Object push method, device, electronic device and storage medium

Abstract

The present disclosure relates to an object push method, apparatus, electronic device, and storage medium. The method comprises: determining target associated objects corresponding to a to-be-pushed object based on preset object relationship information and association weight information corresponding to the to-be-pushed object, wherein each to-be-pushed object corresponds to multiple association relationships. Aggregating the target associated objects to obtain target aggregate information corresponding to the to-be-pushed object, and executing a push operation on the to-be-pushed object based on the target aggregate information. This method can fully utilize multiple association relationships and the importance levels corresponding to different association relationships to discover potential association relationships between users, thereby improving the accuracy and effectiveness of object push.

Description

Technical Field

[0001] The present disclosure relates to the field of computer application technology, and in particular to an object push method, device, electronic device, and storage medium. Background Art

[0002] In recent years, with the rise of deep learning, more and more deep learning algorithms have been proposed to replace the original traditional machine learning algorithms. However, since the essential problem of user push is how to identify people that users may know in a complex user relationship network, it is inherently complex user relationships and a large amount of data. As a result, existing technologies have difficulties such as the complexity of user relationships and the large amount of data. In the field of user push, deep learning algorithms are prone to obtaining inaccurate user push information because they cannot fully utilize complex user relationships, thereby reducing the accuracy of user push. Summary of the Invention

[0003] The present disclosure provides an object push method, apparatus, electronic device, and storage medium to at least address the problem in related technologies of low utilization of complex user relationships, resulting in low accuracy of user push. The technical solutions of the present disclosure are as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, a method for pushing an object is provided, the method comprising:

[0005] Determine the target associated object corresponding to the object to be pushed based on preset object relationship information and association weight information corresponding to the object to be pushed, wherein the object relationship information is a graph structure with preset objects as nodes and multiple association relationships corresponding to each object to be pushed as edges, and the association weight information is the degree of association of the multiple association relationships corresponding to the object to be pushed;

[0006] Aggregate the target-related objects to obtain target aggregation information corresponding to the object to be pushed;

[0007] Based on the target aggregation information, a push operation is performed on the object to be pushed.

[0008] As an optional embodiment, the determining the target associated object corresponding to the object to be pushed based on the preset object relationship information and the associated weight information corresponding to the object to be pushed includes:

[0009] Based on the preset object relationship information, performing at least one object association on the object to be pushed to obtain an associated object corresponding to the object to be pushed;

[0010] Based on the association weight information, object sampling is performed on the associated objects corresponding to the object to be pushed to obtain the target associated objects.

[0011] As an optional embodiment, the performing object sampling on the associated objects corresponding to the object to be pushed based on the association weight information to obtain the target associated objects includes:

[0012] Determine an initial sampling object from the associated objects corresponding to the object to be pushed;

[0013] Taking the initial sampling object as the current sampling object;

[0014] Determining current object sampling information corresponding to the current sampling object based on the association weight information;

[0015] Based on the current object sampling information, updating the current sampling object;

[0016] Repeating the iterative operations of determining the current object sampling information corresponding to the current sampling object based on the association weight information to updating the current sampling object based on the current object sampling information until the current object sampling information meets a first preset condition;

[0017] In a case where the current object sampling information satisfies the first preset condition, the current sampling object is updated based on the current object sampling information that satisfies the first preset condition to obtain the target associated object.

[0018] As an optional embodiment, determining the current object sampling information corresponding to the current sampling object based on the association weight information includes:

[0019] Acquire a preset number of sampling objects from the current sampling object;

[0020] A preset number of sampled objects are used as first objects to be processed, and the other sampled objects except the first objects to be processed in the current sampled objects are used as second objects to be processed:

[0021] Repeating the iterative operation until the first object to be processed meets a second preset condition, the iterative operation comprising:

[0022] Determining first association feature information corresponding to the first object to be processed based on the association weight information;

[0023] Determining target association feature information from the first association feature information;

[0024] Determining second association feature information corresponding to a second object to be processed based on the association weight information;

[0025] updating the first to-be-processed object and the second to-be-processed object based on the second association characteristic information and the target association characteristic information;

[0026] The current object sampling information is determined based on the first object to be processed that meets the second preset condition.

[0027] As an optional embodiment, the aggregating the target-related objects to obtain target aggregate information corresponding to the object to be pushed includes:

[0028] The target associated object is input into the associated object aggregation model, and based on the degree of association between the target associated object and the object to be pushed, the target associated object is aggregated at least once to obtain the target aggregation information. The associated object aggregation model is a model for performing aggregation processing on the target associated object.

[0029] As an optional embodiment, inputting the target associated object into an associated object aggregation model, performing at least one aggregation process on the target associated object based on the degree of association between the target associated object and the object to be pushed, and obtaining the target aggregation information includes:

[0030] Inputting the target associated objects into the associated object aggregation model, and in the case of a first aggregation process, determining initial objects to be aggregated from the target associated objects based on the association degree;

[0031] Performing aggregation processing on the initial objects to be aggregated to obtain object aggregation information corresponding to the first aggregation processing;

[0032] In the case of a non-first aggregation process, obtain the previous object aggregation information corresponding to each aggregation process;

[0033] Based on the association degree, determining the to-be-aggregated object corresponding to each aggregation process from the target associated objects;

[0034] Aggregating the object to be aggregated and the previous object aggregation information to obtain object aggregation information corresponding to each aggregation process;

[0035] When the number of aggregation processes meets the preset number, the object aggregation information corresponding to the aggregation processes when the preset number of times is met is used as the target aggregation information.

[0036] As an optional embodiment, the associated object aggregation model includes an aggregation processing layer and a feature perception layer, and the aggregation processing of the to-be-aggregated object and the previous aggregation processing information to obtain the object aggregation information corresponding to each aggregation processing includes:

[0037] Inputting the object to be aggregated and the previous aggregation processing information into the aggregation processing layer to perform feature aggregation to obtain aggregated feature information;

[0038] Inputting the aggregated feature information into the feature perception layer for feature perception to obtain perceived feature information;

[0039] The perception feature information is used as the object aggregation information corresponding to each aggregation process.

[0040] As an optional embodiment, the method further includes:

[0041] Get multiple associations corresponding to each object to be pushed;

[0042] Determine the association weight information corresponding to the multiple association relationships

[0043] Taking the multiple objects to be pushed as nodes and based on the multiple association relationships and the association weight information corresponding to the multiple association relationships, an edge corresponding to the node of each object to be pushed is constructed to obtain the object relationship information.

[0044] As an optional embodiment, performing a push operation on the object to be pushed based on the target aggregate information includes:

[0045] In response to a push request corresponding to a target object, obtaining first target aggregate information corresponding to the target object and second target aggregate information corresponding to other objects among the objects to be pushed except the target object;

[0046] determining a similarity between the first target aggregate information and the second target aggregate information;

[0047] Based on the similarity, determining a target push object corresponding to the target object;

[0048] Push the target push object to the target object.

[0049] According to a second aspect of an embodiment of the present disclosure, there is provided an object pushing device, the device comprising:

[0050] a target-associated object determination module configured to determine a target-associated object corresponding to an object to be pushed based on preset object relationship information and association weight information corresponding to the object to be pushed, wherein the object relationship information is a graph structure with preset objects as nodes and multiple association relationships corresponding to each object to be pushed as edges, and the association weight information is the degree of association of the multiple association relationships corresponding to the object to be pushed;

[0051] An aggregation processing module is configured to perform aggregation processing on the target-related objects to obtain target aggregation information corresponding to the objects to be pushed;

[0052] The object pushing module is configured to execute a push operation on the object to be pushed based on the target aggregation information.

[0053] As an optional embodiment, the target-related object determination module includes:

[0054] An object association module is configured to perform at least one object association on the object to be pushed based on preset object relationship information to obtain an associated object corresponding to the object to be pushed;

[0055] The object sampling module is configured to perform object sampling on the associated objects corresponding to the object to be pushed based on the associated weight information to obtain the target associated objects.

[0056] As an optional embodiment, the object sampling module includes:

[0057] an initial sampling object determining unit, configured to determine an initial sampling object from associated objects corresponding to the object to be pushed, and use the initial sampling object as a current sampling object;

[0058] an object sampling information determining unit, configured to determine current object sampling information corresponding to the current sampling object based on the association weight information;

[0059] a sampling object updating unit, configured to update the current sampling object based on the current object sampling information;

[0060] a first iterative unit configured to repeat the iterative operations of determining the current object sampling information corresponding to the current sampling object based on the association weight information and updating the current sampling object based on the current object sampling information until the current object sampling information satisfies a first preset condition;

[0061] The target associated object determining unit is configured to, when the current object sampling information satisfies the first preset condition, update the current sampling object based on the current object sampling information that satisfies the first preset condition to obtain the target associated object.

[0062] As an optional embodiment, the object sampling information determining unit includes:

[0063] The to-be-processed object determining unit is configured to obtain a preset number of sampling objects from the current sampling object; and use the preset number of sampling objects as first to-be-processed objects, and use the other sampling objects in the current sampling object except the first to-be-processed objects as second to-be-processed objects:

[0064] The second iterative unit is configured to repeatedly perform an iterative operation until the first object to be processed meets a second preset condition, and the second iterative unit includes:

[0065] A first association feature determining unit is configured to determine first association feature information corresponding to the first to-be-processed object based on the association weight information;

[0066] a target association feature determination unit, configured to determine target association feature information from the first association feature information;

[0067] A second association feature determining unit is configured to determine second association feature information corresponding to a second object to be processed based on the association weight information;

[0068] A first object to be processed updating unit is configured to update the first object to be processed and the second object to be processed based on the second association feature information and the target association feature information;

[0069] The current object sampling information determining unit is configured to determine the current object sampling information based on the first object to be processed that meets a preset condition.

[0070] As an optional embodiment, the aggregation processing module includes:

[0071] The model processing unit is configured to input the target associated object into the associated object aggregation model, perform at least one aggregation process on the target associated object, and obtain the target aggregation information. The associated object aggregation model is a model for performing aggregation processing on the target associated object.

[0072] As an optional embodiment, the model processing unit includes:

[0073] a first to-be-aggregated object determining unit configured to input the target associated object into the associated object aggregation model, and determine, in the case of a first aggregation process, an initial to-be-aggregated object from the target associated object based on the association degree;

[0074] A first aggregation processing unit is configured to perform aggregation processing on the initial objects to be aggregated, and obtain object aggregation information corresponding to the first aggregation processing;

[0075] The previous aggregation information obtaining unit is configured to obtain the previous object aggregation information corresponding to each aggregation process when the aggregation process is not the first one;

[0076] a second to-be-aggregated object determining unit configured to determine, based on the association degree, an to-be-aggregated object corresponding to each aggregation process from the target associated objects;

[0077] A second aggregation processing unit is configured to perform aggregation processing on the object to be aggregated and the previous object aggregation information to obtain object aggregation information corresponding to each aggregation processing;

[0078] The target aggregation information determining unit is configured to, when the number of aggregation processes meets a preset number, use the object aggregation information corresponding to the aggregation processes when the preset number of times is met as the target aggregation information.

[0079] As an optional embodiment, the associated object aggregation model includes an aggregation processing layer and a feature perception layer, and the second aggregation processing unit includes:

[0080] a feature aggregation unit configured to input the to-be-aggregated object and the previous aggregation processing information into the aggregation processing layer to perform feature aggregation and obtain aggregated feature information;

[0081] a feature perception unit configured to input the aggregated feature information into the feature perception layer for feature perception to obtain perceived feature information;

[0082] The object aggregation information acquisition unit is configured to use the perception feature information as the object aggregation information corresponding to each aggregation process.

[0083] As an optional embodiment, the device further includes:

[0084] An association relationship acquisition module is configured to acquire multiple association relationships corresponding to each object to be pushed;

[0085] an association weight determination module, configured to determine association weight information corresponding to each of the plurality of association relationships;

[0086] The object relationship information construction module is configured to execute, with multiple objects to be pushed as nodes, and based on the multiple association relationships and the association weight information corresponding to the multiple association relationships, to construct the edge of each node corresponding to the object to be pushed, so as to obtain the object relationship information.

[0087] As an optional embodiment, the object pushing module includes:

[0088] a push request response unit configured to execute a push request corresponding to a target object, obtain first target aggregate information corresponding to the target object and second target aggregate information corresponding to other objects among the objects to be pushed except the target object;

[0089] a similarity determining unit, configured to determine a similarity between the first target aggregate information and the second target aggregate information;

[0090] a target push object determining unit, configured to determine a target push object corresponding to the target object based on the similarity;

[0091] The push operation execution unit is configured to execute pushing of the target push object to the target object.

[0092] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising:

[0093] processor;

[0094] a memory for storing instructions executable by the processor;

[0095] The processor is configured to execute the instructions to implement the object pushing method as described above.

[0096] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the object push method described above.

[0097] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program, which implements the object pushing method described above when executed by a processor.

[0098] The technical solutions provided by the embodiments of the present disclosure bring at least the following beneficial effects:

[0099] Based on preset object relationship information and the association weight information corresponding to the object to be pushed, the target associated objects corresponding to the object to be pushed are determined, where each object to be pushed has multiple associations. The target associated objects are aggregated to obtain target aggregate information corresponding to the object to be pushed. Based on the target aggregate information, a push operation is performed on the object to be pushed. This method can fully utilize multiple associations and the importance of different associations to discover potential associations between users, thereby improving the accuracy and effectiveness of object push.

[0100] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute an improper limitation of the present disclosure.

[0102] Figure 1The figure is a schematic diagram of an application scenario of an object push method according to an exemplary embodiment.

[0103] Figure 2 The figure is a flowchart of a method for pushing an object according to an exemplary embodiment.

[0104] Figure 3 The present invention is a flowchart of determining a target associated object in an object push method according to an exemplary embodiment.

[0105] Figure 4 The present invention is a flowchart showing a method for determining current object sampling information in an object pushing method according to an exemplary embodiment.

[0106] Figure 5 The present invention is a flowchart of obtaining target aggregate information in an object push method according to an exemplary embodiment.

[0107] Figure 6 The present invention is a flowchart showing aggregation processing in an object push method according to an exemplary embodiment.

[0108] Figure 7 The present invention is a flowchart of an object pushing method according to an exemplary embodiment.

[0109] Figure 8 The present invention is a flowchart of constructing object relationship information in an object push method according to an exemplary embodiment.

[0110] Figure 9 The figure is a block diagram showing an object pushing device according to an exemplary embodiment.

[0111] Figure 10 It is a block diagram of an electronic device according to an exemplary embodiment (general structure of a server). DETAILED DESCRIPTION

[0112] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0113] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure as detailed in the appended claims.

[0114] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0115] Figure 1 FIG. 1 is a schematic diagram showing an application scenario of an object push method according to an exemplary embodiment. Figure 1 As shown, the application scenario includes a client 110 and a server 120. The server 120 constructs object relationship information based on the object to be pushed sent by the client 110 and the multiple association relationships corresponding to the object to be pushed. The server 120 determines the target associated object corresponding to the object to be pushed based on the object relationship information and the association weight information corresponding to the object to be pushed. The server 120 aggregates the target associated objects to obtain target aggregate information corresponding to the object to be pushed. In response to the object push request, the server 120 obtains the first target aggregate information corresponding to the client 110 and the second target aggregate information corresponding to other objects, compares the first target aggregate information and the second target aggregate information, and can determine the target push object based on the comparison result. The server 120 sends the target push object to the client 110.

[0116] In the embodiment of the present disclosure, the client 110 includes physical devices such as smart phones, desktop computers, tablet computers, laptops, digital assistants, smart wearable devices, etc., and may also include software running in the physical devices, such as applications. The operating system running on the physical device in the embodiment of the present application may include but is not limited to Android, IOS, Linux, Unix, Windows, etc. The client 110 includes a UI (User Interface) layer, and the client 110 provides external display of target push objects and collection of target object associations through the UI layer. In addition, the target object associations are sent to the server 120 based on the API (Application Programming Interface).

[0117] In the disclosed embodiment, server 120 may include a standalone server, a distributed server, or a server cluster consisting of multiple servers. Server 120 may include a network communication unit, a processor, a memory, and the like. Specifically, server 120 may be configured to pre-sample and aggregate objects to be pushed based on object relationship information, obtain target aggregate information corresponding to the objects to be pushed, and push the objects to client 110 based on the target aggregate information.

[0118] Figure 2 FIG. 1 is a flow chart showing a method for pushing an object according to an exemplary embodiment. Figure 2 As shown, the method is used in a server and includes the following steps.

[0119] S210. Based on the preset object relationship information and the association weight information corresponding to the object to be pushed, determine the target associated object corresponding to the object to be pushed. The object relationship information is a graph structure with preset objects as nodes and multiple associations corresponding to each object to be pushed as edges. The association weight information is the degree of association of the multiple associations corresponding to the object to be pushed;

[0120] As an optional embodiment, based on preset object relationship information and the association weight information corresponding to the object to be pushed, the target associated object can be determined from the associated objects corresponding to each object to be pushed. The association relationship can be the interaction relationship between two preset objects within the target application, as well as the interaction relationship between two preset objects outside the target application. For example, the interaction relationship within the target application can be a pun relationship, a like relationship, a favorite relationship, etc., and the interaction relationship outside the target application can be a mobile phone address book relationship, a chat software friend relationship, etc. Different association relationships can correspond to different association weight information.

[0121] As an optional embodiment, determining the target associated object corresponding to the object to be pushed based on preset object relationship information and associated weight information corresponding to the object to be pushed includes:

[0122] Based on the preset object relationship information, perform at least one object association on the object to be pushed to obtain the associated object corresponding to the object to be pushed;

[0123] Based on the association weight information, object sampling is performed on the associated objects corresponding to the object to be pushed to obtain the target associated objects.

[0124] As an optional embodiment, based on preset object relationship information, the object to be pushed is used as the central node, and at least one object association is performed on the central node to obtain an associated node corresponding to the object to be pushed. The object to be pushed corresponding to the associated node is used as the associated object. The number of object associations can be set to a preset number. The starting node of the first object association in the preset number of object associations is the object to be pushed. The starting node of any subsequent object association is the associated node obtained from the previous object association.

[0125] As an optional embodiment, based on the association weight information, the object sampling information during object sampling can be determined. The object sampling information can be a sampling probability distribution. Based on the object sampling information, object sampling is performed on the associated objects corresponding to the pushed objects, and the target associated objects matching the object sampling information can be obtained, that is, the statistical properties of the target associated objects obey the sampling probability distribution.

[0126] When sampling objects based on association weight information, more association objects with larger association weights can be sampled, and fewer association objects with smaller association weights can be sampled, thereby improving the closeness of the association between the target association object and the object to be pushed, facilitating object push in subsequent steps.

[0127] As an alternative embodiment, see Figure 3 Based on the association weight information, object sampling is performed on the associated objects corresponding to the pushed object, and the target associated objects include:

[0128] S310. Determine the initial sampling object from the associated objects corresponding to the object to be pushed;

[0129] S320. The initial sampling object is used as the current sampling object;

[0130] S330. Based on the associated weight information, determine the current object sampling information corresponding to the current sampling object;

[0131] S340. Based on the current object sampling information, update the current sampling object;

[0132] S350 repeats the iterative operation of determining the current object sampling information corresponding to the current sampling object based on the associated weight information to updating the current sampling object based on the current object sampling information until the current object sampling information meets the first preset condition;

[0133] S360. When the current object sampling information satisfies the first preset condition, update the current sampling object based on the current object sampling information that satisfies the first preset condition to obtain a target associated object.

[0134] As an optional embodiment, a sampled object is randomly selected from the associated objects corresponding to the object to be recommended as the initial sampled object. In the first object sampling, the initial sampled object is used as the current sampled object. Based on the association weight information corresponding to the initial sampled object, the first object sampling information corresponding to the initial sampled object can be determined. Based on the first object sampling information, the associated objects corresponding to the object to be recommended are sampled, and the initial sampled object can be updated to obtain a second sampled object. In the second object sampling, the second sampled object is used as the current sampled object. Based on the association weight information corresponding to the second sampled object, the second object sampling information corresponding to the second sampled object can be determined. Based on the second object sampling information, the associated objects corresponding to the object to be recommended are sampled, and the second sampled object can be updated to obtain a third sampled object. In the third object sampling, the third sampled object is used as the current sampled object. Similarly, in each object sampling, the current sampled object is obtained, and based on the association weight information corresponding to the current sampled object, the current object sampling information corresponding to the current sampled object can be determined. Based on the current object sampling information, the associated objects corresponding to the object to be recommended are sampled, and the current sampled object can be updated to obtain an updated sampled object.

[0135] As an optional embodiment, the first preset condition may be a preset difference condition. When the difference between the current object sampling information and the previous object sampling information obtained in each object sampling operation satisfies the preset difference condition, that is, when the difference between the current object sampling information and the previous object sampling information obtained in each object sampling operation is less than a preset difference threshold, that is, when the change in the current object sampling information tends to be stable, it can be determined that the current object sampling information satisfies the first preset condition. The sampled object obtained by updating the current sampling object based on the current object sampling information that satisfies the first preset condition can be used as the target associated object. In other words, based on the current object sampling information that satisfies the first preset condition, the associated objects corresponding to the to-be-recommended object are sampled, and the obtained sampling result is the target associated object.

[0136] Determining the current sampling object information based on the current sampling object obtained by sampling can simplify the process of determining the expectation of the high-dimensional current sampling object, thereby improving the efficiency of determining the current object sampling information that meets the preset conditions, thereby improving the efficiency of determining the target associated object.

[0137] As an alternative embodiment, see Figure 4 , based on the association weight information, determining the current object sampling information corresponding to the current sampling object includes:

[0138] S410. Obtain a preset number of sampling objects from the current sampling object;

[0139] S420. Set a preset number of sample objects as the first objects to be processed, and set the other sample objects except the first objects to be processed in the current sample objects as the second objects to be processed:

[0140] S430. Repeat the iterative operation until the first object to be processed meets the second preset condition, the iterative operation including:

[0141] S4301. Based on the association weight information, determine the first association feature information corresponding to the first object to be processed;

[0142] S4303. Determine target association feature information from the first association feature information;

[0143] S4305. Based on the association weight information, determine the second association feature information corresponding to the second object to be processed;

[0144] S4307. Based on the second associated feature information and the target associated feature information, the first object to be processed and the second object to be processed are updated;

[0145] S440. Determine current object sampling information based on the first object to be processed that meets the second preset condition.

[0146] As an optional embodiment, a preset number of sampling objects are obtained from the current sampling object, and the preset number of sampling objects are used as the first objects to be processed, and the sampling objects other than the first objects to be processed in the current sampling object are used as the second objects to be processed. When obtaining the preset number of sampling objects from the current sampling object, the sampling objects can be randomly selected, or the sampling objects in the current sampling object can be sorted based on the time the current sampling object has paid attention to the object to be pushed, and then the first preset number of sampling objects are selected from the sorted sampling objects.

[0147] As an optional embodiment, based on the association weight information corresponding to the first object to be processed, the first association feature information corresponding to the first object to be processed can be calculated, and based on the association weight information corresponding to the second object to be processed, the second association feature information corresponding to the second object to be processed can be calculated. The calculation formula of the association feature information is as follows:

[0148] where u i =rand(0,1)

[0149] Among them, rand function is a function that generates random numbers, u i is a random number between 0 and 1, k i is the associated feature information corresponding to the sampling object, which can be the first associated feature information or the second associated feature information. i is the associated weight information.

[0150] As an optional embodiment, the first correlation feature information corresponding to each sampling object in the first object to be processed is compared pairwise, and the minimum value in the first correlation feature information corresponding to the first object to be processed is determined to obtain the target correlation feature information.

[0151] Each second correlation feature information is compared with the target correlation feature information. If the second correlation feature information is smaller than the target correlation feature information, the sampling object corresponding to the second correlation feature information smaller than the target correlation feature information is removed from the second object to be processed, and the second object to be processed is updated.

[0152] If the second association feature information is greater than the target association feature information, the sampling object corresponding to the target association feature information is removed from the first object to be processed, and the sampling object corresponding to the second association feature information greater than the target association feature information is removed from the second object to be processed and added to the first object to be processed, and the first object to be processed and the second object to be processed are updated.

[0153] After the first object to be processed is updated, the target-associated feature information can be updated based on the first associated feature information corresponding to the updated first object to be processed. The first associated feature information corresponding to the updated first object to be processed is compared pairwise, and the minimum value in the first associated feature information corresponding to the updated first object to be processed is re-determined, thereby updating the target-associated feature information. Each second associated feature information is compared with the updated target-associated feature information. When the second associated feature information is greater than the updated target-associated feature information, the above-mentioned iterative steps of updating the first object to be processed and the second object to be processed, and updating the target-associated feature information are repeated until the first object to be processed is no longer updated, that is, the second preset condition is met, and the second preset condition indicates that the first object to be processed is no longer updated. When the first object to be processed is no longer updated, the sampling objects in the current sampling object are either deleted during the object processing process or divided into the first object to be processed, that is, all the sampling objects in the current sampling object have been processed.

[0154] When comparing the second correlation feature information with the target correlation feature information, the second correlation feature information may be sorted and compared with the target correlation feature information in ascending order.

[0155] As an optional embodiment, based on the first object to be processed that meets the second preset condition, the sampling expectation information corresponding to the first object to be processed that meets the second preset condition can be determined. The formula for calculating the sampling expectation information is as follows:

[0156]

[0157] Wherein, E represents the sampling expectation information, L represents the number of first objects to be processed that meet the second preset condition, x represents the first objects to be processed that meet the second preset condition, and f(x) represents the objective function corresponding to the first objects to be processed that meet the second preset condition.

[0158] Based on the sampling expectation information, the current object sampling information can be calculated. The formula for calculating the current object sampling information is as follows:

[0159] E=∫f(x)p(x)dx

[0160] Wherein, E represents the sampling expectation information, f(x) represents the objective function corresponding to the first object to be processed that meets the preset conditions, and p(x) represents the probability distribution corresponding to the first object to be processed that meets the preset conditions, which is also the sampling information of the current object.

[0161] Based on different association weight information, the first object to be processed that meets the second preset condition is determined, and then the current object sampling information is determined based on the first object to be processed that meets the second preset condition. This can make full use of multiple association relationships and the importance corresponding to different association relationships, thereby improving the accuracy and effectiveness of the current object sampling information.

[0162] S220. Aggregate the target associated objects to obtain target aggregation information corresponding to the object to be pushed;

[0163] As an optional embodiment, the target associated object is obtained by sampling the associated objects obtained from the recommended object. Therefore, the target associated object also corresponds to multiple object associations. The distance between the target associated object and the object to be pushed can be determined based on the order of the object associations, and thus the degree of association between the target associated object and the object to be pushed can be determined based on the distance. Based on the degree of association, the target associated objects corresponding to each object association are successively aggregated to obtain target aggregate information corresponding to the object to be pushed.

[0164] As an optional embodiment, performing aggregation processing on target-related objects to obtain target aggregation information corresponding to the object to be pushed includes:

[0165] The target associated object is input into the associated object aggregation model. Based on the association between the target associated object and the object to be pushed, the target associated object is aggregated at least once to obtain target aggregation information. The associated object aggregation model is a model for aggregating the target associated object.

[0166] As an optional embodiment, the object to be pushed is input into the associated object determination model, and based on the object relationship information in the associated object determination model and the associated weight information corresponding to the object to be pushed, the target associated object corresponding to the object to be pushed can be determined. The target associated object is input into the associated object aggregation model, and based on the associated object aggregation model, the target associated object can be aggregated at least once to obtain target aggregation information. The associated object determination model and the associated object aggregation model can be models determined based on a graph neural network algorithm, the associated object determination model can be obtained based on the sampling algorithm in the graph neural network algorithm, and the associated object aggregation model can be obtained based on the aggregation algorithm in the graph neural network algorithm. The graph neural network algorithm can be a GraphSAGE (Graph SAmple and aggreGatE) algorithm, and the GraphSAGE algorithm can determine the embedded vector of the object to be pushed based on the aggregated representation information obtained by aggregating the target associated object.

[0167] As an optional embodiment, based on the distance between the object to be pushed and the target-associated object in the object relationship information, the degree of association between the object to be pushed and the target-associated object can be determined, and then based on the size of the association, the target-associated objects are aggregated at least once from small to large to obtain target aggregation information, that is, based on the distance in the object relationship information, the target-associated objects are aggregated from far to near.

[0168] In the associated object aggregation model, the target associated objects are summarized and learned, and based on the size of the association, the target associated objects are aggregated one by one to obtain target aggregation information, which can reduce the calculation amount of the model and thus improve the efficiency of determining the target aggregation information.

[0169] As an alternative embodiment, see Figure 5 , input the target associated object into the associated object aggregation model, and perform at least one aggregation process on the target associated object based on the degree of association between the target associated object and the object to be pushed, and obtain target aggregation information including:

[0170] S510. Input the target associated object into the associated object aggregation model. In the case of the first aggregation process, based on the association degree, determine the initial object to be aggregated from the target associated object;

[0171] S520. Perform aggregation processing on the initial objects to be aggregated to obtain object aggregation information corresponding to the first aggregation processing;

[0172] S530. In the case of a non-first aggregation process, obtaining the previous object aggregation information corresponding to each aggregation process;

[0173] S540. Based on the relevance, determine the object to be aggregated corresponding to each aggregation process from the target associated object;

[0174] S550. Aggregate the object to be aggregated and the previous object aggregation information to obtain the object aggregation information corresponding to each aggregation process;

[0175] S560. When the number of aggregation processes meets the preset number, the object aggregation information corresponding to the aggregation processes when the preset number of times is met is used as the target aggregation information.

[0176] As an optional embodiment, the target associated objects are input into the associated object aggregation model. In the case of the first aggregation process, based on the degree of association, the object obtained in the last object association when the to-be-pushed object is associated with the target associated objects is determined. The object obtained in the last object association is used as the initial object to be aggregated. Aggregation processing is performed on the initial object to be aggregated to obtain object aggregation information corresponding to the first aggregation process.

[0177] In the case of the second aggregation process, based on the degree of association, the object obtained in the penultimate object association when the to-be-pushed object is associated is determined from the target associated objects. The object obtained in the penultimate object association is used as the to-be-aggregated object corresponding to the second aggregation process. The object aggregation information corresponding to the first aggregation process and the to-be-aggregated object corresponding to the second aggregation process are aggregated to obtain the object aggregation information corresponding to the second aggregation process.

[0178] In the case of the third aggregation process, based on the degree of association, the object obtained in the third-to-last object association when the to-be-pushed object is associated with the target associated objects is determined, and the object obtained in the third-to-last object association is used as the to-be-aggregated object for the third aggregation process. The object aggregation information corresponding to the second aggregation process and the to-be-aggregated object corresponding to the third aggregation process are aggregated to obtain the object aggregation information corresponding to the third aggregation process.

[0179] This process is repeated until the object obtained from the first object association is used as the object to be aggregated corresponding to the last aggregation process. The object aggregation information corresponding to the penultimate aggregation process and the object to be aggregated corresponding to the last aggregation process are aggregated to obtain the object aggregation information corresponding to the last aggregation process. At this point, the number of aggregation processes meets the preset number, and the object aggregation information corresponding to the last aggregation process can be used as the target aggregation information. The object aggregation information corresponding to the last aggregation process, that is, the object aggregation information corresponding to the aggregation process that meets the preset number, is used as the target aggregation information.

[0180] As an optional embodiment, the number of aggregation processes is the same as the number of object associations.

[0181] Aggregation is performed starting from the last associated object determined until the first associated object is aggregated. The object aggregation information of the last aggregation process is added to each aggregation process, so that the correlation between the target associated object and the object to be pushed can be fully utilized, and the result of the last aggregation process can be learned, thereby improving the accuracy and effectiveness of the aggregation process.

[0182] As an alternative embodiment, see Figure 6 The associated object aggregation model includes an aggregation processing layer and a feature perception layer. It aggregates the object to be aggregated and the previous aggregation processing information. The object aggregation information corresponding to each aggregation processing includes:

[0183] S610. Input the object to be aggregated and the previous aggregation processing information into the aggregation processing layer for feature aggregation to obtain aggregate feature information;

[0184] S620. Inputting the aggregated feature information into the feature perception layer for feature perception to obtain the perceived feature information;

[0185] S630. Use the perception feature information as the object aggregation information corresponding to each aggregation process.

[0186] As an optional embodiment, the associated object aggregation model includes an aggregation processing layer and a feature perception layer. The feature perception layer can be a multilayer perceptron (MLP). In the aggregation processing layer, feature aggregation is performed on the input object to be aggregated and the previous aggregation processing information to obtain aggregated feature information. In the feature perception layer, feature perception is performed on the input aggregated feature information, that is, feature generalization processing is performed on the aggregated feature information to obtain perceived feature information. The perceived feature information output by the feature perception layer is used as the object aggregation information.

[0187] As an optional embodiment, the target associated object is input into the associated object aggregation model. In the case of the first aggregation process, after the initial objects to be aggregated are determined based on the degree of association, feature aggregation is performed on the initial objects to be aggregated in the aggregation processing layer to obtain aggregated feature information corresponding to the first aggregation process. In the feature perception layer, feature perception is performed on the aggregated feature information corresponding to the first aggregation process to obtain perception feature information corresponding to the first aggregation process, and the perception feature information corresponding to the first aggregation process is used as the object aggregation information corresponding to the first aggregation process.

[0188] The target associated object is input into the associated object aggregation model. In the case of non-first aggregation processing, based on the degree of association, the to-be-aggregated object corresponding to each aggregation processing and the aggregation information of the previous object corresponding to each aggregation processing are determined. In the aggregation processing layer, feature aggregation is performed on the to-be-aggregated object to obtain the aggregation result corresponding to the to-be-aggregated object. The aggregation result corresponding to the to-be-aggregated object and the previous aggregation processing information are feature concatenated to obtain the aggregated feature information corresponding to each aggregation processing. In the feature perception layer, feature perception is performed on the aggregated feature information corresponding to each aggregation processing to obtain the perception feature information corresponding to each aggregation processing. The perception feature information corresponding to each aggregation processing is used as the object aggregation information corresponding to each aggregation processing.

[0189] As an optional embodiment, based on the adjacent objects of the object to be aggregated corresponding to each aggregation process, the representation information of the object to be aggregated can be obtained. In the aggregation processing layer, feature aggregation is performed on the representation information of the object to be aggregated to obtain the aggregation result corresponding to the object to be aggregated.

[0190] As an optional embodiment, in the aggregation processing layer, the calculation formula for performing feature aggregation on the input object to be aggregated and the aggregation information of the previous object is as follows:

[0191]

[0192] in, is the target aggregation information corresponding to the object to be pushed, W l It is the weight information corresponding to each layer of the neural network in the associated object aggregation model, MLP is the feature perception layer, and CONCAT is the feature concatenation function, which is used to perform feature concatenation on the aggregation result corresponding to the aggregated object and the previous aggregation processing information. Aggregate information of the previous object corresponding to each aggregation process, is the adjacent object of the object to be aggregated corresponding to each aggregation process, l represents the number of aggregations, and N(v) represents the number range of target associated objects.

[0193] In the aggregation processing layer, the input objects to be aggregated can be aggregated first, and then the aggregation results of the objects to be aggregated are aggregated with the aggregation information of the previous object. The formula for aggregating the objects to be aggregated is as follows:

[0194]

[0195] in, The aggregation result of the object to be aggregated corresponding to each aggregation process, is the adjacent object of the object to be aggregated corresponding to each aggregation process, N(v) represents the number range of target related objects, AGG represents the average aggregation method, and the calculation formula of AGG is as follows

[0196]

[0197] in, is the adjacent object of the object to be aggregated corresponding to each aggregation process, and N(v) represents the number range of target associated objects.

[0198] Adding a feature perception layer to the associated object aggregation model can improve the generalization of the object aggregation information corresponding to each aggregation process, and adopt different aggregation methods for different types of association relationships, thereby improving the accuracy of the associated object aggregation model.

[0199] S230. Based on the target aggregation information, execute a push operation on the object to be pushed.

[0200] As an optional embodiment, based on the target aggregation information corresponding to each object to be pushed, the similarity between two objects to be pushed can be determined, and then the push operation is performed on the objects to be pushed based on the similarity.

[0201] As an alternative embodiment, see Figure 7 Based on the target aggregation information, the push operations performed on the push object include:

[0202] S710. In response to the push request corresponding to the target object, obtain the first target aggregate information corresponding to the target object and the second target aggregate information corresponding to the other objects to be pushed except the target object;

[0203] S720. Determine the similarity between the first target aggregate information and the second target aggregate information;

[0204] S730. Based on the similarity, determine the target push object corresponding to the target object;

[0205] S740. Push the target push object to the target object.

[0206] As an optional embodiment, in response to a push request corresponding to a target object, first target aggregate information corresponding to the target object and second target aggregate information corresponding to other objects to be pushed, excluding the target object, may be obtained. The target object may be a target user, and the push request may be a request sent by the user to the server or a push request periodically generated by the server to proactively push to the user.

[0207] When determining the similarity between the first target aggregate information and the second target aggregate information, a dot product can be performed on the first target aggregate information and the second target aggregate information to obtain a dot product result, and the dot product result is used as the similarity. A larger dot product result indicates that the target object is more similar to the other objects corresponding to the dot product result, and a smaller dot product result indicates that the target object is less similar to the other objects corresponding to the dot product result. The formula for calculating similarity is as follows:

[0208]

[0209] Among them, z a represents the transpose of the first target aggregation information, Represents the transpose of the second target aggregation information, a represents the target object, and b represents other objects.

[0210] As an optional embodiment, when the similarity is greater than or equal to a preset similarity threshold, other objects corresponding to similarities greater than or equal to the similarity threshold are determined as target push objects corresponding to the target object; when the similarity is less than the preset similarity threshold, other objects corresponding to similarities less than the similarity threshold are not determined as target push objects corresponding to the target object.

[0211] Objects can be pushed based on target aggregation information. The target aggregation information is obtained by sampling and aggregating objects based on a variety of different association relationships, so that the potential connection between two users can be determined more accurately, thereby improving the accuracy and effectiveness of object push.

[0212] As an alternative embodiment, see Figure 8 , the method further comprises:

[0213] S810. Get multiple associations corresponding to each object to be pushed;

[0214] S820. Determine the association weight information corresponding to each association relationship;

[0215] S830. Taking multiple objects to be pushed as nodes, and based on multiple association relationships and association weight information corresponding to each association relationship, construct an edge corresponding to each node of each object to be pushed to obtain object relationship information.

[0216] As an optional embodiment, multiple associations corresponding to each object to be pushed are obtained, such as mobile phone address book relationships, pun relationships, etc., and corresponding association weight information is set for different associations. For example, the association weight information corresponding to the mobile phone address book relationship can be set to 8, the association weight information corresponding to the chat software friend relationship can be set to 4, and the association weight information corresponding to the pun relationship can be set to 2. When all associations are bidirectional, the object relationship information constructed is an undirected weighted graph.

[0217] As an optional embodiment, if in the process of model training of the associated object determination model and the associated object aggregation model, the object to be pushed can be filtered when constructing the object relationship information used in the model training, and the objects to be pushed whose number of associated relationships is greater than a first preset number threshold are removed, and the objects to be pushed whose number of associated relationships is less than a second preset number threshold are removed. For example, the objects to be pushed whose number of associated relationships is greater than 10,000 are filtered out from the objects to be pushed, and the objects to be pushed whose number of associated relationships is less than 2 are filtered out.

[0218] When constructing object relationship information, we add association weight information to integrate multiple associations, fully utilizing complex associations during object push and improving object push accuracy. Furthermore, by performing data processing during object relationship information construction, we avoid focusing solely on objects with many associations during model training, preventing inadequate training and improving model training accuracy.

[0219] The disclosed embodiments provide an object push method, comprising: determining target associated objects corresponding to a to-be-pushed object based on preset object relationship information and association weight information corresponding to the to-be-pushed object, wherein each to-be-pushed object corresponds to multiple association relationships. Aggregating the target associated objects to obtain target aggregate information corresponding to the to-be-pushed object; and performing a push operation on the to-be-pushed object based on the target aggregate information. This method can fully utilize multiple association relationships and the importance levels corresponding to different association relationships to discover potential association relationships between users, thereby improving the accuracy and effectiveness of object push.

[0220] Figure 9 FIG. 1 is a block diagram of an object pushing device according to an exemplary embodiment. Figure 9 , the device comprises:

[0221] Target associated object determination module 910 is configured to determine a target associated object corresponding to an object to be pushed based on preset object relationship information and association weight information corresponding to the object to be pushed. The object relationship information is a graph structure with preset objects as nodes and multiple associations corresponding to each object to be pushed as edges. The association weight information is the degree of association of the multiple associations corresponding to the object to be pushed.

[0222] Aggregation processing module 920, configured to perform aggregation processing on target-related objects to obtain target aggregation information corresponding to the object to be pushed;

[0223] The object push module 930 is configured to execute a push operation on the object to be pushed based on the target aggregation information.

[0224] As an optional embodiment, the target-related object determination module 910 includes:

[0225] The object association module is configured to perform at least one object association on the object to be pushed based on preset object relationship information to obtain an associated object corresponding to the object to be pushed;

[0226] The object sampling module is configured to perform object sampling on the associated objects corresponding to the object to be pushed based on the associated weight information to obtain the target associated objects.

[0227] As an optional embodiment, the object sampling module includes:

[0228] an initial sampling object determining unit, configured to determine an initial sampling object from associated objects corresponding to the object to be pushed, and use the initial sampling object as a current sampling object;

[0229] an object sampling information determining unit, configured to determine current object sampling information corresponding to a current sampling object based on the association weight information;

[0230] The sampling object updating unit is configured to update the current sampling object based on the current object sampling information;

[0231] A first iterative unit is configured to repeatedly perform iterative operations of determining the current object sampling information corresponding to the current sampling object based on the association weight information and updating the current sampling object based on the current object sampling information, until the current object sampling information meets a first preset condition;

[0232] The target associated object determining unit is configured to, when the current object sampling information satisfies the first preset condition, update the current sampling object based on the current object sampling information that satisfies the first preset condition to obtain the target associated object.

[0233] As an optional embodiment, the object sampling information determining unit includes:

[0234] The unit for determining objects to be processed is configured to obtain a preset number of sampling objects from the current sampling object; and use the preset number of sampling objects as first objects to be processed, and use the other sampling objects in the current sampling object except the first objects to be processed as second objects to be processed:

[0235] The second iterative unit is configured to repeatedly perform the iterative operation until the first object to be processed meets the second preset condition, and the second iterative unit includes:

[0236] A first association feature determining unit is configured to determine first association feature information corresponding to the first object to be processed based on the association weight information;

[0237] a target association feature determination unit, configured to determine target association feature information from the first association feature information;

[0238] A second association feature determining unit is configured to determine second association feature information corresponding to the second object to be processed based on the association weight information;

[0239] A first object to be processed updating unit is configured to update the first object to be processed and the second object to be processed based on the second association feature information and the target association feature information;

[0240] The current object sampling information determining unit is configured to determine the current object sampling information based on the first object to be processed that meets the second preset condition.

[0241] As an optional embodiment, the aggregation processing module 920 includes:

[0242] The model processing unit is configured to input the target associated object into the associated object aggregation model, perform at least one aggregation process on the target associated object, and obtain target aggregation information.

[0243] As an optional embodiment, the model processing unit includes:

[0244] A first to-be-aggregated object determining unit is configured to input the target associated object into the associated object aggregation model, and in the case of a first aggregation process, determine an initial to-be-aggregated object from the target associated object based on the degree of association;

[0245] A first aggregation processing unit is configured to perform aggregation processing on the initial objects to be aggregated, and obtain object aggregation information corresponding to the first aggregation processing;

[0246] The previous aggregation information obtaining unit is configured to obtain the previous object aggregation information corresponding to each aggregation process when the aggregation process is not the first one;

[0247] The second to-be-aggregated object determining unit is configured to determine, based on the relevance, the to-be-aggregated object corresponding to each aggregation process from the target related objects;

[0248] The second aggregation processing unit is configured to perform aggregation processing on the object to be aggregated and the aggregation information of the previous object, and obtain object aggregation information corresponding to each aggregation processing;

[0249] The target aggregation information determining unit is configured to, when the number of aggregation processes meets the preset number, use the object aggregation information corresponding to the aggregation processes when the preset number of times is met as the target aggregation information.

[0250] As an optional embodiment, the associated object aggregation model includes an aggregation processing layer and a feature perception layer, and the second aggregation processing unit includes:

[0251] The feature aggregation unit is configured to perform feature aggregation by inputting the object to be aggregated and the previous aggregation processing information into the aggregation processing layer to obtain aggregated feature information;

[0252] The feature sensing unit is configured to input the aggregated feature information into the feature sensing layer for feature sensing to obtain the sensed feature information;

[0253] The object aggregation information acquisition unit is configured to use the perception feature information as the object aggregation information corresponding to each aggregation process.

[0254] As an optional embodiment, the device further includes:

[0255] An association relationship acquisition module is configured to acquire multiple association relationships corresponding to each object to be pushed;

[0256] An association weight determination module is configured to determine association weight information corresponding to a plurality of association relationships;

[0257] The object relationship information construction module is configured to execute, with multiple objects to be pushed as nodes, and based on multiple association relationships and association weight information corresponding to the multiple association relationships, construct edges corresponding to the nodes of each object to be pushed to obtain object relationship information.

[0258] As an optional embodiment, the object push module 930 includes:

[0259] a push request response unit configured to execute a push request corresponding to a target object, obtain first target aggregate information corresponding to the target object and second target aggregate information corresponding to other objects in the objects to be pushed except the target object;

[0260] A similarity determination unit is configured to determine the similarity between the first target aggregate information and the second target aggregate information;

[0261] a target push object determining unit, configured to determine a target push object corresponding to the target object based on similarity;

[0262] The push operation execution unit is configured to execute pushing of a target push object to a target object.

[0263] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0264] Figure 10 This is a block diagram of an electronic device for object push according to an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as shown in FIG. Figure 10 As shown. The electronic device includes a processor, a memory, and a network interface connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, an object push method is implemented.

[0265] Those skilled in the art will understand that Figure 10 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present disclosure, and does not constitute a limitation on the electronic device to which the scheme of the present disclosure is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0266] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 1004 including instructions, and the instructions can be executed by the processor 1020 of the electronic device 1000 to perform the above method. Alternatively, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0267] In an exemplary embodiment, a computer program product is further provided, including a computer program, which implements the above-mentioned object pushing method when executed by a processor.

[0268] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0269] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An object push method, characterized in that: Applied to a server, the method includes: Based on the preset object relationship information, perform at least one object association on the object to be pushed to obtain the associated object corresponding to the object to be pushed; Based on the association weight information, object sampling is performed on the associated objects corresponding to the object to be pushed to obtain a target associated object, wherein the object relationship information is a graph structure with preset objects as nodes and multiple association relationships corresponding to each object to be pushed as edges, and the association weight information is the degree of association of the multiple association relationships corresponding to the object to be pushed; the association relationships corresponding to the objects to be pushed are the interaction relationships between the preset objects in the target application and the interaction relationships between the preset objects outside the target application; the interaction relationships between the preset objects in the target application include at least one of a pun relationship, a like relationship, and a favorite relationship; the interaction relationships between the preset objects outside the target application include at least one of a mobile phone address book relationship and a chat software friend relationship; Inputting the target associated objects into an associated object aggregation model, and in the case of a first aggregation process, determining initial objects to be aggregated from the target associated objects based on the association degree, wherein the associated object aggregation model is a model for performing an aggregation process on the target associated objects, and the associated object aggregation model includes an aggregation processing layer and a feature perception layer; Performing aggregation processing on the initial objects to be aggregated to obtain object aggregation information corresponding to the first aggregation processing; In the case of a non-first aggregation process, obtain the previous object aggregation information corresponding to each aggregation process; Based on the association degree, determining the to-be-aggregated object corresponding to each aggregation process from the target associated objects; Inputting the object to be aggregated and the previous aggregation processing information into the aggregation processing layer to perform feature aggregation to obtain aggregated feature information; Inputting the aggregated feature information into the feature perception layer for feature perception to obtain perceived feature information; Using the perception feature information as object aggregation information corresponding to each aggregation process; When the number of aggregation processes meets the preset number, the object aggregation information corresponding to the aggregation processes meeting the preset number is used as the target aggregation information; Based on the target aggregation information, a target push object is pushed to the object to be pushed; the target push object is a user account pushed to the object to be pushed.

2. The object push method according to claim 1, characterized in that: The performing object sampling on the associated objects corresponding to the object to be pushed based on the association weight information to obtain the target associated objects includes: Determine an initial sampling object from the associated objects corresponding to the object to be pushed; Taking the initial sampling object as the current sampling object; Determining current object sampling information corresponding to the current sampling object based on the association weight information; Based on the current object sampling information, updating the current sampling object; Repeating the iterative operations of determining the current object sampling information corresponding to the current sampling object based on the association weight information to updating the current sampling object based on the current object sampling information until the current object sampling information meets a first preset condition; In a case where the current object sampling information satisfies the first preset condition, the current sampling object is updated based on the current object sampling information that satisfies the first preset condition to obtain the target associated object.

3. The object push method according to claim 2, characterized in that: The determining, based on the association weight information, current object sampling information corresponding to the current sampling object includes: Acquire a preset number of sampling objects from the current sampling object; A preset number of sampled objects are used as first objects to be processed, and the other sampled objects except the first objects to be processed in the current sampled objects are used as second objects to be processed: Repeating the iterative operation until the first object to be processed meets a second preset condition, the iterative operation comprising: Determining first association feature information corresponding to the first object to be processed based on the association weight information; Determining target association feature information from the first association feature information; Determining second association feature information corresponding to a second object to be processed based on the association weight information; updating the first to-be-processed object and the second to-be-processed object based on the second association characteristic information and the target association characteristic information; The current object sampling information is determined based on the first object to be processed that meets the second preset condition.

4. The object push method according to claim 1, characterized in that: The method further comprises: Get multiple associations corresponding to each object to be pushed; Determining association weight information corresponding to the plurality of association relationships; Taking the multiple objects to be pushed as nodes and based on the multiple association relationships and the association weight information corresponding to the multiple association relationships, an edge corresponding to the node of each object to be pushed is constructed to obtain the object relationship information.

5. The object push method according to claim 1, characterized in that: The pushing of the target push object to the object to be pushed based on the target aggregation information includes: In response to a push request corresponding to the object to be pushed, obtaining first target aggregate information corresponding to the object to be pushed and second target aggregate information corresponding to other objects among the objects to be pushed except the object to be pushed; determining a similarity between the first target aggregate information and the second target aggregate information; Based on the similarity, determining a target push object corresponding to the to-be-pushed object; Push the target push object to the object to be pushed.

6. An object pushing device, characterized in that: Applied to a server, the device includes: An object association module is configured to perform at least one object association on the object to be pushed based on preset object relationship information, and obtain an associated object corresponding to the object to be pushed; An object sampling module is configured to perform object sampling on associated objects corresponding to the object to be pushed based on association weight information to obtain target associated objects, wherein the object relationship information is a graph structure with preset objects as nodes and multiple association relationships corresponding to each object to be pushed as edges, and the association weight information is the degree of association of the multiple association relationships corresponding to the object to be pushed; the association relationships corresponding to the objects to be pushed are interaction relationships between two preset objects within the target application, and interaction relationships between two preset objects outside the target application; the interaction relationships between two preset objects within the target application include at least one of a pun relationship, a like relationship, and a favorite relationship; the interaction relationships between two preset objects outside the target application include at least one of a mobile phone address book relationship and a chat software friend relationship; a first to-be-aggregated object determining unit configured to input the target associated objects into an associated object aggregation model and, in the case of a first aggregation process, determine initial to-be-aggregated objects from the target associated objects based on the association degree, wherein the associated object aggregation model is a model for performing an aggregation process on the target associated objects and includes an aggregation processing layer and a feature perception layer; A first aggregation processing unit is configured to perform aggregation processing on the initial objects to be aggregated, and obtain object aggregation information corresponding to the first aggregation processing; The previous aggregation information obtaining unit is configured to obtain the previous object aggregation information corresponding to each aggregation process when the aggregation process is not the first one; a second to-be-aggregated object determining unit configured to determine, based on the association degree, an to-be-aggregated object corresponding to each aggregation process from the target associated objects; a feature aggregation unit configured to input the to-be-aggregated object and the previous aggregation processing information into the aggregation processing layer to perform feature aggregation and obtain aggregated feature information; a feature perception unit configured to input the aggregated feature information into the feature perception layer for feature perception to obtain perceived feature information; an object aggregation information acquiring unit, configured to use the perception feature information as the object aggregation information corresponding to each aggregation process; A target aggregation information determining unit is configured to, when the number of aggregation processes meets a preset number, use the object aggregation information corresponding to the aggregation processes that meet the preset number as the target aggregation information; The object pushing module is configured to push a target pushing object to the object to be pushed based on the target aggregation information; the target pushing object is a user account pushed to the object to be pushed.

7. The object pushing device according to claim 6, characterized in that: The object sampling module includes: an initial sampling object determining unit, configured to determine an initial sampling object from associated objects corresponding to the object to be pushed, and use the initial sampling object as a current sampling object; an object sampling information determining unit, configured to determine current object sampling information corresponding to the current sampling object based on the association weight information; a sampling object updating unit, configured to update the current sampling object based on the current object sampling information; a first iterative unit configured to repeat the iterative operations of determining the current object sampling information corresponding to the current sampling object based on the association weight information and updating the current sampling object based on the current object sampling information until the current object sampling information satisfies a first preset condition; The target associated object determining unit is configured to, when the current object sampling information satisfies the first preset condition, update the current sampling object based on the current object sampling information that satisfies the first preset condition to obtain the target associated object.

8. The object pushing device according to claim 7, characterized in that: The object sampling information determining unit includes: The to-be-processed object determining unit is configured to obtain a preset number of sampling objects from the current sampling object; and use the preset number of sampling objects as first to-be-processed objects, and use the other sampling objects in the current sampling object except the first to-be-processed objects as second to-be-processed objects: The second iterative unit is configured to repeatedly perform an iterative operation until the first object to be processed meets a second preset condition, and the second iterative unit includes: A first association feature determining unit is configured to determine first association feature information corresponding to the first to-be-processed object based on the association weight information; a target association feature determination unit, configured to determine target association feature information from the first association feature information; A second association feature determining unit is configured to determine second association feature information corresponding to a second object to be processed based on the association weight information; A first object to be processed updating unit is configured to update the first object to be processed and the second object to be processed based on the second association feature information and the target association feature information; The current object sampling information determining unit is configured to determine the current object sampling information based on the first object to be processed that meets the second preset condition.

9. The object pushing device according to claim 6, characterized in that: The device further comprises: An association relationship acquisition module is configured to acquire multiple association relationships corresponding to each object to be pushed; an association weight determination module, configured to determine association weight information corresponding to the plurality of association relationships; The object relationship information construction module is configured to execute, with multiple objects to be pushed as nodes, and based on the multiple association relationships and the association weight information corresponding to the multiple association relationships, to construct the edge of each node corresponding to the object to be pushed, so as to obtain the object relationship information.

10. The object pushing device according to claim 6, characterized in that: The object push module includes: a push request response unit configured to execute a push request corresponding to the object to be pushed, and obtain first target aggregate information corresponding to the object to be pushed and second target aggregate information corresponding to other objects among the objects to be pushed except the object to be pushed; a similarity determining unit, configured to determine a similarity between the first target aggregate information and the second target aggregate information; a target push object determining unit, configured to determine a target push object corresponding to the to-be-pushed object based on the similarity; The push operation execution unit is configured to execute pushing of the target push object to the object to be pushed.

11. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the object pushing method according to any one of claims 1 to 5.

12. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the object pushing method according to any one of claims 1 to 5.

13. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the object pushing method according to any one of claims 1 to 5 is implemented.

Images