface for playing short videos in the social application. In this application interface g1, the short videos that the user is currently viewing can be played. Furthermore, the application interface g1 can also contain a control "Friends are watching today". Clicking on the control "Friends are watching today" will display the application interface g2.

[0157] In application interface g2, the short video played in application interface g1 can be minimized so that the user can still see the full view of the short video played in application interface g1 in application interface g2. Application interface g2 can also contain a recommendation list triggered by the control "What Friends Are Watching Today." This recommendation list can contain data pushed to the user, which can be short videos. The data pushed to the user in the recommendation list could include the T trending data points selected above. For example, these T trending data points could include the short video "The Charm of Technology~" uploaded by creator "Xiao Xin," the short video "Let's Play Basketball Together!" uploaded by creator "Xiao Le," the short video "The Charm of Intelligent Driving~" uploaded by creator "Xiao Li," and the short video "<Healthy Recipes>" uploaded by creator "Xiao Kang."

[0158] Users can click on short videos recommended to them in the recommendation list, allowing the social application to play the clicked short videos for the user to watch.

[0159] The above methods of pushing data to target objects are merely examples. In practical applications, data can be pushed to target objects in any way that incorporates the relationship weight of the target object within each target object hierarchy. The following examples illustrate several more scenarios for pushing data to objects based on object hierarchies, as described below:

[0160] 1. In-circle content recommendation: Representative content consumed (such as purchased or viewed) by people within a target audience's circle (e.g., content with high consumption popularity) can be recommended to people within that target audience's circle who have not yet consumed this content. This can improve the reach and efficiency of the pushed content, significantly increase the viewing time of the pushed content, and at the same time, make the user's content experience richer and more surprising.

[0161] 2. In-circle red dot push: In the form of "friends are watching", representative content consumed by users in the target circle is pushed to other users, enhancing users' sense of participation in the pushed content.

[0162] 3. Summary of Content Within the Target Audience: Summarize the content consumed within the target audience's social circle and, based on the characteristics of the target audience's social circle (such as "colleagues are watching"), push the content to other users within the target audience's social circle to improve the relevance and attractiveness of the pushed content.

[0163] 4. Content Push to Users Within the Target Audience: Push content published by users within the target audience's circle to other users within the target audience's circle, promote interaction and content dissemination among users, encourage users to publish more content, and encourage more users to publish content, thus building a better development ecosystem for the content platform.

[0164] Similar recommendation logic can be extended to other application scenarios, such as gift-giving in e-commerce platforms. When an individual needs to give a gift on an e-commerce platform, gifts or products given by other individuals within that individual's social circle can be recommended (i.e., pushed) to the individual who needs to give the gift. This provides an important supply of candidate items (such as candidate gifts or products) for the cold start phase of gift-giving scenarios on e-commerce platforms, significantly improving user experience and product conversion rates.

[0165] Please refer to Figures 11 and 12. Figure 11 is a schematic diagram of an interface for gift-giving on an e-commerce platform provided by an embodiment of this application, and Figure 12 is a schematic diagram of another interface for gift-giving on an e-commerce platform provided by an embodiment of this application. As shown in Figure 11, interface h1 can be a gift-giving interface on an e-commerce platform. In interface h1, a list of products that the current object has historically gifted can be displayed. This product list can include the product "Nut Gift Pack" displayed in box v1. In addition, interface h1 can also display a list of "Everyone is Giving Gifts". This list of "Everyone is Giving Gifts" can include products that other objects have historically gifted (such as the products "Ceramic Cup" and "Professional Fishing Gear" displayed in box v2). The products that other objects have historically gifted can include products that other objects within the object circle of the current object have historically gifted, or / and products that other objects in the object set have historically gifted.

[0166] Furthermore, users can swipe on interface h1. For example, swiping up will display interface h2 in Figure 12. Interface h2 and interface h1 are actually the same interface. Compared to interface h1, interface h2 displays more products that other objects have delivered in the past. In addition to the products "ceramic cup" and "professional fishing gear" shown in box v2 above, it can also include products "breakfast milk" and "pastries and snacks", etc.

[0167] When the "Everyone's Giving Gifts" list on interface h2 displays various items that other objects have historically given gifts to, it can also display notification information related to the historical gifts given to each item. This notification information can be gift information for the item from all objects in the object set (such as the number of gifts given and / or gift comments), or gift information for the item from objects in the current object's social circle. For example, if other objects in the current object's social circle have not given a gift to a certain item, but an object in the object set has given a gift to that item, the notification information for that item can be gift information for that item from all objects in the object set; and if other objects in the current object's social circle have given a gift to a certain item, the notification information for that item can be gift information for that item from objects in the current object's social circle.

[0168] For example, the prompt message v4 in interface h2 could include the total number of times all objects in the object set have given the product "ceramic mug" as a gift (999+ times), as well as the most frequent comments from objects in the object set regarding the product "ceramic mug," such as "exquisite and practical." Similarly, the prompt message v5 in interface h2 could include the total number of times all objects in the object set have given the product "professional fishing gear" as a gift (99+ times). In this case, the number of comments from objects in the object set regarding the product "professional fishing gear" is very small (e.g., less than the set target number), making it impossible for the product "professional fishing gear" to have any valid comments. Therefore, the prompt message v5 could not include any comments regarding the product "professional fishing gear." The prompt message v6 in interface h2 could be "given as a gift by a friend," which could refer to when... If other people in the same social circle as the previous recipient have given "Breakfast Milk" as a gift, the notification message "Friends have given it" can be displayed when the total number of times other people have given "Breakfast Milk" as a gift is less than a set threshold. This notification message can also include the most frequent comments from other people regarding "Breakfast Milk," such as "Smooth and delicious." Similarly, notification message v7 in interface h2 can include the total number of times other people in the same social circle have given "Pastries and Snacks" as a gift (11 times) and the most frequent comments from other people regarding "Pastries and Snacks," such as "Rich in milk flavor." This notification message v7 can be displayed when the total number of times other people have given "Pastries and Snacks" as a gift is greater than or equal to the aforementioned set threshold.

[0169] Through the above process, products that other objects in the current object's social circle have previously gifted, or products that other objects in the object set have previously gifted, are recommended to the current object. The current object can then use these recommended products to gift their friends, such as those who have a direct social relationship with the current object.

[0170] Using the method provided in this application, one or more object circles can be obtained for each object in the object set. This allows for more accurate data push to each object based on the object circles to which they belong. Furthermore, it solves the problem of accurate data push to users in cold start scenarios where a user lacks data operation records (such as browsing history). The method in this application uses data from user operations (such as browsing) within a user's object circle as a reference for accurate data push, improving the method and accuracy of data push and thus enhancing the user's data browsing experience.

[0171] This application proposes a method for mining object circles based on personal social networks. By mining homogeneous relationship communities in different personal communities (such as object relationship communities) and performing community diffusion and merging, it achieves the mining of multi-attribute circles in massive and complex social networks, possessing applicability-level computational feasibility and circle mining accuracy. Since the homogeneous relationship communities discovered in this application are very likely to come from different subsets of the same object circle, extremely accurate object circles can be mined by merging relationship communities with their homogeneous relationship communities.

[0172] Furthermore, based on the verification results, the data recommendation logic based on object circles in this application solves the problem of not being able to capture the gap in users' social interest associations, and brings significant improvements to several core indicators of the data platform, including daily active user growth, increased content publication volume, and increased content browsing time, etc., which greatly improves the user experience and promotes the healthy development of the content platform.

[0173] Please refer to Figure 13, which is a flowchart illustrating a resource processing method provided in an embodiment of this application. The executing entity in this embodiment can be a resource processing device, which can be a single computer device or a cluster of multiple computer devices. The computer device can be a server or other devices; this application does not impose any limitations on this. As shown in Figure 13, the method may include:

[0174] Step S1301: Obtain the object layer. The object layer includes multiple objects obtained by aggregating objects in the global object set.

[0175] Specifically, the resource processing device can acquire an object hierarchy, which is obtained by aggregating objects from a global object set. This object hierarchy can contain multiple objects aggregated from the global object set. The global object set can contain all objects; for example, all objects could be all registered users in an application (such as a social application). An object can be a user, such as a user account (also called an object account) in the application. The resource processing device can be a background device for this application.

[0176] There can be many object layers obtained from aggregation. Since the processing logic for each object layer is the same in this embodiment, the following description will focus on processing one object layer as an example. An object layer can contain some objects from the global object set, and different object layers can contain the same objects.

[0177] The specific process of constructing and acquiring the object layer can be found in the relevant description in the embodiment corresponding to Figure 19 below.

[0178] Step S1302: Select the first content resource from the resource set. The resource set contains multiple content resources. The first content resource is the content resource that the target audience has historically consumed.

[0179] Specifically, the resource processing device can select a first content resource from the resource set, and there can be one or more such first content resources. The resource set can contain multiple content resources, and it can include all the content resources that need to be examined. For example, the content resources in this application can be any resource that can be consumed by any object, such as videos, articles, audio, goods, activities, services, or gifts given away on online e-commerce platforms. The type of content resource can be flexibly determined according to the actual application scenario, and this application does not impose any restrictions on it.

[0180] The first content resource refers to the content resources historically consumed by the target audience. In this application, the consumption of content resources by an object can be an operation on that content resource, i.e., consumption is manifested as a behavior. For example, the consumption of content resources by an object can be any of the following: liking, forwarding, collecting, commenting, browsing, using, placing an order, exchanging content resources for general resources, etc. Here, the general resource can refer to a general asset, and exchanging content resources for general resources can refer to purchasing content resources using general resources.

[0181] In one implementation, the method of selecting a first content resource from a resource set may include: the resource processing device can acquire the consumption frequency of objects in an object circle for each content resource in the resource set, which can be referred to as the first consumption frequency. An object in the object circle may have a first consumption frequency for a content resource in the resource set. For example, the first consumption frequency can be the total frequency (i.e., the total number of times) that objects in the object circle consume the content resource within a specified period (which can be called the target period) (such as within the past month or the past week). If the consumption is browsing, the first consumption frequency can be the cumulative number of times all objects in the object circle click and browse the content resource within that period. The target period is a pre-set time period used to statistically analyze the consumption data of objects on content resources, such as within the past month or the past week. Both the first consumption data and the second consumption data are statistically analyzed based on this time period.

[0182] Similarly, the resource processing device can also obtain the consumption frequency of objects in the global object set for content resources in the resource set, which can be referred to as the second consumption frequency. An object in the global object set can have a second consumption frequency for a single content resource in the resource set. This second consumption frequency can also be the total frequency (i.e., the total number of times) that objects in the global object set consume that content resource within a target time period (such as within the past month or week). If the consumption is browsing, then the second consumption frequency can be the cumulative number of times all objects in the global object set click and browse that content resource within the target time period.

[0183] Through the above process, each content resource in the resource set can obtain its own first consumption frequency and second consumption frequency. The resource processing device can identify content resources in the resource set whose first consumption frequency is greater than or equal to a set first frequency threshold, and whose second consumption frequency is greater than or equal to a set second frequency threshold, as first content resources. Both the first and second frequency thresholds can be set according to the actual application scenario. The first frequency threshold can be set in the range of 1-10 times, and the second frequency threshold can be set in the range of 10-100 times. The specific values ​​can be dynamically adjusted based on factors such as the size and activity level of the object circle and the global object set. Since the second frequency threshold is for the global object set, and the first frequency threshold is for the object circle, the second frequency threshold can be greater than the first frequency threshold. Alternatively, in some special application scenarios, the second frequency threshold and the first frequency threshold can have other relative values.

[0184] The aforementioned first content resource can be understood as a content resource selected from the resource set that is effective (e.g., effectively consumed) for the target audience, i.e., a content resource with reference value.

[0185] The above method of selecting the first content resource with consumption validity (i.e., confidence) is only an example. In actual application scenarios, other methods can also be used to select the first content resource.

[0186] Step S1303: Obtain the associated consumption data of the first content resource. The associated consumption data includes the first consumption data generated by objects in the object circle consuming the first content resource, and the second consumption data generated by objects in the global object set consuming the first content resource.

[0187] Specifically, the resource processing device can acquire associated consumption data of the first content resource. This associated consumption data can include consumption data generated by objects in the object circle consuming the first content resource (which can be referred to as first consumption data), and consumption data generated by objects in the global object set consuming the first content resource (which can be referred to as second consumption data). That is, the associated consumption data of the first content resource can include both the first consumption data and the second consumption data.

[0188] Both the first and second consumption data can be any of the following: frequency of likes, frequency of shares, frequency of favorites, frequency of comments, frequency of views, frequency of use (e.g., if the content resource is an emoji, the frequency of use could be the number of times the emoji is used), frequency of redemption, and the amount of general resources redeemed (e.g., the total amount of general resources redeemed for a content resource; if the content resource is a product, the amount of general resources redeemed could be the total sales revenue of that product), etc. The types of the first and second consumption data can be the same, such as both being view frequency, both being redemption frequency, or both being general resource redemption volume, etc.

[0189] The type of consumption and the type of consumption data can correspond. For example, if consumption is the act of adding to favorites, then the consumption data could be the frequency of adding to favorites; if consumption is the act of exchanging general resources for content resources, then the consumption data could be the frequency of exchanging or the amount of general resources exchanged; and if consumption is the act of placing an order for content resources, then the consumption data could be the number of orders (i.e., the total number of orders placed for content resources).

[0190] Both the first and second consumption data mentioned above can be statistically obtained from the first content resource within a specified period of time (such as the target period mentioned above). For example, if consumption is the action of liking, and the consumption data is the type of like frequency, then the first consumption data can be the total number of times all objects in the target circle like the first content resource within the target period, and the second consumption data can be the total number of times all objects in the global object set like the first content resource within the target period.

[0191] For example, if consumption is the action of browsing, and consumption data is the type of browsing frequency, then the first consumption data can be the total number of times all objects in the object circle within the target time period browse the first content resource, and the second consumption data can be the total number of times all objects in the global object set within the target time period browse the first content resource.

[0192] For example, if consumption is the act of exchanging general resources for content resources, and consumption data is the type of general resource exchange volume, then the first consumption data can be the total amount of general resources exchanged for the first content resource by all objects in the target period's object circle, and the second consumption data can be the total amount of general resources exchanged for the first content resource by all objects in the global object set within the target period; and so on.

[0193] Through the above process, the associated consumption data of the first content resource can be obtained. This associated consumption data may include the first consumption data of the object circle for the first content resource, and the second consumption data of the global object set for the first content resource. There can be multiple first content resources, and the process of obtaining the associated consumption data of each first content resource can be the same as described above. The process of obtaining the associated consumption data of each first content resource can be the same and independent of each other.

[0194] In one embodiment, the resource processing device can obtain the associated consumption data of each content resource in the resource set in step S1302 according to the same principle as the above-described acquisition of the associated consumption data of the first content resource. The associated consumption data of each content resource can include the first consumption data and the second consumption data of each content resource. Therefore, the first content resource in step S1302 can also be selected from the resource set through the associated consumption data (including the first consumption data and the second consumption data) of each content resource in the resource set.

[0195] For example, content resources whose first consumption data value is greater than or equal to a set first value threshold, and whose second consumption data value is greater than or equal to a set second value threshold, can be selected as the first content resources. The first value threshold and the second value threshold can be different and can be set according to the actual application scenario. For example, the first value threshold can be set to 15 and the second value threshold can be set to 100.

[0196] Step S1304: Based on the associated consumption data of the first content resource, perform circle preference prediction processing on the first content resource to obtain the interest preference degree of the target circle for the first content resource.

[0197] Specifically, the resource processing device can use the associated consumption data obtained from the first content resource to perform segmented preference prediction processing on the first content resource, that is, predict the interest preference of the target segment on the first content resource, thereby obtaining the interest preference degree of the target segment on the first content resource. The process of predicting the interest preference degree of the target segment on each first content resource is also the same and independent of each other, and a corresponding interest preference degree can be predicted for each first content resource.

[0198] Among them, the interest preference of the target audience for a first content resource can reflect the target audience's interest preference for that first content resource, that is, reflect the characteristics (such as preference) of that first content resource in the target audience's circle. The higher the interest preference, the more interested the target audience is in that first content resource, and the more characteristic that first content resource is in the target audience's circle; conversely, the lower the interest preference, the less interested the target audience is in that first content resource, and the less characteristic that first content resource is in the target audience's circle.

[0199] In one implementation, the process of predicting the interest and preference of target demographics for the first content resource using associated consumption data may include:

[0200] The resource processing device can calculate the ratio between the value of the first consumption data and the value of the second consumption data, and can use this ratio as the initial interest preference of the target audience for the first content resource. For example, the first consumption data can be denoted as vv. circle This first consumption data can also be understood as consumption within the circle, and the second consumption data can be recorded as vv. global This second consumption data can also be understood as global consumption; therefore, the initial interest preference P0 can be:

[0201] Since the consumption level of content resources within an object circle is also affected by the popularity of the content resource in the global (e.g., the global object set), the content resources that are highly consumed within the object circle (e.g., ranked high in consumption frequency) are very likely to be the content resources that are highly consumed globally. In order to weaken the influence of this global popularity on the popularity of content resources within the object circle (which can also be understood as the influence of characteristics), this application introduces the denominator (i.e., global consumption) in the above formula (1), thereby weakening the global popularity of content resources and obtaining the initial interest preference degree. This initial interest preference degree can better reflect the object circle's preference for the first content resource.

[0202] The resource processing device can optimize the initial interest preference obtained above to obtain the interest preference of the target audience for the first content resource. This process may include: the resource processing device can obtain the total number of objects contained in the target audience (i.e., the number of objects contained in the target audience). Since the size of different target audiences (i.e., the number of objects contained) is different, the larger the audience, the higher the consumption enthusiasm for a certain content resource will be in the prior sense. In order to balance (and weaken) the different impacts of the different sizes of different target audiences on the preference for content resources, this application can introduce the audience size to achieve normalization processing of the initial interest preference obtained above, as described below.

[0203] The resource processing device can normalize the initial interest preference using the number of objects in the object circle obtained above. This normalized initial interest preference can be referred to as the intermediate interest preference. If the resource processing device can obtain a circle size balance parameter and calculate the product between this parameter and the initial interest preference, and the ratio of this product to the number of objects in the object circle can be used as the intermediate interest preference, then the circle size balance parameter can be denoted as c1, and the number of objects in the object circle can be denoted as uv. circle Therefore, the intermediate interest preference degree P1 can be:

[0204] The circle size balancing parameter c1 can be a constant factor, typically set to a value greater than or equal to 1. To more effectively reduce the impact of different circle sizes, c1 can be set smaller; conversely, to less effectively reduce the impact of different circle sizes, c1 can be set larger. For example, in scenarios with large differences in circle size, c1 can be set to 2; when the differences in circle size are small, c1 can be set to 1.2, etc. The specific value can be determined through experiments or experience based on factors such as the degree of difference in circle size and data distribution.

[0205] Because the consumption volume of top-tier content (i.e., high-frequency content resources) far exceeds that of other content resources (especially UGC content resources), the target audience, in practical application scenarios, prefers to promote content resources from ordinary creators to incentivize them to create more and better content. However, due to the significant difference in consumption volume between top-tier content and other content resources, to narrow the gap in interest preferences among the target audience and improve the recognizability of their interest in other content resources, this application can smooth the numerator and denominator of the intermediate interest preference using a non-linear function. That is, the resource processing device can smooth the aforementioned intermediate interest preference to obtain the interest preference score, as described below.

[0206] The aforementioned intermediate interest preference can include a first preference parameter of the object circle for the first content resource, and a second preference parameter of the global object set for the first content resource. The first preference parameter can be the numerator of this intermediate interest preference (because the numerator of the intermediate interest preference is mainly obtained through consumption within the circle), such as vv. circle ·c1 or (vv circle ·c1) / uv circle (soon UV) circle (Place it in the numerator); the second preference parameter can be the denominator of the intermediate interest preference (because the denominator of the intermediate interest preference is mainly obtained through global consumption), such as vv global Or vv global ·uv circle (soon UV) circle (Place it in the denominator). This application can use a smoothing function to smooth the first preference parameter and the second preference parameter to obtain the interest preference degree CP, as shown in the following formula.

[0207] Where f1 represents the smoothing function used to smooth the first preference parameter (which can be called the first smoothing function), f2 represents the smoothing function used to smooth the second preference parameter (which can be called the second smoothing function), and c2 represents a constant factor (which can be called the global balancing parameter), used to further balance (e.g., reduce) the magnitude of the global second preference parameter. c2 can be set to a value less than 1. The first smoothing function and the second smoothing function can be the same smoothing function or different smoothing functions, depending on the actual application requirements. For example, both the first smoothing function and the second smoothing function can be logarithmic smoothing functions (shown in the formula), or other applicable smoothing functions.

[0208] This application can smooth the first preference parameter using a first smoothing function to obtain a smoothed first preference parameter, which is represented as the numerator of the interest preference degree; and can smooth the second preference parameter using a second smoothing function combined with a constant factor, c2, to obtain a smoothed second preference parameter, which is represented as the denominator of the interest preference degree. The interest preference degree is composed of the smoothed first preference parameter and the smoothed second preference parameter, as detailed in the above formula (3).

[0209] Therefore, the optimization process for the initial interest preference degree in this application may include the normalization process and the smoothing process described above.

[0210] Please refer to Figure 14, which is a schematic diagram illustrating the principle of predicting the interest preference of an object circle towards a first content resource according to an embodiment of this application. As shown in Figure 14, the resource processing device can acquire first consumption data generated by the object circle consuming the first content resource, and second consumption data generated by the global object set consuming the first content resource. The resource processing device can combine the first and second consumption data to calculate the initial interest preference of the object circle towards the first content resource. The resource processing device can normalize the initial interest preference to obtain an intermediate interest preference, and smooth the intermediate interest preference to obtain the predicted interest preference of the object circle towards the first content resource. Subsequently, the obtained interest preference can also be used to realize application tasks such as merging and filtering various object circles and recommending resources within the circle.

[0211] The method provided in this application has strong scenario scalability. It can be applied to various scenarios of resource recommendation through object circles, which can improve the overall product experience (such as the experience of consuming content resources).

[0212] In this application, there can be multiple object layers and multiple first content resources for any object layer. The above exemplarily describes the process of obtaining the interest preference degree of any first content resource in any object layer. It can be understood that the resource processing device can obtain the interest preference degree of each first content resource in each object layer according to the same principle. For a first content resource in an object layer, a corresponding interest preference degree can be obtained, which reflects the interest preference (i.e., characteristic) of the first content resource in that object layer. Different object layers may have the same first content resource, but the interest preference degree of the first content resource in different object layers may be different and calculated separately.

[0213] This application can obtain an object circle, which includes multiple objects obtained by aggregating objects in a global object set; it can select a first content resource from a resource set, which contains multiple content resources, and the first content resource is the content resource that the object circle has historically consumed; and it can obtain the associated consumption data of the first content resource, which includes first consumption data generated by objects in the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource; thus, based on the associated consumption data of the first content resource, circle preference prediction processing can be performed on the first content resource to obtain the interest preference degree of the object circle for the first content resource. Therefore, the method proposed in this application can obtain the first content resources historically consumed by the target audience's social circle. By combining the first consumption data of objects in the target audience's social circle and the second consumption data of objects in the global object set for the first content resource, the method can comprehensively perform circle preference prediction processing on the first content resource to obtain the accurate interest preference of the target audience's social circle for the first content resource. This interest preference can reflect the target audience's interest preference for the first content resource based on the global object set, thereby improving the accuracy of obtaining the target audience's interest preference for content resources. Furthermore, it can further improve the accuracy of recommending content resources to objects in the target audience's social circle based on the target audience's interest preference for content resources.

[0214] Please refer to Figure 15, which is a flowchart illustrating a process for obtaining a preference resource sequence for an object layer according to an embodiment of this application. Since the principles for obtaining preference resource sequences for each object layer are the same and independent, this embodiment will use the process of obtaining a preference resource sequence for one object layer as an example for explanation. As shown in Figure 15, the process may include:

[0215] Step S1501: Select the first content resources among the multiple first content resources whose interest preference degree is greater than or equal to the set preference degree threshold as the preferred content resources of the target circle.

[0216] In one implementation, there may be multiple first content resources for an object circle. The resource processing device can identify the first content resource among these multiple first content resources where the interest preference is greater than or equal to a set preference threshold as the preferred content resource for that object circle, i.e., the content resource for which the object circle has an interest preference. The specific value of the preference threshold can be set according to the actual application scenario.

[0217] Among them, there can be at least one, that is, one or more, of the preferred content resources of the target circle.

[0218] Step S1502: Sort at least one preferred content resource according to the order of interest preference degree from large to small to obtain the preferred resource sequence of the object circle.

[0219] In one embodiment, the resource processing device can sort the at least one preferred content resource according to its interest preference level from largest to smallest, thereby constructing a preference resource sequence for the object circle. Since representing the object circle's interest preference for content resources through a single content resource is rather weak, this application can construct a preference resource sequence for the object circle to more completely and accurately reflect the object circle's interest preference for content resources.

[0220] The preference resource sequence of an object circle includes the first content resource of that object circle whose interest preference degree is greater than or equal to the preference degree threshold, and each first content resource in the preference resource sequence is sorted in descending order of its interest preference degree in that object circle.

[0221] Since there can be multiple object layers, the resource processing device can construct the preferred resource sequence for each object layer according to the same principle described above. Each object layer can have a corresponding constructed preferred resource sequence. This application can uniformly represent each object layer through its preferred resource sequence; that is, the characteristics of an object layer (including the characteristics of consuming content resources) can be represented by the preferred resource sequence of that object layer.

[0222] After constructing the preference resource sequence for each object circle, this application can also use the preference resource sequence for each object circle to perform applications such as merging the object circles, filtering the object circles, and recommending content resources within the circles, as described below.

[0223] In one implementation, the filtering process for object circles may include: if the number of content resources in the preferred resource sequence of an object circle is greater than or equal to a set quantity threshold, then the preferred resource sequence can be used to recommend content resources to objects within that object circle. This indicates that the number of content resources with which the object circle has interests is sufficient to serve as a reference for recommending content resources to objects within that object circle; therefore, it can be understood that the object circle has been retained. This quantity threshold can be set according to the actual application scenario, such as 20. For example, multiple content resources ranked first in the preferred resource sequence can be recommended to objects within the object circle that have not yet consumed these multiple content resources. The recommended content resources can be those that the object has not yet consumed, or other content resources related to (e.g., of the same or similar type) the multiple content resources ranked first in the preferred resource sequence can be obtained and recommended to objects within the object circle that have not yet consumed these other content resources.

[0224] If the number of content resources in the preferred resource sequence of an object circle is less than a set threshold, the object circle can be frozen. This indicates that the number of content resources with which the object circle has interests is too small to be used as a reference for content resource recommendations; therefore, it can be understood as filtering out the object circle.

[0225] Because an object circle may be inactive during certain periods, with its members consuming fewer content resources and showing little interest in those resources, it is highly likely that the object circle will be frozen during these inactive periods. Subsequently, if a large-scale event related to this object circle is held, its members are very likely to become unusually active around the time of the event. Therefore, the frozen object circle can be activated at this point. For example, an object circle could be an alumni circle, whose members are alumni of the same school. If the large-scale event is a school anniversary celebration, then around the time of the anniversary celebration, the members of the alumni circle are very likely to become very active.

[0226] In other words, the frozen object layers can wait to be activated rather than being completely deleted (i.e. removed), as described below.

[0227] The resource processing device can reacquire the preference resource sequence of an object layer (which can be described using any object layer as an example) according to a set target period, and can use the reacquired preference resource sequence as the updated preference resource sequence for that object layer. This target period can be a set period for updating and reacquiring the preference resource sequence of the object layer. For example, the target period can be one day, indicating that the preference resource sequence of the object layer can be reacquired and updated every other day, thus ensuring the timeliness of the acquired preference resource sequence.

[0228] Therefore, if the object circle is in a frozen state, and the number of content resources in the update preference resource sequence is greater than or equal to the aforementioned set threshold, the resource processing device can activate the frozen object circle and use the update preference resource sequence to recommend content resources to objects in the activated object circle. For example, multiple content resources ranked first in the update preference resource sequence can be recommended to objects in the object circle that have not yet consumed these multiple content resources; the recommended content resources can be those that the object has not yet consumed.

[0229] If the object circle is in a frozen state, and the number of content resources in the update preference resource sequence is less than the aforementioned threshold, then the object circle can remain frozen. Conversely, if the object circle is not frozen, and the number of content resources in the update preference resource sequence is less than the aforementioned threshold, then the object circle can be frozen. Similarly, a frozen object circle can wait to be activated later.

[0230] Please refer to Figure 16, which is a schematic diagram of the effect of an effective layer provided in an embodiment of this application. As shown in Figure 16, circles containing the letter 'd' can represent objects, and ellipses enclosing the circles can represent object layers. Here, object layer Q1 is used as an example for explanation. Object layer Q1 can contain 6 objects: d5, d7, d8, d9, d10, and d11. The consumption resource pool of object layer Q1 (including the various content resources consumed by object layer Q1) can include content resource 1, content resource 2, content resource 3, and content resource 4; and the arrows pointing from objects to content resources in the figure can represent the consumption of content resources by objects. Among them, content resources 1 and 4 are content resources consumed in small quantities by object circle Q1, while content resources 2 and 3 are content resources consumed frequently by object circle Q1. Content resources 2 and 3 can be content resources in the preferred resource sequence of object circle Q1. Here, there are enough content resources in the preferred resource sequence of object circle Q1 (such as content resources with a quantity greater than or equal to the above quantity threshold), indicating that object circle Q1 has sufficient preferred content resources. Therefore, object circle Q1 can be an effective circle, that is, an object circle that can be retained.

[0231] Please refer to Figure 17, which is a schematic diagram of the effect of an invalid layer provided in an embodiment of this application. As shown in Figure 17, similarly, circles containing the letter 'e' can represent objects, and ellipses enclosing the circles can represent object layers. Here, object layer Q2 is used as an example for explanation. Object layer Q2 can contain 6 objects: d4, d5, d6, d7, d8, and d9. The consumption resource pool of object layer Q2 can contain content resources 5, 6, 7, and 8; and the arrows pointing from objects to content resources in the figure can represent the consumption of content resources by objects. Among them, content resources 5, 6, 7, and 8 are all content resources consumed in small quantities by object layer Q2. The preferred resource sequence of object layer Q2 can be empty, that is, it indicates that object layer Q2 does not have preferred content resources. Therefore, object layer Q2 is an invalid layer, and object layer Q2 can be frozen. The frozen object layer Q2 can be understood as being filtered out.

[0232] Any two of the above multiple object layers may include a first object layer and a second object layer, that is, the first object layer and the second object layer may be any two object layers obtained in step S1301 above.

[0233] The resource processing device can obtain the sequence similarity between the preference resource sequences of the first object layer and the preference resource sequences of the second object layer. For example, the resource processing device can compare the content resources contained in the preference resource sequences of the first and second object layers to obtain the content resource overlap between them. This content resource overlap can be the greater of the overlap between the preference resource sequences of the first and second object layers (which can be called the first resource overlap) and the overlap between the preference resource sequences of the second and first object layers (which can be called the second resource overlap).

[0234] For example, the first resource overlap degree of the preference resource sequence of the first object layer with respect to the preference resource sequence of the second object layer can be: the number of content resources in the preference resource sequence of the first object layer that overlap with the content resources in the preference resource sequence of the second object layer, divided by the number of content resources contained in the preference resource sequence of the first object layer. And, the second resource overlap degree of the preference resource sequence of the second object layer with respect to the preference resource sequence of the first object layer can be: the number of content resources in the preference resource sequence of the second object layer that overlap with the content resources in the preference resource sequence of the first object layer, divided by the number of content resources contained in the preference resource sequence of the second object layer.

[0235] For example, the preference resource sequence of the first object layer includes five content resources: z1, z2, z3, z4, and z5. The preference resource sequence of the second object layer includes three content resources: z2, z5, and z6. Therefore, the overlapping content resources between the preference resource sequences of the first and second object layers are content resources z2 and z5. The overlap between the preference resource sequences of the first and second object layers is 2 / 5, and the overlap between the preference resource sequences of the second and first object layers is 2 / 3. Therefore, the overlap between the preference resource sequences of the first and second object layers is the larger of 2 / 5 and 2 / 3, which is 2 / 3.

[0236] Alternatively, when calculating the content resource overlap between the preference resource sequences of the first and second object layers, it can be done using the top content resources (such as the first L content resources) in the preference resource sequence of the first object layer, and the top content resources (such as the first L content resources) in the preference resource sequence of the second object layer. L is a positive integer, and its specific value can be determined based on the actual application scenario. This calculation method is the same as the method described above, which uses the entire preference resource sequence of the first and second object layers.

[0237] For example, the header content resources in the preference resource sequence of the first object layer can include five content resources: z1, z2, z3, z4, and z5. Similarly, the header content resources in the preference resource sequence of the second object layer can include five content resources: z2, z5, z6, z7, and z8. Therefore, the overlapping content resources between the header content resources of the first and second object layer preference resource sequences are content resources z2 and z5. The overlap between the preference resource sequences of the first and second object layers is 2 / 5, and the overlap between the preference resource sequences of the second and first object layers is also 2 / 5. Therefore, the overlap of the aforementioned content resources between the preference resource sequences of the first and second object layers is also 2 / 5. In this way, we can focus on and consider the overlap between the head content resources in the preference resource sequence of the first object circle and the head content resources in the preference resource sequence of the second object circle.

[0238] The resource processing device can use the content resource overlap between the preferred resource sequence of the first object layer and the preferred resource sequence of the second object layer as the sequence similarity between the preferred resource sequence of the first object layer and the preferred resource sequence of the second object layer.

[0239] Alternatively, the preference resource sequence of the first object layer can contain identifiers (such as names or numbers) for each content resource, and the preference resource sequence of the second object layer can also contain identifiers (such as names or numbers) for each content resource. The resource processing device can also predict the sequence similarity between the preference resource sequences of the first and second object layers using a similarity prediction model. This similarity prediction model can be a trained model capable of predicting and outputting the similarity between two sequences. For example, the similarity prediction model can generate a representation vector (which can be called a first feature vector) of the preference resource sequence of the first object layer, and can generate a representation vector (which can be called a second feature vector) of the preference resource sequence of the second object layer. Thus, the vector similarity (such as cosine similarity) between the first and second feature vectors can be obtained as the sequence similarity between the preference resource sequences of the first and second object layers.

[0240] The above method is only an example. In actual implementation, other methods can be used to obtain the sequence similarity between the preference resource sequence of the first object layer and the preference resource sequence of the second object layer. The specific method used to obtain the sequence similarity between the preference resource sequence of the first object layer and the preference resource sequence of the second object layer can be determined according to the actual application scenario. This application does not impose any restrictions on this.

[0241] If the sequence similarity between the obtained preference resource sequence of the first object layer and the preference resource sequence of the second object layer is greater than or equal to the set similarity threshold, then the first object layer and the second object layer can be considered to be similar object layers; or, if the overlap between the preference resource sequence of the first object layer and the preference resource sequence of the second object layer, and the overlap between the preference resource sequence of the second object layer and the preference resource sequence of the first object layer, are both greater than or equal to the similarity threshold, then the first object layer and the second object layer can be considered to be similar object layers.

[0242] If the first object layer and the second object layer are similar object layers, the resource processing device can directly classify the first object layer and the second object layer as duplicate object layers. Alternatively, if the first object layer and the second object layer are similar object layers, the resource processing device can further determine the duplication of the similar first object layer and the second object layer. The result of the determination can be that the first object layer and the second object layer are duplicate object layers, or that the first object layer and the second object layer are not duplicate object layers.

[0243] The method provided in this application offers a practically feasible solution, whether directly treating similar object layers as duplicate object layers or further performing duplicate judgments on similar object layers. In practical applications, the number of objects in the global object set is extremely large, potentially reaching hundreds of millions. Therefore, the number of object layers aggregated from these global object sets is also very large, and the same object can simultaneously reside in multiple different object layers. Directly comparing objects within each object layer to merge them would be extremely labor-intensive and computationally demanding, exceeding the capabilities of computers in practical implementations. Therefore, this application reflects the similarity between object layers by analyzing the similarity of the preference resource sequences constructed from each object layer. This reduces the number of similar object layers requiring merging or duplicate judgments to a practically feasible range, making the merging of object layers feasible.

[0244] In one implementation, the process by which the resource processing device performs a repeatability determination on similar first and second object layers may include: the resource processing device comparing objects in the first object layer with objects in the second object layer to obtain the object overlap degree between the first and second object layers. Similarly, the resource processing device can obtain the first object overlap degree of objects in the first object layer with respect to objects in the second object layer, and the second object overlap degree of objects in the second object layer with respect to objects in the first object layer, and can use the larger of the first and second object overlap degrees as the object overlap degree between the first and second object layers.

[0245] For example, the first object overlap degree can be the value obtained by dividing the number of overlapping objects in the first object layer and the second object layer by the number of objects in the first object layer, and the second object overlap degree can be the value obtained by dividing the number of overlapping objects in the first object layer and the second object layer by the number of objects in the second object layer.

[0246] For example, the first object layer includes five objects: d1, d2, d3, d4, and d5. The second object layer includes four objects: d3, d4, d5, and d6. The overlapping objects in the first and second object layers are objects d3, d4, and d5. Therefore, the overlap rate of the first object layer is 3 / 5, the overlap rate of the second object layer is 3 / 4, and the overlap rate between the first and second object layers is 3 / 4.

[0247] If the obtained object overlap is greater than or equal to the set overlap threshold, then the first object layer and the second object layer can be considered to be overlapping object layers. If the object overlap is less than the overlap threshold, then the first object layer and the second object layer can be considered not to be overlapping object layers. This overlap threshold can be set according to the actual application scenario, such as 90%.

[0248] If the first and second object circles mentioned above are duplicate object circles, then a circle merging process can be performed on the first and second object circles, that is, merging the first and second object circles into a single object circle (which can be called a merged object circle). After merging the first and second object circles to obtain the merged object circle, the preference resource sequence of the merged object circle can be re-acquired according to the principles described above, and the circle merging and content resource recommendation within the circle can be performed through the re-acquired preference resource sequence.

[0249] This application effectively filters out object layers that are ineffective for the current application scenario by filtering and deduplicating object layers, thereby significantly reducing the computational load of object layers in the recommendation process, greatly improving the application efficiency of object layers, and avoiding invalid computation of object layers.

[0250] Furthermore, in the cold start scenario of mid-to-long-tail content (such as UGC content published by ordinary creators, i.e., self-created content resources), the local dissemination effect can be linked to the target audience's social circle, realizing diffusion from personal friendships to the target audience's social circle. For example, a fitness video published by an ordinary creator can spread rapidly through the "fitness enthusiast" circle, thereby significantly improving the growth speed and dissemination effect of UGC content. Here, the cold start scenario can refer to the initial stage where basic operating conditions need to be re-established due to the lack of historical data / initial state information. For example, in the cold start scenario of consuming UGC content, it can mean that the UGC content does not yet exist or has very little data generated by consumption (such as consumption frequency, consumption time, consumer group, etc.), resulting in the lack of basic data (i.e., data generated by consumption) for subsequent wider dissemination of the UGC content in the cold start scenario. However, through the method provided in this application, even in the cold start scenario of consuming UGC content, effective and widespread dissemination of UGC content to users can still be achieved through the target audience's social circle.

[0251] Furthermore, after obtaining the preferred resource sequence of an object circle, the resource processing device can also sum the interest preferences of each content resource in the preferred resource sequence to obtain the overall interest characteristic value of the object circle for the preferred resource sequence. This interest characteristic value can be used to reflect the characteristics (i.e., preferences) that the preferred resource sequence of the object circle possesses relative to the global object for the object circle. The larger the interest characteristic value, the more characteristic the preferred resource sequence of the object circle is for the object circle, that is, there is a greater difference between the consumption of the preferred resource sequence by the global object and the consumption by objects within the object circle; conversely, the smaller the interest characteristic value, the less characteristic the preferred resource sequence of the object circle is for the object circle, that is, there is less difference between the consumption of the preferred resource sequence by the global object and the consumption by objects within the object circle. In other words, the larger the value of this interest characteristic, the more it indicates that the content resources in the preference resource sequence are content resources with interest preferences that are unique to the object circle. The preference resource sequence can better reflect and characterize the object circle's interest preferences and consumption semantics (such as the type of content resources consumed).

[0252] Therefore, if the interest characteristic value is greater than or equal to the set threshold, it indicates that the preferred resource sequence is unique to the target audience, and the preferred resource sequence can be used to recommend content resources to objects within that target audience. Conversely, if the interest characteristic value is less than the threshold, it indicates that the preferred resource sequence is less unique to the target audience, and the difference between its consumption by global objects and by objects within that target audience is small. In this case, content resource recommendations can be directly made to objects within that target audience based on the global consumption data of objects in the global object set for content resources in the resource set.

[0253] In this context, an object in the global object set can have global consumption data for a content resource within the resource set. The type of this global consumption data can be the same as or different from the types of the first and second consumption data mentioned above, depending on the specific application scenario. For example, this global consumption data could directly be the second consumption data (such as consumption frequency, or browsing frequency).

[0254] Therefore, the process of recommending content resources to objects in an object sphere based on the consumption data of objects in a global object set for content resources in a resource set can include: the resource processing device sorts each content resource in the resource set according to the descending order of the global consumption data values ​​of each content resource in the resource set, to obtain sorted content resources. The resource processing device can then recommend the top K content resources from this sorted list to objects in the object sphere, such as to objects in the object sphere that have not yet consumed these K content resources. The content resources recommended to any object can include those K content resources that the object has not yet consumed.

[0255] In one implementation, when recommending content resources to objects within an object circle using a preference resource sequence (such as the original preference resource sequence or an updated preference resource sequence), the preference resource sequence can be combined with the global consumption data of objects in the global object set for the content resources in the resource set. For example, each content resource in the resource set can be sorted in descending order of its global consumption data value to obtain sorted content resources. Then, the first S content resources in the preference resource sequence and the first K content resources in the sorted content resources can be recommended to objects in the object circle that have not yet consumed the S and K content resources. The content resources recommended to any object can include those not yet consumed by the object from the S and K content resources. S is a positive integer, and the numerical relationship between S and K can be set according to the actual application scenario. For example, if 10 content resources need to be recommended to an object, S can be equal to 2, and K can be equal to 8. This allows you to set the proportion of each of the preferred resource sequence and the sorted content resources within the total amount of content resources to be recommended to an object. The content resources recommended to the object can be referred to as the secondary content resources.

[0256] Please refer to Figure 18, which is a schematic diagram of a scenario for recommending content resources to objects in an object circle according to an embodiment of this application. As shown in Figure 18, the resource processing device can obtain the preferred resource sequence of the object circle and the global consumption data of objects in the global object set for the content resources in the resource set. The resource processing device can select the top content resources preferred by the object circle through the preferred resource sequence. For example, the top content resources may include multiple content resources ranked first in the preferred resource sequence, such as content resources z1, z2, z3, and z4.

[0257] The resource processing device can also select the most consumed content resources from the global object set based on the global consumption data of the content resources in the global object set. For example, the most consumed content resources may include multiple content resources with the highest values ​​of the global consumption data in the resource set. Here, these multiple content resources may include content resource z5, content resource z6, content resource z7, and content resource z8.

[0258] The resource processing device can merge the top content resources preferred by the target audience and the high-consumption content resources of the global target audience to obtain a recommended resource set for the target audience. The recommended resource set may include the selected content resources z1, z2, z3, z4, z5, z6, z7 and z8.

[0259] The resource processing device can use this recommended resource set to recommend content resources to each object in the object circle. For example, the resource processing device can recommend content resources that each object in the recommended resource set has not yet consumed to each object in the object circle.

[0260] Alternatively, when calculating the interest preference scores of each first content resource in the target audience, this application can directly use these interest preference scores to recommend content resources to objects within the target audience. The interest preference scores of each first content resource can serve as the weight for recommending each first content resource. The higher the interest preference score of a content resource, the higher its recommendation weight, and the more frequently it will be recommended; conversely, the lower the interest preference score of a content resource, the lower its recommendation weight, and the less frequently it will be recommended.

[0261] Through the process described above in this application, namely, by predicting the interest and preference of the first content resources of each object circle, effective circle merging, circle filtering, and recommendation of content resources within the circle are achieved, thereby enhancing the diversity and richness of processing and applying object circles.

[0262] In one implementation, the resource set described in this application can be of many types. One resource set can correspond to one content ecosystem (i.e., resource ecosystem), and one resource set can contain several content resources of the same type (or one resource set can also contain a combination of custom content resources, which can be determined according to the actual application scenario). For example, a resource set can be a video set (i.e., the content resource is video), an article set (i.e., the content resource is article), an audio set (i.e., the content resource is audio), or a product set (i.e., the content resource is product), and so on. The process described in this application can use any resource set as a sampling sample to calculate the interest preference of content resources in the target audience's social circle, and to implement a series of processing on the target audience's social circle (such as merging, filtering, and recommending content resources within the circle). In actual application scenarios, for each resource set, the interest preference of content resources in various resource sets in the target audience's social circle can be calculated according to the process described above, thereby implementing different processing on the target audience's social circle according to different types of resource sets.

[0263] The method described in this application allows for more accurate and specific acquisition of content resources preferred by various target audience groups, enabling precise resource recommendations to be made within each group. For example, in a video recommendation scenario (where the content resource is video), within the "fitness enthusiast" group, priority can be given to recommending videos related to fitness tutorials and healthy eating, or high-quality videos with rich content and high production quality, rather than general fitness content. This can increase the willingness of content consumers (i.e., the target audience) to interact with and consume content resources, not only improving the user's consumption experience but also significantly increasing user consumption metrics (such as viewing time and interaction rate).

[0264] This application can be applied to recommendation scenarios for any type of content resources, such as, but not limited to, video recommendation scenarios, gift-giving scenarios on e-commerce platforms, article recommendation scenarios, music recommendation scenarios, news recommendation scenarios, book recommendation scenarios, and advertising recommendation scenarios, etc.

[0265] For example, in this video recommendation scenario, top videos favored by a specific audience segment (such as the top-ranked videos in the preferred resource sequence) and frequently consumed videos from the global audience set can be recommended to audiences within that segment who haven't yet consumed those videos. Similarly, in the gift-giving scenario on an e-commerce platform, when an audience needs to give a gift, gifts previously given by other audience members can be recommended (i.e., pushed) to the audience who needs to give a gift. This provides a crucial supply of candidate items (such as candidate gifts) for the cold start phase of the gift-giving scenario on the e-commerce platform, significantly improving user experience and gift conversion rates. For example, the recommendation effect can be seen in the descriptions in Figures 10, 11, and 12 below.

[0266] Please refer to Figure 10, which is a schematic diagram of the application interface of a social application provided in an embodiment of this application. As shown in Figure 10, the application interface g1 can be an interface for playing short videos in the social application. In this application interface g1, the short videos that the user is currently viewing can be played. Furthermore, the application interface g1 can also contain a control "Friends are watching today". Clicking on the control "Friends are watching today" will display the application interface g2.

[0267] In application interface g2, the short video played by the user in application interface g1 can be minimized so that the user can still see the full view of the short video played in application interface g1 in application interface g2. Application interface g2 can also contain a recommendation list triggered by the control "What Friends Are Watching Today". This recommendation list can contain short videos pushed to the user, including top short videos preferred by the user's social circle and short videos with high consumption across the entire social group. For example, this recommendation list could include short videos uploaded by creator "Xiao Xin" ("The Charm of Technology~"), "Xiao Le" ("Let's Play Basketball!"), "Xiao Li" ("The Charm of Intelligent Driving~"), and "Xiao Kang" ("<Healthy Recipes>").

[0268] Users can click on short videos recommended to them in the recommendation list, allowing the social application to play the short videos they click for them to watch.

[0269] Please refer to Figures 11 and 12. Figure 11 is a schematic diagram of an interface for gift-giving on an e-commerce platform provided by an embodiment of this application, and Figure 12 is a schematic diagram of another interface for gift-giving on an e-commerce platform provided by an embodiment of this application. As shown in Figure 11, interface h1 can be a gift-giving interface on an e-commerce platform. In interface h1, a list of products that the current object has historically gifted can be displayed. This product list can include the product "Nut Gift Pack" displayed in box v1. In addition, interface h1 can also display a list of "Everyone is Giving Gifts". This list of "Everyone is Giving Gifts" can include products that other objects have historically gifted (such as the products "Ceramic Cup" and "Professional Fishing Gear" displayed in box v2). The products that other objects have historically gifted can be products that other objects within the object circle of the current object have historically gifted (such as the top gift products preferred by objects within the object circle, or multiple products ranked first in the preferred resource sequence of the object circle), or / and products that other objects in the global object set have historically gifted (such as the products that objects in the global object set frequently gift).

[0270] Furthermore, users can swipe on interface h1. For example, swiping up will display interface h2 in Figure 12. Interface h2 and interface h1 are actually the same interface. Compared to interface h1, interface h2 displays more products that other objects have delivered in the past. In addition to the products "ceramic cup" and "professional fishing gear" shown in box v2 above, it can also include products "breakfast milk" and "pastries and snacks", etc.

[0271] When the "Everyone's Giving Gifts" list on interface h2 displays various items that other objects have historically given gifts to, it can also display notification information related to the historical gifts given to each item. This notification information can be gift-giving information for the item from all objects in the global object set (such as the number of gifts given and / or gift comments), or gift-giving information from objects in the current object's object circle. For example, if other objects in the current object's object circle have not given a gift to a certain item, but an object in the object set has given a gift to that item, the notification information displayed for that item can be gift-giving information for that item from all objects in the global object set; and if other objects in the current object's object circle have given a gift to a certain item, the notification information displayed for that item can be gift-giving information for that item from objects in the current object's object circle.

[0272] For example, the prompt message v4 in interface h2 could include the total number of times all objects in the global object set have gifted the product "Ceramic Cup" (999+ times), as well as the most frequent comments from objects in the global object set regarding the product "Ceramic Cup," such as "Exquisite and practical." Similarly, the prompt message v5 in interface h2 could include the total number of times all objects in the global object set have gifted the product "Professional Fishing Gear" (99+ times). In this case, the number of comments from objects in the global object set regarding the product "Professional Fishing Gear" is very small (e.g., less than the set target number), resulting in no valid comments for the product "Professional Fishing Gear." Therefore, the prompt message v5 could not include any comments for the product "Professional Fishing Gear." The prompt message v6 in interface h2 could be "Gifted by a friend," which could include... This refers to other people in the current user's social circle who have given "Breakfast Milk" as a gift. The notification message "Friends gave it as a gift" can be displayed when the total number of times other people have given "Breakfast Milk" as a gift is less than a set threshold. The notification message can also include the most frequent comments from other people regarding "Breakfast Milk," such as "smooth and delicious." Similarly, the notification message v7 in interface h2 can include the total number of times other people in the current user's social circle have given "Pastries and Snacks" as a gift (11 times) and the most frequent comments from other people regarding "Pastries and Snacks," such as "rich in milk flavor." This notification message v7 can be displayed when the total number of times other people have given "Pastries and Snacks" as a gift is greater than or equal to the aforementioned set threshold.

[0273] Through the above process, products that other objects in the current object's social circle have previously gifted, or products that other objects in the global object set have previously gifted, are recommended to the current object. The current object can then use these recommended products to gift its friends, such as those who have a direct social relationship with the current object.

[0274] Furthermore, this application also provides a mathematical explanation of the validity of the above-mentioned mathematical modeling (i.e., the above-mentioned process for predicting the interest preference degree) of the first content resource of the object circle (also referred to as interest preference score), as described below.

[0275] In mathematical modeling, the essence of solving the problem of selecting the preferred resources (i.e., preferred content resources) of a target audience segment is to find suitable behavioral samples (such as consumed content resources) to estimate the likelihood of the target audience segment (i.e., user segment). The process of selecting the preferred resources of a target audience segment can be equivalent to constructing a distribution under certain conditions to fit the behavioral characteristics (i.e., consumption characteristics) of the target audience segment towards content resources. The corresponding business requirement for this mathematical interpretation can be understood as: finding a specific set of consumed content (such as a sequence of preferred resources) for the target audience segment that can represent the characteristics of this target audience segment. The process of solving the maximum likelihood can be equivalent to finding the minimum KL divergence (also known as relative entropy, a measure of the difference between two distributions).

[0276] Therefore, we can denote any content resource in the preference resource sequence of the object circle (which can be represented as T) as x, and the probability of consuming this content resource x within the object circle can be represented as the parameter q(T, x), and the probability of consuming this content resource x within the global object set (which can be represented as G) can be represented as the parameter p(G, x). Thus, the KL divergence D between the consumption probability distribution of content resources in the preference resource sequence of the object circle within the object circle and the consumption probability distribution of each content resource within the global object set is... KL (q‖p) can be:

[0277] It should be noted that the parameters q(T, x) and p(G, x) in the above formula (4) can both be ideal, unmeasurable abstract values. This application only uses the semantics of these parameters q(T, x) and p(G, x) to interpret the validity of the interest preference degree provided by this application. (See the formula in above formula (4)). In this context, log q(T, x) can refer to (e.g., characterize) the numerator of the interest preference for content resources in this application (i.e., the part corresponding to the object circle). In this context, log p(G, x) can refer to the denominator of the interest preference degree of the content resources in this application (i.e., the part corresponding to the global object set), therefore this formula... This can be used to refer to the sum of interest preferences for each content resource in a sequence of preferred resources within an object's social circle (i.e., the summation value, which is the aforementioned interest characteristic value). KL divergence D... KL (q‖p) can characterize the difference between the consumption behavior of each content resource (i.e., each content resource x) in the object circle's preference resource sequence within the object circle (which can be understood as the distribution of each x's q(T, x)) and the consumption behavior of each content resource within the global object set (which can be understood as the distribution of each x's p(G, x)), which is the formula in the above formula (4). The difference between the numerator and denominator in the formula.

[0278] Therefore, combining the above formula (3), we can give the following conclusions regarding the circle likelihood preference score (i.e., interest preference score):

[0279] Where ∝ represents proportional to, ≈ represents approximately equal to, CP(T, x) represents the interest preference score of content resource x in the preference resource sequence of object circle T, and ∑ x CP(T, x) represents the sum of the interest preference scores of each content resource in the preference resource sequence of the object circle (i.e., the interest characteristic value of the object circle mentioned above), vv circle (T, x) represents the first consumption data of content resource x within the object circle T, vv global (G, x) represents the second consumed data of content resource x within the global object set G. KL (q‖p) represents D KL The minimum value of (q‖p), the ∑ x CP(T, x) approximates the expression in formula (4) above. And approximately equal to D KL The minimum value of (q‖p) (because this expression) No matter what, it will be less than or equal to D. KL (q‖p), therefore, it can be D. KL The minimum value of (q‖p), i.e., ∑ x CP(T, x) is the D KL The lower bound of (q‖p). Therefore, ∑ x The larger CP(T, x) is, the larger D is. KL The larger (q‖p) is, the greater the difference between the consumption behavior of content resources in the object's preferred resource sequence within that object's object and its consumption behavior in the global object set; conversely, the smaller (q‖p) is, the greater the difference between the consumption behavior and the consumption behavior within the global object set. x The smaller CP(T, x) is, the better D KL The smaller (q‖p) is, the smaller the difference between the consumption behavior of content resources in the object sphere's preferred resource sequence within that object sphere and the consumption behavior within the global object set.

[0280] In other words, by summing the CP(T, x) of each content resource in the preference resource sequence of the object layer, this application is actually estimating the lower bound between the probability distribution of each content resource being consumed within the object layer and the probability distribution of each content resource being consumed within the global object set.

[0281] Therefore, this application can calculate the interest characteristic value of an object circle through the preference resource sequence of the object circle. The larger the interest characteristic value, the greater the difference between the consumption behavior of objects in the object circle for the preference resource sequence and the consumption behavior of objects in the global object set for the preference resource sequence. The more distinctive the object circle is, the more suitable it is for recommending content resources using content recommendation strategies specific to the object circle (such as strategies based on preference resource sequences). Conversely, the smaller the interest characteristic value, the smaller the difference between the consumption behavior of objects in the object circle for the preference resource sequence and the consumption behavior of objects in the global object set for the preference resource sequence. The less distinctive the object circle is, the less suitable it is for recommending content resources using content recommendation strategies specific to the object circle (such as strategies based on preference resource sequences).

[0282] Using the method described above in this application, the effectiveness of recommending content resources to objects within an object circle can be explained and proven through mathematical derivation, based on the interest and preference of content resources consumed by the object circle within that object circle.

[0283] In this application, by constructing a preference resource sequence for object circles, the object circles gain interpretability: for example, we can clearly and convincingly determine the commonalities of the groups within the object circles through this sequence, such as the objects in the object circles being mostly employees of a certain company, or mostly from a certain industry, and so on. Furthermore, by constructing a preference resource sequence for object circles, the object circles gain measurability: that is, the preference resource sequence provides a measurable basis for filtering or merging object circles. In addition, by constructing a preference resource sequence for object circles, the object circles also gain dynamic timeliness: by dynamically acquiring and updating the preference resource sequence according to a target period, the object circles can dynamically switch between an application state (i.e., an unfrozen state, also known as a usage state) and a frozen state.

[0284] This application innovatively proposes a cross-application domain object behavior preference evaluation system, aiming to solve key problems in large-scale object circle applications, including circle effectiveness assessment, circle similarity analysis, circle merging condition determination, and circle content matching (such as matching content resources preferred by object circles). Furthermore, this application mines circle-specific content (such as content resources in the object circle's preference resource sequence) by comparing the behavioral tendencies of users within a circle with the behavioral distribution of global users. By analyzing the similarity of characteristic content (such as the similarity between preference resource sequences of different object circles), the necessity of merging object circles is determined; simultaneously, based on the magnitude of preference-characteristic content (which can be reflected by interest characteristic values), the effectiveness of object circles in the current application scenario (such as resource recommendation scenario) is evaluated, providing the application end (such as the resource recommendation end) with more accurate and purer circle data (including object circles and the content resources preferred by object circles), thereby improving application efficiency and effectiveness.

[0285] In this application, object circles can be constructed and diffused through social relationships between objects. Therefore, the object circles obtained in this application can also be called social circles. The specific diffusion logic is as follows: a first object is selected from the global object set as a seed object, and a second object with a social relationship with the first object is obtained. Based on the first relationship community to which the first object belongs and the second relationship community to which the second object belongs, a homogeneous relationship community of the first relationship community is selected from at least one second relationship community of the second object. The first relationship community and the homogeneous relationship community are merged to obtain the object circle. The basis is the social hierarchy; the smaller the social hierarchy with the first object, the closer the social relationship. By setting different social hierarchy ranges (such as M-level social hierarchy), the diffusion range and the density of objects within the object circle are controlled. The smaller the M value, the closer the social relationships within the diffused object circle, but the lower the coverage; the larger the M value, the higher the diffusion efficiency, but the slightly weaker the social relationships between objects. The seed object refers to the object selected from the global object set as the starting point for object circle diffusion. Through the social relationships of the seed object, a larger object circle is gradually formed.

[0286] Please refer to Figure 19, which is a flowchart illustrating a method for constructing object layers according to an embodiment of this application.

[0287] As shown in Figure 19, the method may include:

[0288] Step S1901: Select a first object from the global object set and obtain a second object that has a social relationship with the first object.

[0289] Specifically, the resource processing device can select a first object from the global object set. This first object can be a seed object used for subsequent object layer diffusion. There can be one or more first objects, i.e., there can be one or more seed objects. Since the principle of object layer diffusion through each first object is the same, the following explanation can be based on the process of object layer diffusion through a single first object.

[0290] When selecting the first object, if it is the first object for the first round, the device can randomly select objects from the global object set whose number of direct social objects (objects with direct social relationships, such as friends) is greater than or equal to a set second threshold, based on a set selection quantity. This second threshold can be set according to the actual application scenario, such as 40. If it is the first object for rounds after the first round, the processing device can obtain the currently diffused object circles and filter at least one candidate object from the global object set based on these circles. Specifically, the processing device obtains the circle number of each object in the global object set within the diffused object circles. Objects with a circle number less than H are designated as intermediate objects. Then, intermediate objects whose number of direct social objects is greater than or equal to the set second threshold are designated as selected candidate objects. Finally, the processing device randomly selects the first object from these at least one candidate object according to the set selection quantity. If the number of candidate objects is less than the selection quantity, the number of selected first objects equals the number of candidate objects. H is a positive integer, and its specific value can be set according to the actual application scenario. Candidate objects are a set of objects that meet certain conditions and are selected from the global object set during the process of selecting the first object (seed object). Subsequently, these candidate objects will be randomly selected as the first object in the next round.

[0291] In this application, the first object can be selected in one or more rounds. After each round of diffusion of the first object selected, if the resulting object layers do not cover all or most of the objects in the global object set (such as the target proportion set for the objects in the global object set, such as 90%), the first object of the next round can be selected to continue to diffuse more object layers through the selected first object, so that the diffused object layers can cover (i.e. contain) more objects.

[0292] The objects in the global object set can be users in a social application. This social application can be any application where users can have social relationships, such as a web application, a mini-program, or an app. Users within this social application can have direct social relationships, such as being friends or following each other. For example, users A and B being friends means that both users are in each other's friend list; similarly, users A and B following each other means that both users are in each other's following list. The type of this direct social relationship (such as the type of friend relationship or the type of following relationship) can be determined based on the actual application scenario; this application does not restrict the type of direct social relationship.

[0293] Therefore, if this is the selection of the first object in the first round, the first object can be randomly selected from the global object set by setting a selection quantity. The number of selected first objects is equal to this selection quantity, which can be the set number of first objects selected in one round. Therefore, this selection quantity can also be called the unit quantity. Alternatively, to ensure the effectiveness of subsequent object circle diffusion, this application can also select objects with a large number of direct social objects (objects with direct social relationships, such as friends) as objects used for object circle diffusion. Therefore, this application can also randomly select objects from the global object set whose number of social objects is greater than or equal to a set second quantity threshold by setting the selection quantity, as the first object. The second quantity threshold can be set according to the actual application scenario, such as setting it to 40. Specifically, the number of direct social objects of each object in the global object set can be counted, and objects with a number greater than or equal to the set second quantity threshold can be filtered out. Then, a random number generator can be used to randomly select objects corresponding to the set selection quantity from these filtered objects as the first object.

[0294] If the first object is selected in a round after the first round, the processing device can obtain the currently diffused object layer, which can include the various object layers that have been diffused through the first object selected in the previous rounds.

[0295] The processing device can filter at least one candidate object from the global object set using the diffused object layers. This process may include: the processing device obtaining the number of object layers in the diffused object layers for each object in the global object set, that is, counting the number of object layers in the currently diffused object layers that contain each object. An object can be counted as having a number of object layers, which is the number of object layers in the diffused object layers that contain that object.

[0296] In real-world social scenarios, the more times a user appears in the expanded object circles (i.e., the number of object circles they belong to), the higher their social relevance within those circles. Conversely, the fewer times a user appears in the expanded object circles (i.e., the number of object circles they belong to), the lower their social relevance within those circles. Therefore, the processing device can use objects with fewer than H circles as intermediate objects for filtering. These intermediate objects are considered to have lower social relevance within the currently expanded object circles. H is a positive integer, and its specific value can be set according to the actual application scenario. The hit count refers to the number of times an object appears in the expanded object circles, i.e., the number of object circles it belongs to. The hit count reflects the social relevance of an object within each object circle; the higher the hit count, the stronger the social relevance.

[0297] Therefore, the resource processing device can select intermediate objects as candidate objects if the number of objects with direct social relationships is greater than or equal to the aforementioned second threshold. There can be one or more such candidate objects. The direct social relationships of an intermediate object are those objects with which the intermediate object has a direct social relationship.

[0298] The processing device can randomly select a first object from the at least one candidate object according to the set selection quantity, and the number of selected first objects can be equal to the selection quantity. If the number of candidate objects is less than the selection quantity, the number of selected first objects can be equal to the number of candidate objects, that is, each candidate object can be treated as a first object for object layer diffusion.

[0299] In one implementation, if this is selecting the first object in the first round, the first object can also be selected using the method described above for selecting candidate objects through the already diffused object layers. However, in this case, there can be zero already diffused object layers, meaning there are no already diffused object layers yet. In this case, all objects in the global object set can be considered as candidate objects before selecting the first object.

[0300] The aforementioned diffusion of object circles can refer to continuously expanding to more objects, thereby creating an object circle composed of multiple objects. This application can continuously expand and strengthen the object circle through the relational communities of objects, as detailed in the descriptions of the following embodiments.

[0301] Furthermore, the processing device can also acquire a second object that has a social relationship with the first object. The specific definition of this social relationship, and which objects with social relationships with the first object are included in the second object, can be found in the relevant description in step S1902 below. This second object can be an object that originates from the first object and undergoes diffusion and merging of relationship communities to obtain object circles.

[0302] There may be one or more second objects. Since the principle of processing each second object is the same based on the first object, the following can be explained using one second object as an example, as described below.

[0303] Step S1902: Obtain the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object.

[0304] Specifically, the processing device can acquire a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs. The first relationship community can be used to describe the first object and other objects with social relationships to the first object. The second object can belong to any of the other objects described by the first relationship community; that is, the first relationship community contains the second object. Similarly, the second relationship community can be used to describe the second object and other objects with social relationships to the second object. In other words, the first relationship community can be used to describe the objects it contains; the objects described by the first relationship community are the objects contained within the first relationship community; and the first relationship community can be used to describe the social relationships between the objects it contains. The second relationship community is similar; it can also be used to describe the social relationships between the objects it contains.

[0305] The process by which the processing device obtains the first relational community to which the first object belongs may include: the processing device obtaining the object relational network of the first object, which may contain various objects that have social relationships with the first object, centered on the first object. The object relational network of the first object may also be referred to as the first object's personal social network (ego-network).

[0306] This object relationship network can contain nodes and edges. A node represents an object, and an edge represents a direct social relationship between objects. Specifically, if two objects have a direct social relationship, their nodes can be connected by an edge. The object relationship network of the first object can contain nodes of the first object and nodes of objects that have social relationships with the first object. These social relationships can include, but are not limited to, direct social relationships. Edges are elements in the network structure used to represent direct social relationships between objects. In both the object relationship network and the global relationship network, if there is an edge between two object nodes, it indicates that these two objects have a direct social connection.

[0307] The processing device can partition (or split) the object relationship network centered on the first object by analyzing the connections between nodes in the first object's object relationship network. This partitioning yields at least one relationship community for the first object. Nodes with closer connections are more likely to be assigned to the same relationship community of the first object. Each relationship community of the first object contains nodes of the first object, but different relationship communities of the first object may not contain overlapping nodes of other objects. In other words, an object (not the first object) with a social relationship with the first object will not exist in multiple relationship communities of the first object simultaneously, but rather in only one of the first object's relationship communities.

[0308] In one implementation, the processing device can employ the fast-unfolding (Louvain) method (an unsupervised community detection algorithm based on modularity optimization) to partition the object relation network of the first object into communities, thereby obtaining at least one relation community for the first object. Specifically, each node in the object relation network is first initialized as an independent community. Then, the modularity Q is calculated, where Q is calculated using the formula: Among them l i Let L represent the number of edges within community i, L represent the total number of edges in the network, and d represent the total number of edges in the network. i Let represent the sum of the degrees of all nodes in community i. Then, for each node, try moving it to an adjacent community, calculate the change in modularity ΔQ after the move. If ΔQ > 0, move the node to the adjacent community that maximizes ΔQ. Repeat this step until moving any node no longer increases the modularity. After that, merge each community into a new node, construct a new network, and repeat the above process again until the modularity no longer increases. Finally, complete the community partitioning and obtain at least one relational community for the first object.

[0309] This fast-unfolding method can be used to mine tightly connected groups of nodes (i.e., "communities," referred to as relational communities in this application) from complex social networks. Its core idea is to maximize the optimization of the network's grouping structure through modularity, achieving linear time complexity and making it suitable for handling large-scale social networks. Modularity is a metric calculated based on the connectivity between network nodes, used to measure the quality of community partitioning in the network and reflecting the tightness between node groups. By maximizing modularity, the network grouping structure can be optimized, thereby discovering and partitioning relational communities composed of tightly connected nodes. Specifically, modularity can be calculated based on the connectivity between nodes, reflecting the tightness between the searched node groups. This allows for the merging of highly connected node groups to discover and partition relational communities composed of tightly connected nodes.

[0310] Through the above process, at least one relational community of the first object can be obtained by dividing the object relational network of the first object. The first relational community can be any one of the at least one relational communities divided for the first object. It should be noted that during the diffusion process of object layers, object layers are diffused through each relational community of the first object to obtain one or more object layers. Since the principle of object layer diffusion through each relational community of the first object is the same, this application uses the process of object layer diffusion through any one relational community of the first object (i.e., the first relational community) as an example to explain the relevant process in detail.

[0311] In one embodiment, the processing device can also acquire the object relationship network of the second object, and can divide the object relationship network of the second object with the second object as the center, according to the same principle as the division of the relationship community of the first object, to obtain at least one relationship community of the second object. This at least one relationship community can be referred to as at least one second relationship community of the second object. That is, the at least one second relationship community includes all the relationship communities obtained by dividing the object relationship network of the second object.

[0312] In this context, the object relationship network of an object (such as the first object or the second object) can be a local relationship network within the global relationship network of all objects in the global object set; that is, the object relationship network of an object can be a part of the global relationship network. This global relationship network can be constructed through edges connecting the nodes of various objects in the global object set and the nodes of objects with direct social relationships. This global relationship network can be used to represent the social connections among all users in the global object set.

[0313] In this application, the definition of the object relation network for each object is the same, that is, the principle of obtaining the object relation network of each object from the global relation network is the same. Therefore, the definition of the object relation network is explained here by taking the object relation network of the first object as an example, as described below.

[0314] The nodes in the object relationship network of the first object include the nodes of the first object and the nodes of objects that have social relationships with the first object. The objects that have social relationships with the first object may include: objects whose social hierarchy with the first object is within the Nth social hierarchy, that is, objects whose social hierarchy with the first object is from the 1st social hierarchy to the Nth social hierarchy.

[0315] Here, an object with a social hierarchy of level i to the first object refers to an object in the global relationship network whose shortest link to the first object contains i edges, where i is less than or equal to N and is a positive integer. An object with a social hierarchy of level 1 to the first object refers to an object with a direct social relationship to the first object, that is, an object represented by a node in the global relationship network that is directly connected to a node of the first object.

[0316] In a global relational network, the shortest link from one object to another is the shortest path from the nodes of one object to the other object based on the edges between the nodes. This shortest link can be formed by the nodes and edges that are traversed from the nodes of one object to the nodes of the other object with the fewest steps.

[0317] Based on the above description, for example, if the direct social relationship is a friend relationship, then the relationship between the first object's friend (i.e., the object with a direct social relationship with the first object) and the first object is the first level of social hierarchy; the relationship between the first object's friend's friend and the first object is the second level of social hierarchy; the relationship between the first object's friend's friend's friend and the first object is the third level of social hierarchy, and so on. The object with a social hierarchy of the i-th level with the first object can also be called the first object's i-th degree friend.

[0318] Therefore, the second object can include: objects within the first relationship community that have a social hierarchy of M levels with the first object, where M is a positive integer and less than or equal to N. That is, M can be any positive integer within the range of N. If M equals 1, it indicates that the second object includes objects within the first relationship community that have a direct social relationship with the first object; if M equals 2, it indicates that the second object includes objects within the first relationship community that have a first-level social hierarchy with the first object, and objects that have a second-level social hierarchy with the first object; and so on.

[0319] In this context, a smaller social hierarchy level between the user and the first object indicates a closer social connection, while a larger social hierarchy level indicates a more distant social connection. Therefore, a smaller M results in higher accuracy in subsequent object-based diffusion based on the first object, closer social connections between objects within the resulting object layer, and lower coverage (i.e., fewer objects diffused within the object layer). Conversely, a larger M results in lower accuracy in subsequent object-based diffusion based on the first object, weaker social connections between objects within the resulting object layer, but higher coverage (i.e., more objects diffused within the object layer) and higher diffusion efficiency. Therefore, the value of M can be flexibly set according to actual application needs. M can be any positive integer less than or equal to N, and M is an adjustable value.

[0320] Step S1903: Select a homogeneous relation community of the first relation community from at least one second relation community; the first object belongs to other objects described by the homogeneous relation community.

[0321] Specifically, the processing device can select a homogeneous relation community of the first relation community from at least one second relation community of the second object. The homogeneous relation community can be a second relation community in which the at least one second relation community contains the first object. That is, the first object can be another object described by the homogeneous relation community. In other words, the first object is an object among the other objects described by the homogeneous relation community. The homogeneous relation community contains the first object (i.e., a node containing the first object).

[0322] In practical applications, if the first object and the second object are colleagues, the first relationship community of the first object that includes the second object can be the colleague community of the first object, and the second relationship community of the second object that includes the first object can also be the colleague community of the second object. Therefore, merging the first relationship community that includes the second object and the second relationship community that includes the first object can yield a larger colleague circle (an object circle belonging to the colleague type) of the first object and the second object.

[0323] As can be seen from the above, the first relational community of the first object includes the second object, and the homogeneous relational community of the first relational community selected from one or more second relational communities of the second object also includes the first object. In other words, the relational communities of different objects that mutually include each other are likely to have the same community attributes (i.e., communities of the same nature, such as colleagues, or hometown, classmates, etc.). These relational communities belong to each other's homogeneous relational communities. This application can select the homogeneous relational communities of a certain relational community (such as the first relational community) through the social association between objects.

[0324] Step S1904: Merge the first relationship community with the homogeneous relationship community to obtain the object circle based on the first object.

[0325] Specifically, the processing device can merge the first relation community of the first object and the homogeneous relation community selected by the second object for the first relation community to obtain an object circle based on the first object, which may include the first relation community and the homogeneous relation community.

[0326] There can be many second objects within the first relation community. Therefore, the processing device can select each second object's homogeneous relation community from the second relation community of each second object according to the above principle. The first relation community can be merged with each homogeneous relation community of the first relation community (including the homogeneous relation communities of the first relation community selected from the second relation community of each second object) to obtain an object circle that is finally diffused through the first relation community of the first object.

[0327] For example, the second objects within the first relational community include objects 1, 2, and 3. The processing device can select homogeneous relational communities of the first relational community from at least one relational community of object 1, from at least one relational community of object 2, and from at least one relational community of object 3. Thus, the processing device can merge the first relational community, the homogeneous relational communities of the first relational community selected from the relational communities of object 1, the homogeneous relational communities of the first relational community selected from the relational communities of object 2, and the homogeneous relational communities of the first relational community selected from the relational communities of object 3, to obtain an object layer diffused through the first relational community of the first object.

[0328] Furthermore, since the first object can have multiple relational communities, the processing device can, according to the principle described above (i.e., the principle of object layer diffusion through the first relational community), diffuse object layers through each relational community of the first object separately, thereby obtaining object layers diffused through each relational community of the first object separately. One object layer can be obtained through diffusion from one relational community of the first object. Therefore, the number of object layers that can ultimately be obtained based on the number of relational communities of the first object is directly proportional to the number of relational communities it has.

[0329] Furthermore, the processing device can acquire at least one object circle derived from the first object, which may include various object circles derived from the various relational communities of the first object. The processing device can further filter the at least one object circle to obtain a filtered target object circle. This target object circle may be an object circle derived from the first object and ultimately put into use (i.e., adopted). There may be one or more target object circles. The process of filtering the target object circle can be described as follows.

[0330] The processing device can optimize each of the at least one object layer to obtain at least one optimized object layer, wherein one of the at least one object layers can be optimized to obtain a corresponding optimized object layer.

[0331] Any one of the at least one object layer can be called a candidate object layer. Since the principle of optimizing each object layer to obtain the corresponding optimized object layer is the same, the process of optimizing the candidate object layer to obtain the corresponding optimized object layer will be used as an example for specific explanation.

[0332] The candidate object layer can be obtained by merging E relational communities. These E relational communities can include one relational community of the first object and homogeneous relational communities of the selected relational community. E is an integer greater than 1, and the specific value of E can be determined according to the actual application scenario. The processing device can count the number of communities in the E relational communities that each node in the candidate object layer belongs to. The number of communities corresponding to a node can be counted, which is the number of relational communities of that node included in the E relational communities.

[0333] The more times a node (or object) in a candidate object circle hits (i.e., belongs to one of the E relational communities used to merge and obtain the candidate object circle), the higher its social closeness within that candidate object circle; conversely, the fewer times a node (or object) in a candidate object circle hits (i.e., belongs to one of the E relational communities used to merge and obtain the candidate object circle), the lower its social closeness within that candidate object circle. Therefore, the processing device can remove nodes in the candidate object circle whose corresponding number of communities (i.e., the number of relational communities they belong to) is less than R, to obtain the optimized object circle corresponding to the candidate object circle. This ensures a high confidence that the users retained in the optimized object circle belong to the same object circle. R is a positive integer, and R is less than E. The value of R can be set according to the actual application scenario, such as E / 7. The objects represented by nodes whose corresponding number of communities is less than R can be considered as objects with weak social connections within the candidate object circle. Hit count refers to the number of times an object appears in multiple relationship communities used to merge and form candidate object circles. Hit count measures the social closeness of the object within the candidate object circles; a higher hit count indicates a higher degree of closeness.

[0334] The processing device can obtain the optimized object layer corresponding to each object layer obtained based on the diffusion of the first object, i.e., at least one optimized object layer, according to the above principle.

[0335] The processing device can select the final target object layer from the at least one optimized object layer. This process may include: the processing device can obtain the number of nodes contained in each optimized object layer. An optimized object layer can obtain a number of nodes, which is the number of nodes contained in the optimized object layer.

[0336] This application can also remove excessively small object concentric circles obtained from diffusion. Therefore, the processing device can perform a rejection process (i.e., removal process) on optimized object concentric circles whose corresponding number of nodes is less than or equal to a set first quantity threshold, to obtain the target object concentric circle based on the first object. The number of nodes contained in the target object concentric circle can be greater than the first quantity threshold. The first quantity threshold can be set according to the actual application scenario, such as setting the first quantity threshold to 50, etc.

[0337] The processing device can follow the principle described above (i.e., the principle of obtaining the target object circle based on the first object) to finally diffuse, optimize and filter the target object circle based on each seed object (i.e. each first object). This target object circle is the final object circle in which the objects in the global object set are located, which is the object circle obtained in step S1301 above.

[0338] Please refer to Figure 2, which is a schematic diagram of a scenario for constructing an object hierarchy according to an embodiment of this application. As shown in Figure 2, the global object set can contain a large number of objects. The resource processing device can select a first object from the global object set and obtain a second object that has a social relationship with the first object. The resource processing device can obtain the first relationship community to which the first object belongs (i.e., where it is located), and at least one (or more) second relationship communities to which the second object belongs. The first relationship community can be used to describe the first object and other objects that have a social relationship with the first object, and these other objects may include the second object; and the second relationship community can be used to describe the second object and other objects that have a social relationship with the second object.

[0339] The resource processing device can select a homogeneous relation community of the first object's first relation community from at least one second relation community of the second object, wherein other objects described by this homogeneous relation community include the first object. That is, the second relation community of the first object can be used as a homogeneous relation community of the first object's first relation community to describe objects of the second object. The resource processing device can merge the first relation community and its homogeneous relation communities to obtain an object layer based on the first object. This object layer can simultaneously include all objects described by the first object's first relation community and all objects described by its homogeneous relation communities.

[0340] Furthermore, the methods described above in this application are not only applicable to social circles built through social relationships between users, but also, in practical application scenarios, to various circles built through commonalities among object attributes. For example, age circles (i.e., object circles divided by the age group of the object, where objects of the same age group can be grouped into the same object circle), or regional circles (e.g., object circles divided by the region of the object, where objects of the same region can be grouped into the same object circle), or abstract concepts of user groups built through other unified signals (such as some commonality among objects), and so on.

[0341] Using the method described in this application, a wide range of target object circles can be obtained by expanding the reach of objects through their social networks. Furthermore, the same object can belong to multiple different target object circles, satisfying the multi-dimensional circle identity of a single object, such as simultaneously belonging to a colleague circle, a classmate circle, and an interest circle. Through the process described above, objects can be rapidly aggregated, thereby efficiently expanding the reach of various object circles.

[0342] This application, through an innovative object-centric behavior evaluation system, effectively addresses key issues in recommendation systems such as accuracy, computational efficiency, and dynamic adjustment, significantly improving user experience and business performance. Furthermore, its broad application scope and flexible scalability provide efficient and accurate solutions for various application scenarios, demonstrating significant technological value.

[0343] Please refer to Figure 20, which is a schematic diagram of the structure of an object processing device provided in an embodiment of this application. As shown in Figure 20, the object processing device 2000 may include: an object selection module 2001, a community acquisition module 2002, a community selection module 2003, and a merging module 2004.

[0344] The object selection module 2001 is used to select a first object from the object set and obtain a second object that has a social relationship with the first object;

[0345] The community acquisition module 2002 is used to acquire the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object;

[0346] Community selection module 2003 is used to select a homogeneous relation community of the first relation community from at least one second relation community; the first object belongs to other objects described by the homogeneous relation community.

[0347] The merging module 2004 is used to merge the first relationship community with the homogeneous relationship community to obtain the object circle based on the first object.

[0348] In one implementation, the process by which the community acquisition module 2002 acquires the first relational community to which the first object belongs includes:

[0349] Obtain the object relationship network of the first object; the object relationship network contains nodes and edges, nodes represent objects, and edges represent direct social relationships between objects. The object relationship network contains the nodes of the first object and the nodes of objects that have social relationships with the first object. Social relationships include direct social relationships.

[0350] The object relation network is divided based on the connection relationships between nodes in the object relation network to obtain at least one relation community of the first object;

[0351] The first relational community is any one of at least one relational community.

[0352] In one implementation, the object relationship network is a local relationship network within the global relationship network of all objects in the object set; objects in the object relationship network that have social associations with the first object include: objects whose social hierarchy with the first object is within N social hierarchies, where N is a positive integer;

[0353] Here, the object with the i-th social level relative to the first object refers to an object in the global relationship network whose shortest link to the first object contains i edges, where i is less than or equal to N and i is a positive integer; and,

[0354] The social level of the object with respect to the first object is the first level of social level, which refers to the object with a direct social relationship with the first object.

[0355] In one implementation, the second object includes: an object in the first relationship community, whose social hierarchy with the first object is within the M-level social hierarchy;

[0356] Where M is a positive integer, and M is less than or equal to N.

[0357] In one embodiment, the object processing apparatus 2000 further includes a filtering module 2005, which is used for:

[0358] Obtain at least one object concentric circle based on the diffusion of the first object;

[0359] At least one object layer is filtered to obtain the filtered target object layer.

[0360] In one implementation, the filtering module 2005 performs filtering processing on at least one object layer to obtain the filtered target object layers, including:

[0361] Each object layer in at least one object layer is optimized to obtain at least one optimized object layer; one object layer is used to optimize to obtain a corresponding optimized object layer.

[0362] Select the target object layer from at least one optimized object layer.

[0363] In one implementation, any one of the at least one object circle is a candidate object circle, which is obtained by merging L relation communities, where L is an integer greater than 1; the filtering module 2005 optimizes each of the at least one object circle to obtain at least one optimized object circle, including:

[0364] Count the number of communities in which each node in the candidate object's circle belongs in the L relationship communities;

[0365] Nodes with fewer than K communities are removed from the candidate object layers to obtain the optimized object layers corresponding to the candidate object layers; K is a positive integer and K is less than L.

[0366] In one implementation, the filtering module 2005 selects a target object layer from at least one optimized object layer in the following manner:

[0367] Obtain the number of nodes contained in each optimization object layer;

[0368] At least one optimization object circle with a number of nodes less than or equal to a set first number threshold is removed to obtain the target object circle.

[0369] Among them, the number of nodes contained in the target object's circle is greater than the first quantity threshold.

[0370] In one implementation, any object in the object set is a target object, and the target object exists within Q target object layers obtained by diffusion. The Q target object layers are object layers optimized based on the object layers obtained by diffusion, and Q is a positive integer.

[0371] The filtering module 2005 is also used for:

[0372] Obtain the relationship weight of the target object in each of the Q target object layers;

[0373] Data is pushed to the target object based on the relationship weight of the target object in each target object circle.

[0374] In one implementation, the first object is a seed object in the target object layer that is diffused based on the first object;

[0375] The filtering module 2005 obtains the relationship weight of a target object in each of the Q target object layers using the following methods:

[0376] Get the link number between the target object and the seed object in the first target object circle. The first target object circle is any target object circle in Q target object circles where the seed object is not the target object.

[0377] The number of links is used as the relation weight of the target object in any target object layer.

[0378] In one implementation, the filtering module 2005 pushes data to the target object based on the relationship weight of the target object in each target object circle, including:

[0379] Obtain multiple hotspot data points for each target object's social circle and the popularity score of each hotspot data point;

[0380] By calculating the relational weight of the target object in each target object circle, the popularity score of the hot data in each target object circle is weighted to obtain the weighted popularity score of each hot data.

[0381] Data is pushed to the target audience based on the weighted popularity score of each hot data point.

[0382] In one implementation, the filtering module 2005 pushes data to the target object based on the weighted popularity score of each hot data point, including:

[0383] Sort each hot data point in descending order of its weighted popularity score to obtain the sorted hot data.

[0384] Push the top T hot data points from the sorted hot data to the target object, where T is a positive integer.

[0385] In one implementation, the object selection module 2001 selects a first object from the object set in the following manner:

[0386] Obtain the object layers that have been diffused;

[0387] Based on the already diffused object layers, at least one candidate object is selected from the object set.

[0388] The first object is randomly selected from at least one candidate object according to the set selection number.

[0389] In one implementation, there are G object layers that have been diffused, where G is a non-negative integer; the object selection module 2001 selects at least one candidate object from the object set based on the diffused object layers, including:

[0390] Get the number of object layers that each object in the object set belongs to in the G already diffused object layers;

[0391] Objects with a corresponding number of circles less than H are selected as intermediate objects; H is a positive integer and H is less than G;

[0392] Intermediate objects whose number of direct social objects is greater than or equal to a set second threshold are selected as candidate objects; the direct social objects of an intermediate object are objects with which the intermediate object has a direct social relationship.

[0393] According to one embodiment of this application, the steps involved in the object processing method shown in FIG4 can be executed by various modules in the object processing apparatus 2000 shown in FIG20. For example, step S401 shown in FIG4 can be executed by the object selection module 2001 in FIG20; step S402 shown in FIG4 can be executed by the community acquisition module 2002 in FIG20; step S403 shown in FIG4 can be executed by the community selection module 2003 in FIG20; and step S404 shown in FIG4 can be executed by the merging module 2004 in FIG20.

[0394] This application can select a first object from an object set and obtain a second object that has a social relationship with the first object; it can also obtain a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have a social relationship with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have a social relationship with the second object; this application can select a homogeneous relationship community of the first relationship community from the at least one second relationship community; the first object belongs to other objects described by the homogeneous relationship community; thus, the first relationship community and the homogeneous relationship community can be merged to obtain an object circle based on the first object. Therefore, the device proposed in this application can merge the relationship communities of the first object (such as the first relationship community) and the relationship communities of the second object (such as the selected homogeneous relationship community) through the social association between objects (such as the first object and the second object) to efficiently obtain the object circle based on the first object, thereby improving the efficiency of obtaining the object circle. Furthermore, the multiple relationship communities used to merge the object circle (such as the first relationship community and the homogeneous relationship community) also contain the objects to which the other belongs, so that the multiple relationship communities have a close social relationship. This results in a close social relationship within the object circle obtained by merging the multiple relationship communities, thereby improving the accuracy of the obtained object circle.

[0395] According to one embodiment of this application, the modules in the object processing apparatus 2000 shown in FIG20 can be individually or entirely merged into one or more units, or some of the units can be further divided into multiple functionally smaller sub-units to achieve the same operation without affecting the technical effect of the embodiment of this application. The above modules are based on logical function division. In practical applications, the function of one module can also be implemented by multiple units, or the function of multiple modules can be implemented by one unit. In other embodiments of this application, the object processing apparatus 2000 may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0396] Please refer to Figure 21, which is a schematic diagram of the structure of a resource processing device provided in an embodiment of this application. As shown in Figure 21, the resource processing device 2100 may include: a first acquisition module 2101, a selection module 2102, a second acquisition module 2103, and a prediction module 2104.

[0397] The first acquisition module 2101 is used to acquire object layers, which include multiple objects obtained by aggregating objects in the global object set.

[0398] The selection module 2102 is used to select the first content resource from the resource set. The resource set contains multiple content resources, and the first content resource is the content resource that the target circle has historically consumed.

[0399] The second acquisition module 2103 is used to acquire the associated consumption data of the first content resource. The associated consumption data includes the first consumption data generated by objects in the object circle consuming the first content resource, and the second consumption data generated by objects in the global object set consuming the first content resource.

[0400] The prediction module 2104 is used to perform circle preference prediction processing on the first content resource based on the associated consumption data of the first content resource, so as to obtain the interest preference degree of the target circle for the first content resource.

[0401] In one implementation, the object consumes content resources in any of the following ways:

[0402] Likes, shares, favorites, comments, views, and usage are all ways to exchange content resources using common resources.

[0403] The first consumption data and the second consumption data are each one of the following:

[0404] Frequency of likes, frequency of shares, frequency of favorites, frequency of comments, frequency of views, frequency of use, frequency of redemption, and the amount of general resources redeemed.

[0405] In one implementation, the prediction module 2104 performs segmented preference prediction processing on the first content resource based on the associated consumption data of the first content resource, and obtains the interest preference of the target segment on the first content resource in a manner including:

[0406] The ratio between the value of the first consumption data and the value of the second consumption data is calculated as the initial interest preference of the target group for the first content resource.

[0407] The initial interest preference score is optimized to obtain the interest preference score.

[0408] In one implementation, the prediction module 2104 optimizes the initial interest preference level to obtain the interest preference level in the following ways:

[0409] Get the number of objects contained in the object circle;

[0410] The initial interest preference was normalized using the number of objects to obtain the intermediate interest preference.

[0411] The intermediate interest preference scores are smoothed to obtain the interest preference scores.

[0412] The optimization process includes normalization and smoothing.

[0413] In one implementation, the intermediate interest preference includes a first preference parameter of the object circle for the first content resource, and a second preference parameter of the global object set for the first content resource;

[0414] The prediction module 2104 smooths the intermediate interest preference scores to obtain the interest preference scores in the following ways:

[0415] The first preference parameter in the intermediate interest preference score is smoothed to obtain the smoothed first preference parameter; and,

[0416] The second preference parameter in the intermediate interest preference degree is smoothed to obtain the smoothed second preference parameter;

[0417] Among them, the interest preference degree is composed of the smoothed first preference parameter and the smoothed second preference parameter.

[0418] In one implementation, the prediction module 2104 normalizes the initial interest preference using the number of objects to obtain the intermediate interest preference, including:

[0419] Obtain the circle size balance parameters;

[0420] Calculate the product between the circle size balance parameter and the initial interest preference degree;

[0421] The ratio between the product value and the number of objects is used as the intermediate interest preference.

[0422] In one embodiment, there are multiple first content resources, and the device 2100 further includes a construction module 2105, which is used for:

[0423] The first content resource among multiple first content resources whose interest preference degree is greater than or equal to the set preference degree threshold is regarded as the preferred content resource of the target circle.

[0424] Sort at least one preferred content resource according to its interest preference in descending order to obtain the preferred resource sequence of the object circle.

[0425] In one implementation, there are multiple object layers, and any two of the multiple object layers are the first object layer and the second object layer; the construction module 2105 is further configured to:

[0426] Obtain the sequence similarity between the preference resource sequences of the first object layer and the preference resource sequences of the second object layer;

[0427] If the sequence similarity is greater than or equal to the set similarity threshold, then the first object layer and the second object layer are determined to be similar object layers.

[0428] Perform a repetition check on similar first and second object layers;

[0429] If the first object layer and the second object layer are duplicate object layers, then the first object layer and the second object layer are merged.

[0430] In one implementation, the method by which the construction module 2105 performs a repetition determination on similar first object layers and second object layers includes:

[0431] The objects in the first object layer are compared with the objects in the second object layer to obtain the object overlap between the first object layer and the second object layer.

[0432] If the overlap of objects is greater than or equal to the set overlap threshold, then the first object layer and the second object layer are determined to be overlapping object layers.

[0433] If the overlap of objects is less than the overlap threshold, then the first object layer and the second object layer are determined to be non-overlapping object layers.

[0434] In one implementation, the construction module 2105 is further configured to:

[0435] If the number of content resources in the preferred resource sequence is greater than or equal to the set threshold, then the preferred resource sequence is used to recommend content resources to objects in the object circle.

[0436] If the number of content resources in the preferred resource sequence is less than the quantity threshold, the object layer will be frozen.

[0437] In one implementation, the construction module 2105 is further configured to:

[0438] The preferred resource sequence of the object circle is reacquired according to the set target period, and the reacquired preferred resource sequence is used as the updated preferred resource sequence of the object circle.

[0439] If the object circle is in a frozen state, and the number of content resources in the update preference resource sequence is greater than or equal to the quantity threshold, then the frozen object circle will be activated.

[0440] The algorithm updates the preferred resource sequence and recommends content resources to objects in the activated object sphere.

[0441] In one implementation, the construction module 2105 is further configured to:

[0442] The interest preference scores of the content resources in the preference resource sequence are summed to obtain the interest characteristic values ​​of the object circle.

[0443] If the interest characteristic value is greater than or equal to the set numerical threshold, the preferred resource sequence is used to recommend content resources to objects in the object circle.

[0444] If the interest characteristic value is less than the numerical threshold, then based on the global consumption data of objects in the global object set for the content resources in the resource set, the content resources are recommended to objects in the object circle.

[0445] In this context, each object in the global object set has a global consumption data for a content resource in the resource set.

[0446] In one implementation, the construction module 2105 performs content resource recommendation processing on objects in the object hierarchy based on the consumption data of objects in the global object set for content resources in the resource set, including:

[0447] The content resources in the resource set are sorted in descending order of the global consumption data values ​​of each content resource to obtain the sorted content resources.

[0448] The top K content resources from the sorted list are recommended to objects within the target audience's social circle, where K is a positive integer.

[0449] In one implementation, the method by which the selection module 2102 selects a first content resource from the resource set includes:

[0450] Obtain the first consumption frequency of each object in the object hierarchy for each content resource in the resource set; and, each object in the object hierarchy has a first consumption frequency for a single content resource in the resource set; and,

[0451] Get the second consumption frequency of an object in the global object set for each content resource in the resource set. An object in the global object set has a second consumption frequency for a content resource in the resource set.

[0452] Content resources whose first consumption frequency is greater than or equal to a set first frequency threshold and whose second consumption frequency is greater than or equal to a set second frequency threshold are designated as the first content resources.

[0453] According to one embodiment of this application, the steps involved in the resource processing method shown in FIG13 can be executed by various modules in the resource processing apparatus 2100 shown in FIG21. For example, step S1301 shown in FIG13 can be executed by the first acquisition module 2101 in FIG21, step S1302 shown in FIG13 can be executed by the selection module 2102 in FIG21, step S1303 shown in FIG13 can be executed by the second acquisition module 2103 in FIG21, and step S1304 shown in FIG13 can be executed by the prediction module 2104 in FIG21.

[0454] This application can obtain an object circle, which is obtained by aggregating objects in a global object set. The object circle contains multiple aggregated objects. It can also select a first content resource from a resource set, which contains multiple content resources. The first content resource is the content resource that the object circle has historically consumed. Furthermore, it can obtain the associated consumption data of the first content resource, which includes first consumption data generated by objects in the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource. Therefore, based on the associated consumption data of the first content resource, it can perform circle preference prediction processing on the first content resource to obtain the interest preference degree of the object circle for the first content resource. Therefore, the method proposed in this application can obtain the first content resources historically consumed by the target audience's social circle. By combining the first consumption data of objects in the target audience's social circle and the second consumption data of objects in the global object set for the first content resource, the method can comprehensively perform circle preference prediction processing on the first content resource to obtain the accurate interest preference of the target audience's social circle for the first content resource. This interest preference can reflect the target audience's interest preference for the first content resource based on the global object set, thereby improving the accuracy of obtaining the target audience's interest preference for content resources. Furthermore, it can further improve the accuracy of recommending content resources to objects in the target audience's social circle based on the target audience's interest preference for content resources.

[0455] According to one embodiment of this application, the modules in the resource processing device 2100 shown in FIG. 21 can be individually or entirely merged into one or more units, or some of the units can be further divided into multiple functionally smaller sub-units to achieve the same operation without affecting the technical effect of the embodiment of this application. The above modules are based on logical function division. In practical applications, the function of one module can also be implemented by multiple units, or the function of multiple modules can be implemented by one unit. In other embodiments of this application, the resource processing device 2100 may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented by multiple units working together.

[0456] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0457] According to one embodiment of this application, a computer program capable of executing the steps involved in the corresponding methods shown in the various embodiments of this application can be run on a general-purpose computer device (which may include processing elements and storage elements such as a central processing unit (CPU), random access storage medium (RAM), and read-only storage medium (ROM)) to construct the resource processing apparatus 2100 shown in FIG21. The aforementioned computer program can be recorded on a computer-readable recording medium, and can be loaded into and run in the aforementioned computer device via the computer-readable recording medium.

[0458] Please refer to Figure 22, which is a schematic diagram of the structure of a computer device provided in an embodiment of this application. As shown in Figure 22, the computer device 2200 may include: a processor 2201, a network interface 2204, and a memory 2205. Furthermore, in some embodiments, the computer device 2200 may also include: a user interface 2203, and at least one communication bus 2202. The communication bus 2202 is used to implement communication between these components. The user interface 2203 may include a display screen and a keyboard; optionally, the user interface 2203 may also include a standard wired interface or a wireless interface. The network interface 2204 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 2205 may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. Optionally, the memory 2205 may also be at least one storage device located remotely from the aforementioned processor 2201. As shown in Figure 22, the memory 2205, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a device control application program.

[0459] In the computer device 2200 shown in Figure 22, the network interface 2204 provides network communication functions; the user interface 2203 is mainly used to provide an input interface for the user; and the processor 2201 can be used to call the device control application stored in the memory 2205 to achieve:

[0460] Obtain the object layer, which includes multiple objects obtained by aggregating objects in the global object set into layers;

[0461] Select the first content resource from the resource set. The resource set contains multiple content resources. The first content resource is the content resource that the target audience has historically consumed.

[0462] Obtain the associated consumption data of the first content resource. The associated consumption data includes the first consumption data generated by objects in the object circle consuming the first content resource, and the second consumption data generated by objects in the global object set consuming the first content resource.

[0463] Based on the associated consumption data of the first content resource, we perform circle preference prediction processing on the first content resource to obtain the interest and preference of the target circle for the first content resource.

[0464] It should be understood that the computer device 1000 described in the embodiments of this application can execute the resource processing methods described in the various embodiments of this application, and can also execute the object processing device 2000 described in the embodiment corresponding to FIG20 above, or the resource processing device 2100 described in the embodiment corresponding to FIG21, which will not be repeated here. In addition, the beneficial effects of using the same method will not be repeated here.

[0465] Furthermore, it should be noted that this application also provides a computer-readable storage medium storing a computer program. When a processor executes the computer program, it can perform the resource processing methods described in the various embodiments of this application; therefore, these descriptions will not be repeated here. Additionally, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the embodiments of the computer storage medium involved in this application, please refer to the description of the method embodiments of this application.

[0466] As an example, the aforementioned computer program can be deployed and executed on a single computer device, or deployed and executed on multiple computer devices located in one location, or executed on multiple computer devices distributed across multiple locations and interconnected via a communication network. These multiple computer devices distributed across multiple locations and interconnected via a communication network can form a blockchain network.

[0467] The aforementioned computer-readable storage medium can be an internal storage unit of the computer device, such as a hard drive or memory. It can also be an external storage device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, or flash card. Furthermore, the computer-readable storage medium can include both internal and external storage units of the computer device. This computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. It can also be used to temporarily store data that has been output or will be output.

[0468] This application provides a computer program product comprising a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the resource processing methods described in the embodiments of this application; therefore, these descriptions will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the embodiments of the computer-readable storage medium involved in this application, please refer to the description of the method embodiments of this application.

[0469] In summary, this application provides an object processing method, an object processing apparatus, a computer program product, a computer device, and a computer-readable storage medium. The computer device selects a first object from a set of objects, obtains a second object with which it has a social relationship, and then obtains a first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs. Homogeneous relationship communities from the first relationship community are selected and merged to obtain an object layer diffused based on the first object. This method operates based on social relationships. From a graph theory perspective, by analyzing the connections between objects, relationship communities can be quickly identified and merged. In terms of computational complexity, compared to the traditional method of traversing all objects to find object layers, this method reduces the computational complexity from O(nlogn) to O(klogk) (where n is the total number of objects and k is the number of seed objects selected in this application), greatly improving the efficiency of diffused object layers and reducing computation time and equipment resource consumption. At the same time, because the merged communities contain each other's objects, the social relationships within the object circles are close, which improves the accuracy of object circles and provides a precise target group for subsequent data push and resource recommendation, thereby enhancing the pertinence and effectiveness of resource allocation.

[0470] Furthermore, when obtaining the first relational community to which the first object belongs, the object relational network of the first object is first obtained, which includes nodes and edges. Nodes represent objects, and edges represent direct social relationships between objects. Then, based on the node connections in the object relational network, the network is partitioned to obtain at least one relational community for the first object, and the first relational community is any one of these communities. This partitioning method based on network topology utilizes the connectivity principle of graphs to objectively and efficiently determine relational communities based on direct social relationships between objects. When processing large-scale social networks, the parallelism of graph algorithms can be used to accelerate the process and improve processing efficiency. At the same time, this partitioning method more accurately reflects the social group structure between objects, which helps in subsequent selection of homogeneous relational communities and merging operations. This makes the obtained object circles more consistent with actual social situations, enhances the authenticity and reliability of the object circles, and provides a solid data foundation for subsequent social analysis and applications.

[0471] Furthermore, the object relationship network is a local relationship network within the global relationship network of all objects in the object set. Objects with social connections to the first object are those whose social hierarchy with the first object is within N levels. This hierarchical setting allows for flexible control of the scope of the object relationship network within the graph structure, similar to the depth constraint in a breadth-first search within a graph. In different application scenarios, the value of N can be adjusted according to needs to obtain social network information of varying scales and densities, providing a richer and more diverse data foundation for subsequent segmentation of relationship communities and construction of object circles. This flexibility allows the construction of object circles to adapt to different business needs and scenarios. For example, in a precision marketing scenario, the value of N can be reduced to obtain a tighter social circle, while in an information dissemination scenario, the value of N can be increased to expand the dissemination scope.

[0472] Furthermore, the second object is an object within the first relationship community whose social hierarchy with the first object is within the M-level social hierarchy. The value of M can be flexibly adjusted to define the scope of the second object. In the graph's hierarchical structure, a smaller M value indicates a stronger social connection between the second and first objects. Subsequent object layer diffusion based on the first object is analogous to expanding along tightly connected paths in the graph, resulting in higher accuracy and stronger social connections between objects within the object layer. A larger M value leads to higher diffusion efficiency and greater coverage of the object layer, similar to a broader search within the graph. This adjustability allows the construction of object layers to balance accuracy and coverage according to actual needs, meeting the requirements of different application scenarios, improving the method's adaptability and flexibility, and enabling it to better cope with complex and ever-changing social network environments.

[0473] Furthermore, after obtaining at least one object concentric circle based on the diffusion of the first object, these object concentric circles are filtered to obtain the target object concentric circle. The filtering process includes optimizing each object concentric circle to obtain optimized object concentric circles, and then selecting the target object concentric circle from the optimized object concentric circles. From a data processing perspective, this filtering and optimization process can remove object concentric circles that do not meet the requirements, much like cleaning and filtering data, reducing the workload and data redundancy in subsequent processing. In large-scale data processing, it can significantly improve the efficiency and accuracy of data processing, allowing resources to be more concentrated on valuable object concentric circles, avoiding the impact of invalid data on system performance, and improving the overall operating efficiency of the system.

[0474] Furthermore, when optimizing the object circle, if the candidate object circle is obtained by merging L relational communities, the number of communities each node in the candidate object circle belongs to in the L relational communities is counted. Nodes with fewer than K communities are removed, resulting in the optimized object circle. This method of optimizing the object circle by the number of nodes hitting each relational community is similar to filtering nodes closely connected to core nodes in a graph. It effectively removes objects with weak social connections within the object circle, increases the homogeneity and density of objects within the object circle, ensures the confidence of objects in the optimized object circle, and improves the quality and usability of the object circle. In social analysis and recommendation systems, high-quality object circles can provide more accurate user profiles, thereby improving the accuracy and effectiveness of recommendations.

[0475] Furthermore, when selecting target object circles from the optimized object circles, the number of nodes in each optimized object circle is obtained. Optimized object circles with a node count less than or equal to a set first threshold are removed, resulting in target object circles with a node count greater than the first threshold. This node-based filtering method removes object circles that are too small, preventing unnecessary interference from these small circles in subsequent operations. In resource allocation and recommendation scenarios, this ensures that target object circles have sufficient size and representativeness, improving the effectiveness of resource recommendation and data push. It allows resources to be more effectively allocated to object circles of a certain size and activity level, avoiding resource waste and improving resource utilization.

[0476] Furthermore, given that any target object in the object set exists within Q target object circles, the relationship weight of the target object within each circle is first obtained, and then data is pushed to the target object based on these relationship weights. By incorporating the relationship weights of the target object within each target object circle, the data recommendation algorithm can more accurately understand the closeness between the target object and different object circles, similar to calculating the shortest path or connection strength between nodes in a graph. This closeness allows for data pushes that better match the target object's interests and social background, improving the targeting and effectiveness of the data push. In an era of information overload, this helps users quickly obtain information of interest, enhancing the user experience.

[0477] Furthermore, if the first object is a seed object within the target object circle that it diffuses into, when obtaining the relationship weight of the target object within the target object circle, the number of links between the target object and the seed object within its own first target object circle is obtained and used as the relationship weight. The number of links can intuitively reflect the social distance and closeness between the target object and the seed object. In graph theory, the number of links is similar to the shortest path length between nodes. Using this as the relationship weight is more objective and accurate, providing a more reliable basis for subsequent data push, enabling data push to more accurately personalize recommendations based on the relationship between the target object and the seed object. In personalized recommendation systems, this link-based weight calculation method can better capture the social relationships between users, improving the accuracy and personalization of recommendations.

[0478] Furthermore, when pushing data to a target audience based on its relationship weight within each audience circle, the algorithm first acquires multiple trending data points and popularity scores for each audience circle. These popularity scores are then weighted using relationship weights to obtain a weighted popularity score, which is used for data push. By weighting the popularity scores of trending data, the algorithm comprehensively considers the relationship between the target audience and its audience circle, as well as the popularity of trending data within that circle, similar to weighting different factors in multi-factor decision-making. This ensures that the pushed data not only matches the social circle characteristics of the target audience but also possesses a certain degree of popularity and attractiveness, improving the quality of data push and user engagement. In social media and content recommendation platforms, this can increase user click-through rates and engagement with recommended content.

[0479] Furthermore, when pushing data to the target audience based on weighted popularity scores, the hot data is sorted from largest to smallest according to the weighted popularity score, and the top T hot data points are pushed. This sorting-based push method, in its application of data sorting algorithms, can quickly filter out the hot data that best matches the needs and interests of the target audience, similar to selecting the top K elements in a sorting algorithm. This improves the efficiency and effectiveness of data push, enabling the target audience to obtain the most valuable data in a timely manner, enhancing user experience and resource utilization. In real-time data push scenarios, it can quickly respond and provide valuable information, meeting users' timeliness needs.

[0480] Furthermore, when selecting the first object from the object set, the already diffused object circles are first obtained. Candidate objects are then selected from the object set based on these object circles, and finally, the first object is randomly selected according to a set number. This method of selecting candidate objects based on already diffused object circles ensures that the selected first object has better representativeness and social relevance in the data filtering and sampling algorithms. Just as in sample selection, choosing representative samples for analysis can play a more effective role in subsequent object circle diffusion, improving the efficiency and quality of object circle diffusion and reducing unnecessary trials and calculations. In large-scale social network analysis, it can quickly locate potential seed objects and accelerate the object circle construction process.

[0481] Furthermore, when filtering candidate objects based on the already diffused object circles, the number of circles each object in the object set belongs to within the already diffused object circles is obtained. Objects with fewer than H circles are designated as intermediate objects, and intermediate objects with a number of direct social objects greater than or equal to a set second threshold are designated as candidate objects. Filtering intermediate objects by the number of circles identifies objects with low social relevance within the already diffused object circles. Further filtering by the number of direct social objects ensures that candidate objects have sufficient social connectivity. In graph connectivity analysis, this filtering method can select nodes with high connection potential, providing a more favorable foundation for the subsequent diffusion of object circles and improving the feasibility and effectiveness of object circle diffusion. In social network expansion and information dissemination scenarios, it can quickly identify key dissemination nodes, promoting the widespread dissemination of information.

[0482] Furthermore, this application provides a resource processing method, a resource processing apparatus, a computer program product, a computer device, and a computer-readable storage medium. The computer device acquires object circles obtained through aggregation of object circles in a global object set. It selects first content resources historically consumed by the object circles from the resource set, and obtains associated consumption data containing first consumption data of objects in the object circles and second consumption data of objects in the global object set. Based on the associated consumption data, it performs circle preference prediction processing on the first content resources to obtain interest preference degrees. At the data processing level, this method comprehensively considers the consumption data of both object circles and the global set, avoiding the limitations of a single perspective, similar to fusing data from different dimensions in multi-dimensional data analysis. In terms of algorithm complexity, through a reasonable data structure and calculation method, it can efficiently process large-scale consumption data, providing a multi-dimensional analytical foundation for the calculation of interest preference degrees, making the calculation of interest preference degrees more scientific and accurate. In resource recommendation systems, accurate interest preference degrees can provide precise decision-making basis for resource allocation, improving the overall efficiency and quality of resource processing and reducing resource waste.

[0483] Furthermore, the consumption of content resources by the user encompasses liking, forwarding, collecting, commenting, browsing, using, and exchanging content resources for general resources. The first and second consumption data include the frequency of likes, forwards, collections, comments, views, usage, exchanging, and the amount of general resources exchanged. This multi-dimensional data collection method enriches the information dimensions of the associated consumption data in terms of data feature extraction, much like extracting more features to describe the data in feature engineering. It more comprehensively reflects the user's consumption behavior patterns of content resources, providing richer feature vectors for subsequent data analysis and model building. In machine learning models, rich feature vectors can improve the model's accuracy and generalization ability, thereby improving the precision of interest preference calculation, enhancing the targeting of resource processing, and enabling resource recommendations to better meet the actual needs of users.

[0484] Furthermore, in the circle preference prediction process, the initial interest preference degree is obtained by first calculating the ratio of the first consumption data to the second consumption data, and then optimization processing is performed. Introducing global consumption data as the denominator effectively balances the impact of global popularity on the popularity of content resources within the target circle in the mathematical model, similar to normalization in statistical analysis. This avoids preference judgment bias caused by interference from globally popular content, making the initial interest preference degree more accurately reflect the unique preferences of the target circle, providing a base value closer to the true preferences for subsequent optimization processing. In recommendation algorithms, accurate initial interest preference degree can improve the accuracy and personalization of recommendations, enhancing the accuracy and stability of the entire prediction model.

[0485] Furthermore, when optimizing the initial interest preference, the number of objects in each target audience sphere is first obtained, and normalization is performed to obtain an intermediate interest preference. Then, smoothing is applied to obtain the final interest preference. Normalization, by introducing a balancing parameter between the number of objects and the size of the target audience sphere, eliminates the impact of differences in the size of different target audience spheres at the data standardization level, making the data features of spheres of different sizes comparable, much like a normalization operation in data preprocessing. This provides a unified standard for subsequent data analysis and model training. Smoothing, on the other hand, adjusts the data distribution, narrowing the gap in interest preferences between target audience spheres for top content and other content resources, avoiding excessive data concentration, and making the data distribution more uniform. In machine learning models, a uniform data distribution can improve the model's training effect and generalization ability, enhance the model's ability to identify different types of content resources, and strengthen the fairness and diversity of resource recommendations.

[0486] Furthermore, the intermediate interest preference score includes the preference parameters of the object circle and the global object set for the first content resource. When smoothing these parameters, the two parameters are processed separately, and the interest preference score is composed of the smoothed parameters. Processing the two parameters separately allows for more precise adjustment of the weights of the object circle and the global object set in the interest preference score calculation at the model parameter optimization level, avoiding weight bias caused by uniform processing. This is similar to optimizing different objective functions separately in multi-objective optimization. This makes the final interest preference score more reflective of the characteristics of the object circle, improving the model's targeting and accuracy. In personalized recommendation systems, it can better meet the personalized needs of different object circles.

[0487] Furthermore, when normalizing the initial interest preference using the number of objects, a circle size balance parameter is obtained, and the ratio of the product value to the number of objects is calculated to obtain the intermediate interest preference. The introduction of the circle size balance parameter provides an adjustable mechanism for model flexibility, allowing for flexible adjustment of the normalization degree according to different data distributions and application scenarios, making the normalization process more adaptable to actual situations. Just like adjusting parameters according to different situations in adaptive algorithms, this improves the model's adaptability and generalization ability. In object circles of different sizes and activity levels, it can better balance the characteristics of the data and improve the accuracy of interest preference calculation.

[0488] Furthermore, when there are multiple primary content resources, the primary content resource with an interest preference degree greater than or equal to the preference degree threshold is designated as the preferred content resource, and sorted by interest preference degree to obtain the preferred resource sequence. At the data structure level, the preferred resource sequence structurally integrates multiple preferred content resources, forming an ordered resource set, avoiding the simplistic representation of interest preferences by a single content resource. Similar to building an ordered data structure in data organization, this provides a more efficient data structure for subsequent operations such as circle merging, filtering, and resource recommendation. In data processing and algorithm design, an ordered data structure can improve the efficiency and accuracy of operations, thus enhancing the efficiency and accuracy of data processing.

[0489] Furthermore, when there are multiple object layers, similar object layers are identified by comparing the similarity of the preference resource sequences of the first and second object layers. These similar object layers are then judged for duplication, and merged if they are duplicate layers. In terms of algorithm complexity, determining the similarity and duplication of object layers by comparing the similarity of preference resource sequences avoids the enormous computational burden of directly comparing objects, reducing the number of object layers to be processed to an operational level. Similar to clustering based on feature similarity in clustering algorithms, this reduces the algorithm's time and space complexity, improves data processing efficiency, and makes layer merging operations more feasible. In large-scale object layer processing, it can quickly identify and merge similar layers, improving resource processing efficiency.

[0490] Furthermore, when determining the repetition of similar first and second object layers, the overlap degree is obtained by comparing the objects, and whether a layer is a duplicate is determined based on the overlap degree. In terms of data accuracy, this method based on object overlap, starting from the essential characteristics of objects, can more accurately identify duplicate object layers, avoiding potential misjudgments that might occur if only based on the similarity of preferred resource sequences. Similar to using precise matching rules in data matching, this ensures the accuracy and effectiveness of layer merging, improving the quality of resource processing. In resource integration and management, it can avoid the waste of duplicate resources and improve resource utilization efficiency.

[0491] Furthermore, based on the comparison between the quantity of content resources in the preferred resource sequence and the quantity threshold, the object sphere is either recommended or frozen. At the resource management level, this filtering mechanism effectively filters out object spheres that are ineffective for the current application scenario, reducing the amount of data and computation in the recommendation process and avoiding resource waste on ineffective spheres. Similar to resource filtering and optimization in resource allocation, it improves the efficiency of object sphere management, making resource processing more targeted and efficient. When resources are limited, it can concentrate resources on valuable object spheres.

[0492] Furthermore, the preferred resource sequence of the target audience is reacquired periodically as an updated preferred resource sequence. Frozen target audiences are activated and content resources are recommended when the updated preferred resource sequence meets certain conditions. In terms of data timeliness, this dynamic update and activation mechanism ensures that the preferred resource sequence of the target audience reflects the latest changes in the audience's interests in a timely manner, guaranteeing the timeliness and effectiveness of resource recommendations. Similar to data updates and state transitions in real-time data processing, this avoids resource recommendation lag caused by changes in the activity level of target audiences, improving user experience and resource utilization. In a rapidly changing market environment, it allows for timely adjustments to resource recommendation strategies to meet the latest user needs.

[0493] Furthermore, an interest characteristic value is calculated by summing the interest preferences of content resources in the preferred resource sequence. A recommendation strategy is then selected based on a comparison between the interest characteristic value and a numerical threshold. At the recommendation algorithm optimization level, flexibly selecting a recommendation strategy based on the interest characteristic value better adapts to the interest characteristics of different target audiences. When the interest characteristic value is high, using the preferred resource sequence for recommendation accurately satisfies the unique interests of the target audience; when the interest characteristic value is low, recommendation based on global consumption data provides content resources with broad appeal to the target audience. Similar to selecting the optimal strategy based on different situations in intelligent decision-making, this improves the accuracy and adaptability of the recommendation algorithm, making resource recommendations more closely aligned with the actual interests of the target audience. In personalized recommendation systems, more suitable recommended content can be provided based on the characteristics of different target audiences.

[0494] Furthermore, when making recommendations based on global consumption data of content resources within a resource set, the top K content resources are selected and recommended to objects within the target audience based on the global consumption data values. At the ranking algorithm application level, this ranking recommendation method utilizes ranking algorithms to quickly filter out popular content resources, providing broadly attractive recommendations to the target audience. Similar to ranking and filtering in information retrieval, this improves the efficiency and effectiveness of resource recommendations, enabling resources to be more effectively allocated to the target audience.

[0495] The terms "first," "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.

[0496] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0497] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. An object processing method, executed by a computer device, the method comprising: Select a first object from the object set, and obtain a second object that has a social relationship with the first object; Obtain the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object; Select a homogeneous relation community from the first relation community from the at least one second relation community; The first object belongs to the other objects described in the homogeneous relationship community; and The first relational community and the homogeneous relational community are merged to obtain an object circle based on the first object.

2. The method as described in claim 1, wherein the process of obtaining the first relational community to which the first object belongs includes: Obtain the object relationship network of the first object; The object relationship network includes nodes and edges. The nodes represent objects, and the edges represent direct social relationships between objects. The object relationship network includes the nodes of the first object and the nodes of objects that have social associations with the first object. The social associations include the direct social relationships. Based on the connection relationships between nodes in the object relation network, the object relation network is divided to obtain at least one relation community of the first object; Wherein, the first relational community is any one of the at least one relational community.

3. The method of claim 2, wherein the object relationship network is a local relationship network in a global relationship network of full objects in the object set; and the objects in the object relationship network having a social association relationship with the first object include: The objects whose social hierarchy with the first object is within N social hierarchy levels, where N is a positive integer; Wherein, the object whose social hierarchy with the first object is the i-th social hierarchy refers to: an object in the global relationship network whose shortest link with the first object contains i edges, where i is less than or equal to N, and i is a positive integer; and, An object whose social hierarchy with the first object is the first level of social hierarchy refers to an object that has a direct social relationship with the first object.

4. The method of claim 3, the second object comprising: The objects in the first relationship community whose social hierarchy with the first object is within the M-level social hierarchy; Where M is a positive integer, and M is less than or equal to N.

5. The method according to any one of claims 1 to 4, wherein the method further comprises: Obtain at least one object concentric circle based on the diffusion of the first object; The at least one object layer is filtered to obtain the filtered target object layer.

6. The method as described in claim 5, wherein the step of filtering the at least one object circle to obtain the filtered target object circle includes: Each object layer in the at least one object layer is optimized to obtain at least one optimized object layer; An object layer is used to optimize and obtain a corresponding optimized object layer; The target object layer is selected from the at least one optimized object layer.

7. The method of claim 6, wherein any one of the at least one object circle layer is a candidate object circle layer, and the candidate object circle layer is obtained by merging L relationship communities, L being an integer greater than 1. The optimization process for each object layer in the at least one object layer to obtain at least one optimized object layer includes: Count the number of communities in the L relationship communities that each node in the candidate object layer belongs to; Nodes with fewer than K communities are removed from the candidate object layer to obtain the optimized object layer corresponding to the candidate object layer; K is a positive integer and K is less than L.

8. The method of claim 6 or 7, wherein selecting the target object layer from the at least one optimized object layer comprises: Obtain the number of nodes contained in each optimization object layer; The optimization object layers whose number of corresponding nodes is less than or equal to a set first quantity threshold are removed to obtain the target object layers. Wherein, the number of nodes contained in the target object layer is greater than the first number threshold.

9. The method according to any one of claims 1 to 8, wherein any object in the object set is a target object, the target object exists within Q target object layers obtained by diffusion, and the Q target object layers are object layers optimized based on the object layers obtained by diffusion, where Q is a positive integer; The method further includes: Obtain the relationship weight of the target object in each of the Q target object layers; Data is pushed to the target object based on the relationship weight of the target object in each of the target object's social circles.

10. The method of claim 9, wherein the first object is a seed object in the target object layer diffused based on the first object; The step of obtaining the relationship weight of the target object in each of the Q target object layers includes: Obtain the link number between the target object and the seed object in the first target object circle where the target object is located, where the first target object circle is any target object circle in the Q target object circles where the seed object is not the target object; The number of links is used as the relation weight of the target object in any target object layer.

11. The method as described in claim 9 or 10, wherein pushing data to the target object based on the relationship weight of the target object in each of the target object's social circles comprises: Obtain multiple hotspot data points and the heat score of each hotspot data point for each target object circle. By using the relation weight of the target object in each target object circle, the heat score of the hot data in each target object circle is weighted to obtain the weighted heat score of each hot data. Data is pushed to the target object based on the weighted popularity score of each hot data point.

12. The method of claim 11, wherein pushing data to the target object based on the weighted popularity score of each hot data point includes: The hot data is sorted in descending order of its weighted popularity score to obtain the sorted hot data. The first T hot data points from the sorted hot data are pushed to the target object, where T is a positive integer.

13. The method of any one of claims 1 to 12, wherein selecting the first object from the object set comprises: Obtain the object layers that have been diffused; Based on the already diffused object layer, at least one candidate object is selected from the object set; The first object is randomly selected from the at least one candidate object according to the set selection quantity.

14. The method of claim 13, wherein the diffused object layers number G, where G is a non-negative integer; the step of selecting at least one candidate object from the object set based on the diffused object layers includes: Obtain the number of object layers in which each object in the object set belongs within the G diffused object layers; Objects with a corresponding number of circles less than H are selected as intermediate objects; H is a positive integer and H is less than G; Intermediate objects whose number of direct social objects is greater than or equal to a set second threshold are selected as candidate objects; the direct social objects of the intermediate objects are objects with which the intermediate objects have direct social relationships.

15. A resource processing method, executed by a computer device, the method comprising: Obtain an object layer, wherein the object layer includes multiple objects obtained by aggregating objects in a global object set; Select a first content resource from the resource set, wherein the resource set contains multiple content resources, and the first content resource is a content resource that the target group has historically consumed. Obtain associated consumption data of the first content resource, wherein the associated consumption data includes first consumption data generated by objects in the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource; and Based on the associated consumption data of the first content resource, the first content resource is subjected to circle preference prediction processing to obtain the interest preference degree of the target circle for the first content resource.

16. The method of claim 15, wherein the object consumes content resources in any of the following ways: Likes, shares, favorites, comments, views, and usage are all ways to exchange content resources using common resources. wherein The first consumption data and the second consumption data are each one of the following: Frequency of likes, frequency of shares, frequency of favorites, frequency of comments, frequency of views, frequency of use, frequency of redemption, and the amount of general resources redeemed.

17. The method as described in claim 15 or 16, wherein the step of performing segmented preference prediction processing on the first content resource based on the associated consumption data of the first content resource to obtain the interest preference degree of the target segment for the first content resource includes: The ratio between the value of the first consumption data and the value of the second consumption data is calculated as the initial interest preference of the target group for the first content resource. The initial interest preference score is optimized to obtain the interest preference score.

18. The method of claim 17, wherein optimizing the initial interest preference to obtain the interest preference comprises: Obtain the number of objects contained in the object layer; The initial interest preference is normalized using the number of objects to obtain the intermediate interest preference. The intermediate interest preference score is smoothed to obtain the interest preference score. The optimization process includes the normalization process and the smoothing process.

19. The method of claim 18, wherein the intermediate interest preference degree includes a first preference parameter of the object circle for the first content resource, and a second preference parameter of the global object set for the first content resource; The process of smoothing the intermediate interest preference scores to obtain the interest preference scores includes: The first preference parameter in the intermediate interest preference degree is smoothed to obtain a smoothed first preference parameter; and, The second preference parameter in the intermediate interest preference degree is smoothed to obtain the smoothed second preference parameter; The interest preference degree is composed of the smoothed first preference parameter and the smoothed second preference parameter.

20. The method as described in claim 18 or 19, wherein normalizing the initial interest preference using the number of objects to obtain an intermediate interest preference includes: Obtain the circle size balance parameters; Calculate the product between the circle size balance parameter and the initial interest preference degree; The ratio between the product value and the number of objects is used as the intermediate interest preference degree.

21. The method of any one of claims 15 to 20, wherein the first content resource comprises a plurality of resources, and the method further comprises: The first content resources among the multiple first content resources whose interest preference degree is greater than or equal to the set preference degree threshold are regarded as the preferred content resources of the object circle. The at least one preferred content resource is sorted in descending order of interest preference to obtain the preferred resource sequence of the object circle.

22. The method of claim 21, wherein there are multiple object layers, and any two of the multiple object layers are a first object layer and a second object layer; the method further includes: Obtain the sequence similarity between the preference resource sequence of the first object layer and the preference resource sequence of the second object layer; If the sequence similarity is greater than or equal to the set similarity threshold, then the first object layer and the second object layer are determined to be similar object layers; Repeatability checks are performed on similar first and second object layers; If the first object layer and the second object layer are duplicate object layers, then the first object layer and the second object layer are merged.

23. The method of claim 22, wherein the repetition determination of the similar first object layer and the second object layer includes: The objects in the first object layer and the objects in the second object layer are compared to obtain the object overlap between the first object layer and the second object layer. If the overlap of the objects is greater than or equal to the set overlap threshold, then the first object layer and the second object layer are determined to be overlapping object layers. If the overlap of the objects is less than the overlap threshold, then it is determined that the first object layer and the second object layer are not overlapping object layers.

24. The method of any one of claims 21 to 23, wherein the method further comprises: If the number of content resources in the preferred resource sequence is greater than or equal to a set threshold, then the preferred resource sequence is used to recommend content resources to objects in the object circle. If the number of content resources in the preferred resource sequence is less than the number threshold, the object circle is frozen.

25. The method of claim 24, further comprising: The preferred resource sequence of the object circle is reacquired according to the set target period, and the reacquired preferred resource sequence is used as the updated preferred resource sequence of the object circle. If the object circle is in a frozen state, and the number of content resources in the update preference resource sequence is greater than or equal to the number threshold, then the frozen object circle is activated. The updated preference resource sequence is used to recommend content resources to objects in the activated object circle.

26. The method of any one of claims 21 to 25, wherein the method further comprises: The interest preference degrees of the content resources in the preferred resource sequence are summed to obtain the interest characteristic value of the object circle; If the interest characteristic value is greater than or equal to the set numerical threshold, then the preferred resource sequence is used to recommend content resources to objects in the object circle. If the interest characteristic value is less than the numerical threshold, then based on the global consumption data of the objects in the global object set for the content resources in the resource set, the content resources are recommended to the objects in the object circle. In this context, each object in the global object set has a global consumption data for a content resource in the resource set.

27. The method of claim 26, wherein the step of recommending content resources to objects in the object circle based on the global consumption data of content resources in the resource set by objects in the global object set includes: The content resources in the resource set are sorted in descending order of the global consumption data values ​​of each content resource to obtain sorted content resources. The top K content resources from the sorted content resources are recommended to objects in the object circle, where K is a positive integer.

28. The method of any one of claims 15 to 27, wherein selecting the first content resource from the resource set comprises: Obtain the first consumption frequency of an object in the object circle for each content resource in the resource set, wherein an object in the object circle has a first consumption frequency for a content resource in the resource set; as well as, Obtain the second consumption frequency of an object in the global object set for each content resource in the resource set, wherein an object in the global object set has a second consumption frequency for a content resource in the resource set; The content resources in the resource set whose first consumption frequency is greater than or equal to a set first frequency threshold and whose second consumption frequency is greater than or equal to a set second frequency threshold are designated as the first content resources.

29. An object processing apparatus, characterized by comprising: The device includes: The object selection module is used to select a first object from the object set and obtain a second object that has a social relationship with the first object; The community acquisition module is used to acquire the first relationship community to which the first object belongs and at least one second relationship community to which the second object belongs; the first relationship community is used to describe the first object and other objects that have social relationships with the first object, the second object belongs to other objects described by the first relationship community, and the second relationship community is used to describe the second object and other objects that have social relationships with the second object; The community selection module is used to select a homogeneous relationship community of the first relationship community from the at least one second relationship community; the first object belongs to other objects described by the homogeneous relationship community; and The merging module is used to merge the first relationship community with the homogeneous relationship community to obtain an object circle based on the first object.

30. A resource processing apparatus, the apparatus comprising: The first acquisition module is used to acquire an object layer, which includes multiple objects obtained by aggregating objects in a global object set. The selection module is used to select a first content resource from a resource set, the resource set containing multiple content resources, and the first content resource being content resources that the target circle has historically consumed. The second acquisition module is used to acquire associated consumption data of the first content resource, wherein the associated consumption data includes first consumption data generated by objects in the object circle consuming the first content resource, and second consumption data generated by objects in the global object set consuming the first content resource; and The prediction module is used to perform circle preference prediction processing on the first content resource based on the associated consumption data of the first content resource, so as to obtain the interest preference degree of the target circle towards the first content resource.

31. A computer program product comprising a computer program stored in a computer-readable storage medium, the computer program being adapted to be read and executed by a processor to cause a computer device having the processor to perform the method of any one of claims 1-28.

32. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1-28.

33. A computer-readable storage medium storing a computer program adapted to be loaded by a processor and to perform the steps of the method of any one of claims 1-28.