Information pushing method, device and equipment and computer storage medium

Abstract

The embodiment of the invention provides an information pushing method and device, equipment and a computer storage medium. The method comprises the steps of obtaining interest attribute information corresponding to a to-be-pushed object, wherein the interest attribute information comprises at least one of associated image information, object brand information, object function information and object style information; based on the interest attribute information, a pushed object corresponding to the to-be-pushed object is determined, and the similarity between the pushed object and the to-be-pushed object is larger than or equal to a preset threshold value; based on the interested crowd corresponding to the pushed object, determining a target pushing crowd corresponding to the to-be-pushed object; and pushing information to the target pushing crowd based on the to-be-pushed object. In the embodiment, the similar objects are effectively determined through the interest attribute information, the target pushing crowd is determined based on the interested crowd corresponding to the similar objects, and then the information is pushed to the target pushing crowd, so that the quality and effect of the information pushing operation are effectively ensured.

Description

Technical Field

[0001] This application relates to the field of information processing technology, and in particular to an information push method, apparatus, device, and computer storage medium. Background Technology

[0002] Push notifications refer to message notifications sent to users' mobile phones or terminal devices. Push notifications include strong-reach and non-strong-reach types. Strong-reach push notifications are characterized by being highly proactive (not triggered by the user) and having strong penetration (the user can receive the notification even when the application is not open).

[0003] However, pushing low-quality information often provokes user resentment, potentially causing users to disable their ability to receive push notifications. Therefore, how to deliver high-quality information to the target audience is a pressing technical problem that needs to be solved. Summary of the Invention

[0004] This application provides an information push method, apparatus, device, and computer storage medium, which can accurately push high-quality information to the public, ensuring the quality and effectiveness of the information push.

[0005] In a first aspect, embodiments of the present invention provide an information push method, including: Obtain interest attribute information corresponding to the object to be pushed, wherein the interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information; Based on the interest attribute information, a previously pushed object is determined that corresponds to the object to be pushed, and the similarity between the previously pushed object and the object to be pushed is greater than or equal to a preset threshold. Based on the interested groups corresponding to the already pushed objects, determine the target push group corresponding to the objects to be pushed; Information is pushed to the target audience based on the object to be pushed to.

[0006] Secondly, embodiments of the present invention provide an information push device, comprising: The first acquisition module is used to acquire interest attribute information corresponding to the object to be pushed, and the interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information; The first determining module is used to determine, based on the interest attribute information, the already pushed object corresponding to the object to be pushed, wherein the similarity between the already pushed object and the object to be pushed is greater than or equal to a preset threshold. The first determining module is further configured to determine the target push audience corresponding to the object to be pushed based on the interest audience corresponding to the already pushed object; The first processing module is used to push information to the target audience based on the object to be pushed to.

[0007] Thirdly, embodiments of the present invention provide an electronic device, including: a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the method in the first aspect described above.

[0008] Fourthly, embodiments of the present invention provide a computer storage medium for storing a computer program, which, when executed by a computer, implements the method described in the first aspect above.

[0009] Fifthly, embodiments of the present invention provide a computer program product, comprising: a computer-readable storage medium storing computer instructions, which, when executed by one or more processors, cause one or more processors to perform the steps of the method in the first aspect described above.

[0010] The information push method, apparatus, device, and computer storage medium provided in this embodiment obtain interest attribute information corresponding to the object to be pushed, determine the already pushed objects corresponding to the object to be pushed based on the interest attribute information, and then determine the target push audience corresponding to the object to be pushed based on the interested groups corresponding to the already pushed objects. This effectively realizes the segmentation of the target push audience interested in the object to be pushed by the interested groups corresponding to similar already pushed objects. Moreover, the similar already pushed objects are determined based on the interest attribute information of the object to be pushed, which effectively ensures the accuracy and reliability of the target push audience segmentation. In this way, when pushing information to the target push audience based on the object to be pushed, the quality and effect of the information push operation are effectively guaranteed, and the probability of users disliking the information push and turning off the information push receiving function due to the push operation of low-quality information is reduced, thereby effectively ensuring the practicality of the method. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A schematic diagram illustrating a scenario of an information push method provided for an exemplary embodiment of this application; Figure 2 A flowchart illustrating an exemplary embodiment of this application for an information push method; Figure 3 A schematic diagram illustrating the process of obtaining interest attribute information corresponding to the object to be pushed, provided as an exemplary embodiment of this application; Figure 4 This is a flowchart illustrating a process for determining a target audience corresponding to a target object based on the audience of interest corresponding to the already pushed object, as provided in an exemplary embodiment of this application. Figure 5 This is a flowchart illustrating a process for determining a target audience corresponding to a target object based on the audience of interest corresponding to the already pushed object, as provided in an exemplary embodiment of this application. Figure 6 A flowchart illustrating an information push method provided in an exemplary application embodiment of this application; Figure 7 A system schematic diagram of an information push method provided in an exemplary application embodiment of this application; Figure 8 A schematic diagram of the structure of an information push device provided for an exemplary embodiment of this application; Figure 9 A schematic diagram of the structure of an electronic device provided for an exemplary embodiment of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] It should be noted that, in the case of user information involved in the embodiments of this application, 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 analysis, data stored, data displayed, etc.) involved in the embodiments of this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0014] The various models involved in this application (including but not limited to language models or large models) comply with relevant laws and standards. Furthermore, the technical solutions provided in the embodiments of this application can employ deep learning models with relatively large parameter scales. The large model is merely an example, and the embodiments of this application do not limit the number of model parameters supported by the deep learning model used, aiming to meet actual needs. The deep learning models involved in the embodiments of this application can be artificial intelligence-based language models (LM) or multimodal models (MM).

[0015] Additionally, it should be noted that when user interaction operations or triggering operations are involved in the embodiments of this application, these operations include, but are not limited to, various interaction methods such as touch operations, gesture operations, voice operations, head movement operations, and eye movement operations. Touch operations include, but are not limited to, click operations, double-click operations, long-press operations, swipe operations, pinch operations, or mouse hover operations. Swipe operations include, but are not limited to, straight-line swipes and curved-line swipes.

[0016] To facilitate understanding of the information push method, apparatus, device, and computer storage medium provided in the embodiments of this application, the relevant technologies are briefly described below: Push notifications (PUSH function) refer to message notifications sent to a user's mobile phone or terminal device. Push notifications include both aggressive and passive methods. Aggressive push notifications are characterized by their strong initiative (not triggered by the user) and high penetration (the user receives the notification even when the application is not open). Because of this, pushing low-quality content often annoys users, increasing the likelihood that they will disable the push notification function.

[0017] For example, in e-commerce platforms, when the predicted click-through rate (PCTR) of push notifications is between 0.4% and 0.6%, research shows that users will report problems such as too many marketing messages, low quality, and lack of interest, resulting in an increase in the number of users turning off the push function year by year.

[0018] Leveraging the powerful reach of push notifications, for e-commerce platforms, the incubation, growth, and explosive success of new products are crucial components of increasing Gross Merchandise Value (GMV). However, in the push marketing domain, new and trending products often lack sales figures, reviews, exposure, and historical user behavior data. Therefore, it's difficult to accurately target the right audience from the outset, making it challenging to transform push notifications into engaging and visible messages for users.

[0019] Currently, the following three methods are often used in related technologies to target the audience for new products and cutting-edge goods: Related Technology 1: Manually segmenting the target audience based on user tags. Specifically, operations and maintenance personnel use their experience to determine the target audience for new products, which can be done through a user tagging system.

[0020] However, the above implementation method has the following drawbacks: 1) Low efficiency: It relies on manual selection, and each product needs to be selected individually; 2) High uncertainty and cannot be scaled: The judgment criteria of operation and maintenance personnel on the new product audience profile are different, and they rely heavily on personal experience, which leads to uncertainty in the selection effect; 3) Long process: The entire operation and maintenance chain needs to add operation and maintenance nodes, which can increase the complexity of the entire operation and maintenance operation.

[0021] Related technology 2: Using the inverse dual-tower algorithm recommendation model for audience segmentation. Specifically, the dual-tower algorithm recommendation model generates user vectors and product vectors. It is generally used for user-to-item (U2I) recommendation operations, that is, recalling nearby product vectors through user vectors. The inverse method is to recall nearby users through product vectors to produce the target audience.

[0022] However, the above implementation method has the following drawbacks: 1) High complexity of the link: It is necessary to add an algorithm engineering link to the current recommendation link, including development, operation and maintenance iteration; 2) The effect is not necessarily good: The accuracy of user identification depends on the iterative operation of the algorithm model, while the current user recommendation operation of products in the PUSH product domain requires a long period of accumulation; 3) High interpretation cost: The user identification results through deep learning depend on feature input and vector output, and the recommendation attribution is relatively vague.

[0023] Related technology 3: Using large language models (LLM) for audience segmentation. Specifically, the large language model is trained through fine-tuning, distillation, etc., and user profiles and product data are input. The large language model then outputs a structured response to user-product interest questions to determine the interested audience.

[0024] However, the above implementation method has the following drawbacks: 1) Low efficiency: Each product needs to be judged for each of the full users. The call delay of the large language model and the superimposed user scale will lead to low efficiency in finding the audience; 2) High cost of improving accuracy: Simple general model and prompt word Prompts project generally cannot obtain a relatively accurate result. Model fine-tuning is often required, which requires graphics card resources and the algorithm background of large models, thereby increasing the model inference cost.

[0025] To address the aforementioned technical problems, embodiments of this application provide an information push method, apparatus, device, and computer storage medium, as detailed in the appendix. Figure 1 As shown, the execution entity of this information push method can be an information push device 200, which can be implemented as a local server, a cloud server, or an edge server. When the information push device 200 is implemented as a cloud server, the information push method can be executed in the cloud. Several computing nodes (cloud servers) can be deployed in the cloud, each with computing, storage, and other processing resources. In the cloud, multiple computing nodes can be organized to provide a certain service; of course, a single computing node can also provide one or more services. The cloud can provide this service by providing a service interface, which users call to use the corresponding service. Service interfaces include Software Development Kits (SDKs) and Application Programming Interfaces (APIs).

[0026] The information push device 200 is communicatively connected to the client 100, which is used by the user to trigger information push operations. The client 100 can be any computing device with a certain information interaction capability. Specifically, the client 100 can be a mobile phone, a personal computer (PC), a tablet computer, a settings application, etc. Furthermore, the basic structure of the client 100 may include at least one processor. The number of processors depends on the client's configuration and type. The client 100 may also include memory, which can be volatile, such as Random Access Memory (RAM), or non-volatile, such as Read-Only Memory (ROM), flash memory, etc., or both types. The memory typically stores the operating system (OS), one or more applications, and may also store program data. In addition to the processing unit and memory, the client 100 also includes some basic configurations, such as a network card chip, an I / O bus, a display component, and some peripheral devices. Optionally, some peripheral devices may include, for example, a keyboard, a mouse, a stylus, a printer, etc. Other peripheral devices are well known in the art and will not be described in detail here.

[0027] Information push device 200 refers to a device capable of performing information push operations in a network virtual environment, typically referring to a device that utilizes a network for information planning and push operations. Specifically, information push device 200 can be an information push model used to implement information push operations. Physically, information push device 200 can be any device capable of providing computing services and performing corresponding information push operations, such as processors, servers, etc. The main components of information push device 200 include a processor, hard disk, memory, system bus, etc., and its architecture is similar to that of a general-purpose computer.

[0028] In the above embodiment, the information push device 200 and the client 100 are connected via a network, which can be a wireless or wired network connection. If the information push device 200 and the client 100 are connected via a communication connection, the network standard of the mobile network can be any one of 2G (Global System for Mobile Communications GSM), 2.5G (General Packet Radio Service GPRS), 3G (Wideband Code Division Multiple Access (WCDMA), Time Division Synchronous Code Division Multiple Access (TD-SCDMA), 4G (Long Term Evolution LTE), 4G+ (Enhanced Long Term Evolution LTE+), Global Microwave Access Interoperability (WiMax), 5G, 6G, etc.

[0029] In this embodiment, the client 100 is used by the user to generate an information push request or information push task to trigger an information push operation. The information push request or information push task can be generated based on human-computer interaction. In order to realize the information push operation, the information push request or information push task can be sent to the information push device 200 so that the information push device 200 can perform the corresponding information push operation based on the information push request or information push task.

[0030] The information push device 200 is used to obtain information push requests or information push tasks sent by the client 100, and then obtain interest attribute information corresponding to the object to be pushed based on the obtained information push request or information push task. The object to be pushed can be a new product to be released or a high-quality product. The number of interest attribute information can be one or more, which can specifically include at least one of the following: associated image information, object brand information, object function information, and object style information. The associated image information can be celebrity IP information, which can include at least one of the following: celebrity identity information, celebrity characteristics (appearance, professional ability, personality charm, etc.), representative works (film and television works, music works, etc.), fan information (fan name, number of fans, etc.).

[0031] After obtaining the interest attribute information, the corresponding pushed objects can be determined based on the interest attribute information. The pushed objects can be objects that have been published or announced. Furthermore, the similarity between the pushed objects and the objects to be pushed is greater than or equal to a preset threshold, that is, the pushed objects and the objects to be pushed are similar objects.

[0032] Since the pushed objects are already published or announced, and the similarity between the pushed objects and the objects to be pushed is high, the target audience corresponding to the objects to be pushed can be determined based on the interested groups corresponding to the pushed objects. This target audience is the adapted group that is interested in the objects to be pushed. Then, information can be pushed to the target audience based on the objects to be pushed, which effectively ensures the accuracy and reliability of the information push operation.

[0033] In this embodiment, the target audience for the object to be pushed is effectively defined by identifying the interested groups corresponding to similar pushed objects. Furthermore, the similar pushed objects are determined based on the interest attribute information of the object to be pushed, which effectively ensures the accuracy and reliability of the target audience definition. In this way, when pushing information to the target audience based on the object to be pushed, the quality and effect of the information push operation are effectively guaranteed, reducing the probability of users disliking the information push and turning off the information push receiving function due to low-quality information push operations, thereby effectively ensuring the practicality of the method.

[0034] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0035] Figure 2 A flowchart illustrating an exemplary embodiment of this application for an information push method; see attached diagram. Figure 2 As shown, this embodiment provides an information push method. The execution subject of this method is an information push device, which can be implemented as software or a combination of software and hardware. When the information push device is implemented as hardware, it can be various electronic devices capable of performing information push operations. In some instances, the information push device can be implemented as an application client, server, cloud server, etc. When the information push device is implemented as software, it can be installed in the electronic devices listed above. Specifically, the information push method provided in this embodiment may include: Step S201: Obtain the interest attribute information corresponding to the object to be pushed. The interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information.

[0036] Step S202: Based on interest attribute information, determine the already pushed objects corresponding to the object to be pushed, and the similarity between the already pushed objects and the object to be pushed is greater than or equal to a preset threshold.

[0037] Step S203: Based on the interested audience corresponding to the already pushed objects, determine the target audience corresponding to the objects to be pushed.

[0038] Step S204: Push information to the target audience based on the target object.

[0039] The specific implementation methods and principles of each of the above steps are explained in detail below: Step S201: Obtain the interest attribute information corresponding to the object to be pushed. The interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information.

[0040] When there is a need for information push, the information push device can obtain interest attribute information corresponding to the target audience. The target audience can refer to an unpublished or actively promoted product / service. In different scenarios, this can take different forms. For example, in e-commerce, the target audience could be unpublished products (new products) or products that require active promotion (premium products). In article publishing, the target audience could be unpublished articles, papers, etc.; in video publishing, the target audience could be unpublished video information, animated images, and pictures, etc.

[0041] In addition, the interest attribute information corresponding to the target audience is used to characterize the tags or features of the target audience in which interest dimensions might attract users. This can include not only the inherent attributes of the target audience itself (e.g., brand information, price information), but also attribute information abstracted from user behavior and preferences to reflect the matching relationship between the target audience and the user. In some instances, when the target audience is a product to be published, the interest attribute information corresponding to the product can include at least one of the following: associated image information, target brand information, target function information, target style information, etc. The aforementioned associated image information can include celebrity IP information. Celebrity IP information refers to the integration and development of a celebrity's personal image, fan influence, and other elements into an intellectual property brand with unique market potential. Simply put, it is the brand asset formed by a celebrity through their own characteristics (such as professional ability, personal charm) and the fame accumulated from their works. In some instances, celebrity IP information typically includes at least one of the following: celebrity identity information, celebrity characteristics (appearance, professional ability, personal charm, etc.), representative works (film and television works, music works, etc.), and fan information (fan names, number of fans, etc.).

[0042] Furthermore, this embodiment does not limit the specific method of obtaining the interest attribute information corresponding to the object to be pushed. In some instances, the interest attribute information corresponding to the object to be pushed can be obtained through human-computer interaction. In this case, obtaining the interest attribute information corresponding to the object to be pushed may include: displaying the human-computer interaction interface; obtaining the execution operation input by the user in the human-computer interaction interface; and obtaining the interest attribute information corresponding to the object to be pushed based on the execution operation. This effectively ensures the accuracy and reliability of obtaining the interest attribute information.

[0043] Step S202: Based on interest attribute information, determine the already pushed objects corresponding to the object to be pushed, and the similarity between the already pushed objects and the object to be pushed is greater than or equal to a preset threshold.

[0044] Since the target audience for push notifications is unpublished or has never been pushed before, it lacks historical data such as sales figures, reviews, exposure, and user behavior. To accurately identify the target audience interested in the target audience, we can first determine similar objects in terms of interest attribute related terms. Therefore, after obtaining the interest attribute information corresponding to the target audience, we perform an object search based on this information to determine the previously pushed objects. A similarity between the target audience and the pushed object is defined as a similarity in terms of interest attribute information if the similarity between the pushed object and the previously pushed object is greater than or equal to a preset threshold.

[0045] In some instances, the pushed objects can be determined based on the matching degree of interest attribute information. In this case, determining the pushed objects corresponding to the object to be pushed based on interest attribute information may include: determining the standard interest keywords corresponding to each of the multiple published objects in the pushed database; determining the similarity between the standard interest keywords and interest attribute information; if the similarity is greater than or equal to a preset threshold, determining the published object corresponding to the similarity as the pushed object corresponding to the object to be pushed; if the similarity is less than the preset threshold, determining the published object corresponding to the similarity as not the pushed object corresponding to the object to be pushed. This ensures the accuracy and reliability of determining the pushed objects.

[0046] In other instances, the pushed objects can be determined not only directly based on the matching degree of interest attribute information, but also, when there are multiple interest attribute information, by combining the priority of interest attribute information. In this case, determining the pushed objects corresponding to the object to be pushed based on interest attribute information can include: obtaining multiple standard interest keywords of a published object in the pushed database; determining the priority of multiple interest attribute information; determining the overall similarity between multiple standard interest keywords and multiple interest attribute information based on the priority of multiple interest attribute information. The overall similarity can be obtained by weighted summation of the similarities between each standard interest keyword and interest attribute information based on the weight information corresponding to the priority; if the overall similarity is greater than or equal to a preset threshold, the published object determined by the overall similarity is determined as the pushed object corresponding to the object to be pushed; if the overall similarity is less than the preset threshold, the published object determined by the overall similarity is determined as not the pushed object corresponding to the object to be pushed. This ensures the accuracy and reliability of determining the pushed objects.

[0047] Step S203: Based on the interested audience corresponding to the already pushed objects, determine the target audience corresponding to the objects to be pushed.

[0048] Since the pushed objects and the objects to be pushed are similar in terms of interest attributes, it means that the target audience of the pushed objects and the target audience of the objects to be pushed overlap or are similar. Therefore, after determining the pushed objects corresponding to the objects to be pushed, we can first determine the target audience corresponding to the pushed objects. The target audience can be determined based on the user profile and object attribute characteristics of the pushed objects. Then, we can analyze and process the target audience corresponding to the pushed objects to determine the target audience corresponding to the objects to be pushed. The target audience can include multiple target users, and target users can refer to users who are inferred to be interested in the objects to be pushed.

[0049] In some instances, the target audience corresponding to the already pushed object is directly determined as the target audience corresponding to the object to be pushed. Alternatively, a portion of the target audience corresponding to the already pushed object is determined as the target audience corresponding to the object to be pushed. In this case, determining the target audience corresponding to the object to be pushed based on the target audience corresponding to the already pushed object may include: when there are multiple interest attribute information, determining the priority of each of the multiple interest attribute information; and determining the target audience corresponding to the object to be pushed from the target audience among the target audience based on the multiple interest attribute information and the priority of each of the multiple interest attribute information.

[0050] After obtaining multiple interest attribute information corresponding to the target audience, the priority of each interest attribute information can be determined. The priority of each interest attribute information can be pre-configured. Since different interest attribute information can correspond to different priorities, the interest attribute information with different priorities can have different degrees of impact on the targeting of the target audience. Therefore, after determining the priority of each interest attribute information, the target audience corresponding to the target audience can be determined from the interested audience based on the multiple interest attribute information and the priority of the interest attribute information. At this time, the target audience can be a part of the interested audience.

[0051] In some instances, the target audience for push notifications can be determined by analyzing and processing multiple interest attributes and their corresponding priorities using a pre-trained audience targeting model. In this case, determining the target audience corresponding to the object to be pushed to from among the interested audience can include: determining the pre-trained audience targeting model; inputting the multiple interest attributes, their corresponding priorities, and the interested audience into the audience targeting model for analysis and processing to obtain the target audience output by the audience targeting model. This effectively ensures the accuracy and reliability of determining the target audience corresponding to the object to be pushed to.

[0052] In other instances, the target audience can be determined not only by analyzing and processing multiple interest attributes and their corresponding priorities using a pre-trained audience targeting model, but also by determining the target audience based on the recommended audiences corresponding to each of the multiple interest attributes. In this case, determining the target audience corresponding to the object to be pushed to from among the interested audiences based on multiple interest attributes and their respective priorities can include: determining the recommended audiences corresponding to each of the multiple interest attributes from among the interested audiences; and determining the target audience corresponding to the object to be pushed to based on the recommended audiences corresponding to each of the multiple interest attributes and their respective priorities.

[0053] After obtaining multiple interest attribute information, the recommended audience corresponding to each of the multiple interest attribute information can be determined from the interested population. Different interest attribute information can correspond to different recommended audiences, and different interest attribute information can correspond to different priorities. Therefore, different recommended audiences can also correspond to different priorities. Then, the recommended audiences corresponding to each of the multiple interest attribute information and the priorities corresponding to each of the multiple interest attribute information can be analyzed and processed to determine the target push audience corresponding to the object to be pushed.

[0054] Specifically, determining the target audience for push notifications, based on the recommended user groups corresponding to each of the multiple interest attributes and their respective priorities, can include: determining parameters to limit the size of the target audience; determining the weights of each interest attribute based on its priority; and determining the number of users in the corresponding recommended user group based on the weights and parameters. Generally, higher-priority interest attributes correspond to a larger number of users in their recommended user group, while lower-priority interest attributes correspond to a smaller number of users. After determining the recommended user groups for each of the multiple interest attributes, all these groups can be identified as the target audience for push notifications, effectively ensuring the accuracy and reliability of the target audience determination.

[0055] Step S204: Push information to the target audience based on the target object.

[0056] After identifying the target audience, information can be pushed to them based on the target recipients. Specifically, push information corresponding to the target recipients can be generated and sent to the terminals of the target users included in the target audience. This achieves stable information push operation.

[0057] Furthermore, during the process of performing the information push operation in accordance with the above embodiments, push link information corresponding to the object to be pushed can be recorded. When the object to be pushed is a product to be pushed, the push link information may include at least one of the following: product transaction information, push information execution record, interest attribute information search record, confirmation record of pushed products, etc.; and then the push link information can be stored in a preset area, which makes it convenient to call and view the push link information.

[0058] The information push method provided in this embodiment obtains interest attribute information corresponding to the object to be pushed, determines the previously pushed objects corresponding to the object to be pushed based on the interest attribute information, and then determines the target push audience corresponding to the object to be pushed based on the interested groups corresponding to the previously pushed objects. This effectively achieves the targeting of the target push audience based on the interested groups corresponding to similar previously pushed objects. Furthermore, since the similar previously pushed objects are determined based on the interest attribute information of the object to be pushed, the accuracy and reliability of the target push audience targeting are effectively guaranteed. In this way, when pushing information to the target push audience based on the object to be pushed, the quality and effect of the information push operation are effectively guaranteed, and the probability of users disliking the information push and turning off the information push receiving function due to the push operation of low-quality information is reduced, thereby effectively ensuring the practicality of the method.

[0059] Figure 3 This application provides a schematic diagram of a process for obtaining at least one interest attribute information corresponding to an object to be pushed, as an exemplary embodiment of the present application; based on the above embodiment, refer to the appendix. Figure 3 As shown, the interest attribute information corresponding to the target object can be obtained not only through human-computer interaction but also through the reasoning and understanding capabilities of a multimodal model. Therefore, obtaining the interest attribute information corresponding to the target object can include: Step S301: Obtain the object identifier information of the object to be pushed.

[0060] To accurately obtain the interest attribute information corresponding to the object to be pushed, the object identification information of the object can be obtained. This object identification information is used as a unique identifier to distinguish the object from other objects. For example, when the object to be pushed is a product, the object identification information can be the product ID information; when the object to be pushed is an article, the object identification information can be the article identification information.

[0061] Step S302: Based on the object identification information, determine the object attribute information corresponding to the object to be pushed.

[0062] After obtaining the object identifier information of the object to be pushed, the corresponding object attribute information can be determined based on this identifier. Object attribute information may include at least one of the following: object characteristics, object specifications, object purpose, object physical / chemical characteristics, etc. In different application scenarios, different objects to be pushed may correspond to different object attribute information. For example, in e-commerce applications, the object to be pushed may be a product, and the object attribute information is the product attribute information, which may include at least one of the following: product title, product name, product purpose, product brand, product price, and product description, etc. In information publishing applications, the object to be pushed may be an article, and the object attribute information is the article attribute information, which may include at least one of the following: article title, article author information, article summary, article keywords, article category, article tags, etc.

[0063] For the object attribute information corresponding to the object to be pushed, it can be determined by information extraction operation through the mapping relationship between object identification information and object attribute information. At this time, determining the object attribute information corresponding to the object to be pushed based on the object identification information may include: determining the data center or database used to store the object attribute information; extracting information from the data center or database based on the object identification information of the object attribute information, so as to reliably obtain the object attribute information corresponding to the object to be pushed.

[0064] Step S303: Use a multimodal model to understand the object's attribute information and determine the interest attribute information.

[0065] Since object attribute information can reflect the preference attributes of the object to be pushed to a certain extent, that is, there is a correlation between object attribute information and interest attribute information, after determining the object attribute information corresponding to the object to be pushed, a multimodal model can be used to understand the object attribute information to determine the interest attribute information. The interest attribute information may include at least one of the following: associated image information, object brand information, object function information, object style information, etc. Among them, associated image information can refer to image information that is associated with the object to be pushed. For example, when the object to be pushed is a product, associated image information can refer to any one of the following: IP information of the celebrity who endorses the product, IP information of the anime character associated with the product, IP information of the virtual character associated with the product, etc.

[0066] The multimodal model involved in this application embodiment can be a Large Language Model (LLM) based on artificial intelligence. This application embodiment does not limit the number of model parameters supported by the model, aiming to meet actual needs. If the model has relatively more parameters, the model size will be relatively larger, and the model performance will be relatively better. Of course, more time and resources will be consumed during inference or training. If the model has relatively fewer parameters, the model size will be relatively smaller. While meeting performance requirements, the model is more lightweight, and consumes relatively less time and resources during inference or training. This multimodal model can be a deep learning model used to process and generate natural language text or multimodal data. It can be implemented based on a neural network architecture and can be pre-trained on large amounts of data. In an optional implementation, the multimodal model can include an encoder, a decoder, a self-attention layer, and a feed-forward neural network, etc. The encoder is mainly used to convert input data (usually in sequence form) into vector representation. This process can capture the semantic features of the input data. The decoder is responsible for converting the intermediate representation generated by the encoder into output data (usually in sequence form). The self-attention layer is a mechanism that allows the model to pay attention to other positions in the sequence to better encode the current position information. The feedforward neural network can perform nonlinear transformations on the output of the self-attention layer to enhance the model's expressive power. All parts work together, enabling the model built on them to perform well in various complex processing tasks, such as natural language processing, computer vision, speech recognition, machine translation, text summarization, and intelligent question answering.

[0067] In some instances, interest attribute information can be directly determined based on the understanding and reasoning results of object attribute information by a multimodal large language model. In this case, using a multimodal model to understand object attribute information and determine interest attribute information may include: inputting object attribute information into a multimodal model for parsing to obtain interest attribute information directly output by the multimodal model.

[0068] In other instances, in order to obtain a more comprehensive target audience, interest attribute information can be obtained by generalizing object attribute information. In this case, using a multimodal model to understand object attribute information and determine interest attribute information may include: generalizing object attribute information to obtain multiple related attribute information, where the correlation between related attribute information and object attribute information is greater than or equal to a preset threshold; and using a multimodal model to understand object attribute information and multiple related attribute information to determine interest attribute information.

[0069] After obtaining the object attribute information, generalization processing can be performed on the object attribute information to obtain multiple related attribute information. In some instances, the generalization operation of object attribute information can be achieved through feature abstraction operations. In this case, generalizing the object attribute information to obtain multiple related attribute information may include: determining the feature information of the object attribute information; generalizing the feature information of the object attribute information to higher-level category information, thereby obtaining multiple related attribute information. The correlation between the obtained related attribute information and the object attribute information is greater than or equal to a preset threshold, which effectively ensures the accuracy and reliability of obtaining multiple related attribute information.

[0070] After obtaining multiple related attribute information, the object attribute information and multiple related attribute information can be input into the multimodal model so that the multimodal model can perform reasoning and parsing on the multiple related attribute information and object attribute information, thereby determining the interest attribute information output by the multimodal model. The number of interest attribute information can be one or more.

[0071] For object attribute information, the generalization operation can be achieved not only through feature abstraction operations, but also based on the hotspot knowledge graph corresponding to the object attribute information. In this case, generalizing the object attribute information to obtain multiple related attribute information can include: determining the hotspot knowledge graph corresponding to the object attribute information, which includes multiple nodes and edges between two nodes. Nodes are used to represent the current popular information, and edges are used to represent the correlation between popular information; generalizing the object attribute information based on the hotspot knowledge graph to obtain the related hotspot information corresponding to the object attribute information; and determining multiple related attribute information based on the related hotspot information.

[0072] To ensure the quality and effectiveness of generalization operations based on object attribute information, a hot topic knowledge graph corresponding to the object attribute information can be determined first. This hot topic knowledge graph includes multiple nodes, each representing current trending information. This trending information can include at least one of the following: current trending topics, current trending events, current trending figures, current trending technologies, or current trending trends, etc. Because hot topic knowledge graphs are highly timely and dynamically evolving, they can be determined based on real-time analysis of hot topic data. In this case, determining the hot topic knowledge graph corresponding to the object attribute information can include: identifying hot topic data sources (news media platforms, social media platforms, academic platforms, etc.); obtaining hot topic data corresponding to the object attribute information through these data sources; and determining the hot topic knowledge graph corresponding to the object attribute information based on the obtained hot topic data. This effectively ensures the timeliness and accuracy of determining the hot topic knowledge graph.

[0073] In other instances, the hotspot knowledge graph can be pre-built and stored in a preset area. The preset area stores multiple knowledge graphs. In this case, the hotspot knowledge graph corresponding to the object attribute information is determined by searching the preset area through the object attribute information. This hotspot knowledge graph is the knowledge graph stored in the preset area that has a relationship with the object attribute information. This also ensures the stability and reliability of determining the hotspot knowledge graph.

[0074] After determining the hotspot knowledge graph, object attribute information can be generalized based on the hotspot knowledge graph to obtain associated hotspot information corresponding to the object attribute information. This associated hotspot information consists of hotspot data in the hotspot knowledge graph that are related to the object attribute information. It can be determined by the degree of association between the nodes in the hotspot knowledge graph and the object attribute information. Specifically, if the degree of association between the node and the object attribute information is greater than or equal to a preset threshold, associated hotspot information is determined based on the information corresponding to that node; if the degree of association between the node and the object attribute information is less than the preset threshold, the node is ignored.

[0075] After identifying the relevant hotspot information, multiple related attribute information can be determined based on the relevant hotspot information. In some instances, the relevant hotspot information is directly identified as multiple related attribute information, which effectively enables the stable identification of multiple related attribute information based on the hotspot knowledge graph.

[0076] In other instances, multiple related attribute information may include not only related hotspot information determined based on hotspot knowledge graphs, but also search interest information obtained from searching on a preset platform. In this case, determining multiple related attribute information based on related hotspot information may include: determining search keywords corresponding to object attribute information; searching on a preset platform based on search keywords to obtain search interest information associated with object attribute information; and determining multiple related attribute information based on related hotspot information and search interest information.

[0077] The multiple related attribute information can include not only trending information outside the preset platform, but also search interest information or trending search information within the preset platform. In this case, the search keywords corresponding to the object attribute information can be determined first. These search keywords can be determined through information extraction or summary analysis of the object attribute information. Then, a search can be performed on the preset platform based on the search keywords. Specifically, the search keywords can be entered into the preset platform to determine the drop-down list corresponding to the search keywords. This drop-down list includes popular search terms related to the search keywords within the preset platform, thus reliably obtaining the search interest information corresponding to the object attribute information. This search interest information is the interest information related to the search keywords found on the preset platform.

[0078] After obtaining search interest information and related hot topic information, multiple related attribute information can be determined based on the related hot topic information and search interest information. In some instances, the related hot topic information and search interest information can be summarized and deduplicated to obtain multiple related attribute information. This effectively ensures the accuracy and reliability of determining multiple related attribute information, thereby facilitating the improvement of the accuracy of determining interest attribute information based on multiple related attribute information.

[0079] In this embodiment, by obtaining the object identification information of the object to be pushed, the object attribute information corresponding to the object to be pushed is determined based on the object identification information, and the multimodal model is used to understand the object attribute information to determine the interest attribute information. This effectively ensures the accuracy and reliability of determining the interest attribute information.

[0080] Figure 4 This is a flowchart illustrating an exemplary embodiment of the present application, showing how to determine the target audience corresponding to the object to be pushed to, based on the audience of interest corresponding to the already pushed object; based on the above embodiment, refer to the appendix. Figure 4 As shown, the target audience for push notifications can be determined not only by a pre-trained audience segmentation model, but also by combining the priorities of multiple interest attributes. In this embodiment, determining the target audience corresponding to the object to be pushed to, based on the interest groups corresponding to the already pushed objects, can include: Step S401: When there are multiple interest attribute information, determine the priority of each of the multiple interest attribute information.

[0081] The number of interest attribute information can be one or more. When there are multiple interest attribute information, different interest attribute information can reflect the user's degree of interest in the push object to varying degrees. Therefore, in order to accurately determine the target push audience corresponding to the push object, the priority of each of the multiple interest attribute information can be determined first. The priority of interest attribute information is used to identify the degree of influence of interest attribute information on the target push audience. The higher the priority of interest attribute information, the higher the degree of influence of interest attribute information on the target push audience; the lower the priority of interest attribute information, the lower the degree of influence of interest attribute information on the target push audience.

[0082] In some instances, the priority of each of the multiple interest attribute information can be a pre-configured parameter, or the priority of each of the multiple interest attribute information can be determined based on the input sorting information of the multiple interest attribute information. Specifically, the input sorting information of the multiple interest attribute information can be determined first, and then the priority of each of the multiple interest attribute information can be determined based on the input sorting information. The earlier the sorting position in the input sorting information, the higher the priority of the interest attribute information; the later the sorting position in the input sorting information, the lower the priority of the interest attribute information.

[0083] Alternatively, the priority of each of the multiple interest attributes can be determined based on human-computer interaction. Specifically, the human-computer interaction interface can be displayed first; the priority configuration operation entered by the user in the human-computer interaction interface can be obtained; and the priority of each of the multiple interest attributes can be determined based on the priority configuration operation. This effectively ensures the accuracy and reliability of determining the priority of each interest attribute.

[0084] Step S402: Based on multiple interest attribute information and the priority of each interest attribute information, determine the target push audience corresponding to the object to be pushed to from the interested audience.

[0085] After determining the priority of each of the multiple interest attribute information, the multiple interest attribute information and their respective priorities can be analyzed and processed to identify the target audience corresponding to the object to be pushed to from the interested population.

[0086] In some instances, the target audience for push notifications can be determined based on a pre-trained audience targeting model. In this case, determining the target audience corresponding to the object to be pushed to from among the interested audience, based on multiple interest attribute information and the priorities corresponding to each of the interest attribute information, can include: determining the pre-trained audience targeting model; inputting multiple interest attribute information, the priorities corresponding to each of the interest attribute information, and the interested audience into the audience targeting model for processing, and determining the target audience output by the audience targeting model. This effectively ensures the accuracy and reliability of determining the target audience for push notifications.

[0087] In other instances, the target audience can be determined not only based on a pre-trained audience targeting model, but also based on recommended audiences corresponding to multiple interest attributes. In this case, determining the target audience corresponding to the object to be pushed to from among the interested audiences can include: determining recommended audiences corresponding to multiple interest attributes from among the interested audiences; and determining the target audience corresponding to the object to be pushed to based on the recommended audiences corresponding to multiple interest attributes and the priorities corresponding to multiple interest attributes.

[0088] For example, when multiple interest attributes include: associated image information, object brand information, object function information, and object style information, the priority of associated image information is higher than that of object brand information, which in turn is higher than that of object function information, and finally, object function information is higher than that of object style information. Then, a recommended audience can be determined from the interested population for each interest attribute. Different interest attributes correspond to different recommended audiences; for example, associated image information corresponds to recommended audience 1, object brand information corresponds to recommended audience 2, object function information corresponds to recommended audience 3, and object style information corresponds to recommended audience 4.

[0089] After determining the priority of each of the multiple interest attribute information, the target push audience corresponding to the object to be pushed can be determined based on the recommended audience corresponding to each of the multiple interest attribute information and the priority of each of the multiple interest attribute information. In some instances, one or more target attribute information is first determined based on the priority of each of the multiple interest attribute information; then the recommended audience corresponding to one or more target attribute information can be determined as the target push audience.

[0090] Implementation Method 1: Determine a target attribute based on the priority of multiple interest attributes. For example, the target attribute could be associated image information, and then the recommended audience 1 corresponding to the associated image information can be determined as the target recommended audience. If the target attribute includes both associated image information and object brand information, then the recommended audience 1 corresponding to the associated image information and the recommended audience 2 corresponding to the object brand information can be determined as the target recommended audience; that is, the target recommended audience can be the combination of recommended audience 1 and recommended audience 2.

[0091] In other instances, weight information corresponding to multiple interest attributes is determined based on their respective priorities; user segmentation is performed within the recommended user groups corresponding to the interest attributes based on the weight information to obtain the segmented user groups corresponding to each interest attribute; and the target recommended user group is determined based on the segmented user groups corresponding to each interest attribute.

[0092] Implementation Method 2: Based on the priority of each of the multiple interest attribute information, determine the corresponding weight information for each interest attribute information. For example, associated image information corresponds to weight information 1, object brand information corresponds to weight information 2, object function information corresponds to weight information 3, and object style information corresponds to weight information 4. Then, based on weight information 1 and recommended audience 1, determine the target audience 1 corresponding to associated image information; based on weight information 2 and recommended audience 2, determine the target audience 2 corresponding to object brand information; based on weight information 3 and recommended audience 3, determine the target audience 3 corresponding to object function information; and based on weight information 4 and recommended audience 4, determine the target audience 4 corresponding to object style information. Finally, based on target audience 1, target audience 2, target audience 3, and target audience 4, determine the target recommended audience. That is, the target recommended audience can be the set of target audience 1, target audience 2, target audience 3, and target audience 4, thus effectively ensuring the accuracy and reliability of determining the target recommended audience.

[0093] In this embodiment, when there are multiple interest attribute information, the priority of each of the multiple interest attribute information is determined, and then the target push audience corresponding to the object to be pushed is determined from the interested population based on the multiple interest attribute information and the priority of each of the multiple interest attribute information. This effectively ensures the accuracy and reliability of the target push audience determination.

[0094] Figure 5 This is a flowchart illustrating an exemplary embodiment of the present application, showing how to determine the target audience corresponding to the object to be pushed to, based on the audience of interest corresponding to the already pushed object; based on the above embodiment, refer to the appendix. Figure 5As shown, the target audience for push notifications can be determined not only through a pre-trained audience segmentation model but also by combining audience filtering strategies. In this embodiment, determining the target audience corresponding to the object to be pushed to, based on the interested audience corresponding to the already pushed object, can include: Step S501: Determine the audience screening strategy corresponding to the target audience.

[0095] The system includes a pre-configured audience filtering strategy capable of audience segmentation. This audience filtering strategy may include at least one of the following: an interest-based recommendation filtering strategy and a tag-based filtering strategy. The interest-based recommendation filtering strategy is used to filter audiences based on the degree of interest between users and objects, while the tag-based filtering strategy is used to filter audiences based on the matching degree between object tags and user behavior tags.

[0096] In different application scenarios, different audience filtering strategies can be configured to filter the target audience for push notifications. Therefore, in order to accurately identify the target audience corresponding to the recipient, the audience filtering strategy corresponding to the recipient can be determined first. In some instances, the audience filtering strategy can be determined based on information input operations. In this case, determining the audience filtering strategy corresponding to the recipient can include: displaying a human-computer interaction interface, wherein the interface displays a list of filtering strategies for the user to select or configure; obtaining the user's strategy selection operation in the list of filtering strategies, and determining the audience filtering strategy corresponding to the recipient. This effectively ensures the accuracy and reliability of obtaining the audience filtering strategy.

[0097] Step S502: Based on the audience filtering strategy and the interested audiences corresponding to the already pushed objects, determine the target audience corresponding to the objects to be pushed.

[0098] Implementation Method 1 can determine the target audience based on a single audience filtering strategy and the interest groups corresponding to the already pushed objects. In this case, determining the target audience corresponding to the object to be pushed, based on the audience filtering strategy and the interest groups corresponding to the already pushed objects, can include: if the audience filtering strategy includes an interest recommendation filtering strategy, determining the recommendation degree of the already pushed objects to different users in the interest groups, where the recommendation degree is used to characterize the degree of interest of users in the interest groups towards the already pushed objects; filtering within the interest groups based on the recommendation degree to determine the target audience corresponding to the object to be pushed, where the recommendation degree of users in the target audience is greater than or equal to a preset threshold.

[0099] Specifically, when the audience screening strategy includes an interest-based recommendation screening strategy, the user information of the pushed objects and the interested audience can be analyzed and processed first to determine the recommendation degree of the pushed objects to different users in the interested audience. The recommendation degree is used to characterize the degree of interest of users in the interested audience in the pushed objects. Generally, the higher the recommendation degree, the higher the user's degree of interest in the pushed objects; the lower the recommendation degree, the lower the user's degree of interest in the pushed objects.

[0100] After determining the user's level of interest in the pushed object, the system can filter users within the interested group based on the recommendation level to select a user group whose recommendation level is greater than or equal to a preset threshold. Alternatively, it can filter users within the interested group based on the recommendation level to select the top N users, where N is a pre-configured parameter used to limit the number of target push groups. This effectively determines the target push group corresponding to the object to be pushed, ensuring the accuracy and reliability of the target push group determination.

[0101] Alternatively, based on the audience filtering strategy and the interested audience corresponding to the already pushed objects, determining the target audience corresponding to the object to be pushed can include: if the audience filtering strategy includes a tag filtering strategy, determining the user behavior tags associated with the already pushed objects; and filtering among the interested audience based on the user behavior tags to determine the target audience.

[0102] Specifically, when the audience screening strategy includes a tag screening strategy, user behavior tags corresponding to the pushed targets can be determined first. User behavior tags can be used to identify the user's actual operational behavior on the product or platform. In some instances, user behavior tags can include at least one of the following: click behavior tags, browsing behavior tags, purchase behavior tags, add-to-cart behavior tags, etc. The user behavior tags can be determined by querying and processing the log server corresponding to the pushed targets.

[0103] After identifying user behavior tags associated with the target audience, users can be filtered within the target audience based on these tags to select those with a behavior matching degree greater than or equal to a preset threshold. This effectively ensures the accuracy and reliability of identifying the target audience.

[0104] Implementation method 2 can determine the target audience based on multiple audience filtering strategies and the interest groups corresponding to the already pushed objects. In this case, the recommendation level of the already pushed object to different users within the interest groups can be determined first. Then, based on the recommendation level, the interest groups are filtered to determine the target audience corresponding to the object to be pushed. Specifically, determining the target audience corresponding to the object to be pushed can include: if the audience filtering strategy also includes a tag filtering strategy, filtering the interest groups based on the recommendation level to determine a first push audience; determining user behavior tags associated with the already pushed object; filtering the interest groups based on user behavior tags to determine a second push audience, where the pushed users in the second push audience correspond to the user behavior tags; and determining the target push audience based on the first and second push audiences. The target push audience can be determined by aggregating the first and second push audiences, thus effectively ensuring the accuracy and reliability of the target push audience determination.

[0105] Alternatively, determining the target audience corresponding to the target object can be based on audience filtering strategies and the interested audiences associated with the already pushed objects. This can include: if the audience filtering strategy includes a tag filtering strategy, first determining the user behavior tags associated with the already pushed objects; then filtering within the interested audiences based on these user behavior tags to determine the target audience. Specifically, determining the target audience based on filtering within the interested audiences based on user behavior tags can include: if the audience filtering strategy also includes an interest recommendation filtering strategy, filtering within the interested audiences based on user behavior tags to determine the first target audience; determining the recommendation level of the already pushed objects to different users within the interested audiences, where the recommendation level characterizes the degree of interest of users within the interested audiences in the already pushed objects; filtering within the interested audiences based on the recommendation level to determine the second target audience; and finally, determining the target audience based on the first and second target audiences. The target audience can be determined by aggregating the first and second target audiences, thus effectively ensuring the accuracy and reliability of determining the target audience.

[0106] In this embodiment, by determining the audience filtering strategy corresponding to the target audience, and then determining the target audience corresponding to the target audience based on the audience filtering strategy and the interested audience corresponding to the already pushed object, the flexibility and reliability of determining the target audience are effectively guaranteed.

[0107] In practical applications, taking new products and top-selling products as the target audience for push notifications, this application embodiment provides a method for pushing information about products. This method can efficiently, accurately, and cost-effectively identify the target audience for a product, and then push information based on that target audience. This achieves targeted information push to specific groups interested in the product, which not only improves the relevance of the pushed content but also addresses the problems of low click-through rates and poor matching effects in traditional push notifications. This information push method may include a new product submission stage, a similar product search stage, and a task deployment stage. Specifically, the information push method may include the following steps: Step 1: Obtain the product ID of the product to be pushed and the audience filtering strategy used to implement the audience selection operation.

[0108] Among them, the products to be pushed are new products that require information push operations, and the product ID is the identification information of the product to be pushed. When it is necessary to target the audience for multiple products, the product IDs corresponding to multiple products to be pushed can be obtained at one time, so that the corresponding target user groups can be determined based on the product IDs of the products to be pushed.

[0109] For details, please refer to the appendix. Figure 6 As shown, in order to obtain the product ID corresponding to the product to be pushed, the data compensation module can first extract product information from the product center using the product ID, thereby obtaining the product attribute information of the product to be pushed. The product attribute information may include at least one of the following: product title, product image, product category, product price, and other basic information.

[0110] In some instances, product attribute information can be implemented as structured subject information. Then, a multimodal product understanding agent can be used to understand the product attribute information and obtain the interest attribute information output by the multimodal product understanding agent. The interest attribute information may include at least one of the following: celebrity IP information, product brand, product function, product style, and contextual tags.

[0111] Furthermore, to improve the accuracy and reliability of determining interest attribute information, after obtaining the product attribute information of the product to be pushed, the product attribute information can be generalized by integrating hot information from inside and outside the site to obtain multiple related attribute information. Among them, the correlation between the related attribute information and the object attribute information is greater than or equal to a preset threshold. Then, the related attribute information and the product attribute information can be integrated into the target product attribute of the product to be pushed. In this way, the multimodal product understanding agent can be used to understand the target product attribute, thereby obtaining the interest attribute information output by the multimodal product understanding agent.

[0112] One approach is to integrate external trending information to generalize product attribute information using a product search agent. This involves the following steps: acquiring external information; constructing or defining a trending knowledge graph corresponding to the product attribute information based on the external information. This graph includes multiple nodes and edges between nodes, where nodes represent current trending information and edges represent the relationships between trending information; and generalizing the product attribute information based on the trending knowledge graph to obtain corresponding external generalized information. For example, if the product attribute information includes celebrity IP information, generalization using an external trending knowledge graph can yield derivative trending information related to the celebrity IP information.

[0113] Integrating trending information within the platform to generalize product attribute information can also be achieved through a product search agent. This involves the following steps: determining search keywords corresponding to the product attribute information; searching on a pre-set platform based on the search keywords to obtain trending search information related to the product attribute information, which can be popular search terms within the platform included in the recommendation list corresponding to the search keywords; and determining generalized information within the platform based on the associated trending information and the search trending information.

[0114] After obtaining external and internal generalized information, the associated attribute information can be determined based on the external and internal generalized information. Specifically, the external and internal generalized information can be summarized and processed to obtain the associated attribute information, which effectively ensures the accuracy and reliability of determining the associated attribute information.

[0115] For example, when interest attribute information includes celebrity IP information, by performing the aforementioned generalization operation on the celebrity IP information, for instance, after generalizing the celebrity IP information, it is possible to identify [celebrity] merchandise (including [celebrity name] dolls, [celebrity name] light sticks, [celebrity name] photocards, [celebrity name] cotton dolls), [celebrity] concerts, [celebrity] dolls, etc. This can increase the recall of strongly related products, thereby helping to identify more niche fan groups, which is beneficial for improving the target audience for products.

[0116] Step 2: Search for similar products based on interest attribute information to identify similar products corresponding to the product to be pushed.

[0117] A pre-configured search service is provided for searching similar products. After obtaining interest attribute information, the search engine within the service can perform a search for similar products based on this information, outputting at least one similar product. Specifically, to ensure the accuracy and reliability of identifying similar products, the search engine can first generate query keywords based on the interest attribute information. Then, the search engine can perform a search for similar products within the pre-defined platform based on these query keywords. This method reliably identifies similar products corresponding to the product to be recommended and significantly reduces the cost of identifying similar products.

[0118] Step 3: Based on similar products and audience filtering strategies, determine the target audience corresponding to the products to be pushed.

[0119] Since the platform has already accumulated multiple historical behavioral data corresponding to similar products, and the audience screening strategy of reverse recommendation engine is a audience segmentation method implemented by a deep learning algorithm, while the audience screening strategy based on user behavior tags is an audience segmentation method based on the dimension of user behavior, in order to ensure the accuracy and reliability of the target audience for push notifications, the audience screening strategy of reverse recommendation engine and the audience screening strategy based on user behavior tags can be combined to perform audience mining operations to determine the target audience corresponding to the products to be pushed.

[0120] By reversing the audience filtering strategy of recommendation engines, a typical e-commerce recommendation engine can score products recommended to a user based on their historical behavior, obtaining a recommendation score for each product. These scores can then be ranked to determine the products the user is interested in. When similar products are found, this process can be reversed to determine the detailed recommendation scores of those similar products within the target audience, allowing for the selection of users with high recommendation scores—that is, obtaining the initial target audience obtained by reversing the recommendation engine's audience filtering strategy.

[0121] User behavior tag-based audience filtering strategy: It can obtain user behavior tags corresponding to similar products. User behavior tags can include at least one of the following: search tags, product tags, and private domain tags. The search tags can be search behaviors in the past n days; product tags can include at least one of the following: add-to-cart behavior, favorite behavior, purchase behavior, etc. in the past n days; private domain tags can include at least one of the following: store membership tags, store follow tags, etc. Then, based on user behavior tags, it can filter the interested audience to obtain the second push audience.

[0122] Then, the first and second push audiences can be aggregated to determine the target audience corresponding to the product to be pushed, thus effectively ensuring the accuracy and reliability of the target audience determination. Furthermore, since different interest attribute information represents different product attributes and user interest preferences, different interest attribute information can correspond to different target audiences. To facilitate viewing and understanding the reasoning process for the target audience, the interest attribute information, the product to be pushed, similar products, and the target audience can be structured to generate and store structured data corresponding to the information reasoning process.

[0123] Furthermore, for the same product, different interest attributes can identify different target audiences. When there are multiple interest attributes, their priorities can be defined first. In product application scenarios, since products often include celebrity endorsements, celebrity attributes are a significant feature in determining the target users of new products and have high weight, resulting in a click-through rate increase of over 100% compared to ordinary products. Therefore, to improve the click-through rate and conversion rate of products, the priority of celebrity IP information can be set higher than that of other interest attributes. That is, the priorities from high to low are: celebrity IP information, target brand information, target function information, target style information, etc. Then, the target audience can be determined based on the above interest attributes with different priorities. In some instances, the target audience can be determined using a pre-configured audience targeting model. In this case, a prompt message for audience targeting is generated based on the order of celebrity IP information, target brand information, target function information, and target style information. This prompt message can then be input into the audience targeting model for inference, thereby accurately determining the target audience corresponding to the product to be promoted.

[0124] Step 4: Push information to the target audience based on the product to be pushed.

[0125] After obtaining the target audience, information can be directly pushed to the target audience based on the product to be pushed, which can ensure the accuracy and reliability of the information push operation.

[0126] In other instances, to ensure the quality and effectiveness of push notifications, after identifying the target audience, filtering can be performed. Specifically, filtering rules are first determined, which may include at least one of the following: users uninterested in the push notification service, users who have turned off push notifications, users whose prices are incompatible with the product being pushed, users whose applicable scenarios are incompatible with the product being pushed, etc. Then, the target audience can be filtered based on these filtering rules to obtain a filtered target group. Push notifications are then sent to this filtered target group based on the product to be pushed, effectively ensuring the quality and effectiveness of the push notification operation.

[0127] Furthermore, during the aforementioned information push operation, the link tracing module can be used to collect and record relevant information from the information push to facilitate subsequent analysis and location operations. Specifically, this can include product submission information, similar product information, Agent execution information, and audience filtering information. The Agent execution information can be used as test case evaluation information. For similar product information, the identified similar product information can be scheduled or updated periodically according to the information push cycle.

[0128] The information push method provided in this application embodiment can obtain hot topic data across the entire domain by acquiring hot topic knowledge graphs from outside the site and hot topic data from the recommendation drop-down list within the site. Then, it analyzes and processes the hot topic data across the entire domain through an AI large language model to identify interest attribute information corresponding to the products to be pushed—especially including brand information, celebrity information, etc. This can solve the problem of low accuracy caused by the lack of external data in traditional algorithm models. At the same time, the reasoning process of the large model also solves the problem of poor interpretability of traditional algorithm models. Subsequently, the interest attribute information of the products to be pushed can be used to locate the target user group, thereby achieving precise location of the target user group. In addition, in the process of determining similar products corresponding to the products to be pushed, the reasoning ability of the large model is used to replace manual judgment, thereby reducing the implementation cost and complexity of information push operation for new products. At the same time, with the continuous upgrading of the large language model, search and recommendation algorithms, the targeting effect of the target user group can be dynamically optimized without repeated investment, and has the ability of "automatic evolution", which further improves the ease of use and usability of the method. This solves the problems of low manual efficiency, high uncertainty and inability to scale in related technologies, and further improves the practicality of the method.

[0129] Furthermore, in some of the processes described in the above embodiments and accompanying drawings, multiple operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or they may be executed in parallel. The operation numbers, such as 11, 12, etc., are merely used to distinguish different operations and do not represent any execution order. Additionally, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0130] Figure 8 A schematic diagram of the structure of an information push device provided for an exemplary embodiment of this application; see attached diagram. Figure 8 As shown, this embodiment provides an information push device, which is used to perform the above-mentioned... Figure 2 The information push method shown herein, specifically, the information push device may include: The first acquisition module 11 is used to acquire interest attribute information corresponding to the object to be pushed. The interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information. The first determining module 12 is used to determine, based on interest attribute information, the already pushed object corresponding to the object to be pushed, and the similarity between the already pushed object and the object to be pushed is greater than or equal to a preset threshold. The first determining module 12 is also used to determine the target push audience corresponding to the object to be pushed based on the interested audience corresponding to the already pushed object; The first processing module 13 is used to push information to the target audience based on the object to be pushed.

[0131] The information push device in this embodiment can also perform the above-described... Figures 1-7 The description of the embodiments shown is for reference only, and will not be elaborated upon here.

[0132] like Figure 9 As shown, this embodiment provides an electronic device for performing the above-described... Figure 2 The information push method shown may include an electronic device that includes a memory 24 and a processor 25.

[0133] Memory 24 is used to store computer programs and can be configured to store various other data to support operation on the electronic device. Examples of this data include instructions for any application or method used to operate on the electronic device, data structures, contact data, phone book data, messages, pictures, videos, etc.

[0134] The processor 25, coupled to the memory 24, is used to execute a computer program in the memory 24 for: acquiring interest attribute information corresponding to the object to be pushed, the interest attribute information including at least one of the following: associated image information, object brand information, object function information, and object style information; determining, based on the interest attribute information, the previously pushed objects corresponding to the object to be pushed, wherein the similarity between the previously pushed objects and the object to be pushed is greater than or equal to a preset threshold; determining, based on the interested groups corresponding to the previously pushed objects, the target push group corresponding to the object to be pushed; and pushing information to the target push group based on the object to be pushed.

[0135] Furthermore, such as Figure 9 As shown, the electronic device also includes other components such as a communication component 26, a display 27, a power supply component 28, and an audio component 29. Figure 9 The diagram only shows some components and does not mean that the electronic device includes only these components. Figure 9 The components shown. Additionally... Figure 9 The components within the center frame are optional, not mandatory, and their specific requirements depend on the product form of the work node. In this embodiment, the work node can be a terminal device such as a desktop computer, laptop computer, smartphone, or IoT device, or a server-side device such as a conventional server, cloud server, or server array. If the work node in this embodiment is implemented as a terminal device such as a desktop computer, laptop computer, or smartphone, it may include... Figure 9 The components within the center frame; if the working node in this embodiment is implemented as a server-side device such as a conventional server, cloud server, or server array, it may not include... Figure 9 The component within the center frame.

[0136] The aforementioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0137] The aforementioned communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel.

[0138] The aforementioned display includes a screen, which may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a Touch Panel, the screen can be implemented as a touchscreen to receive input signals from the user. The Touch Panel includes one or more touch sensors to sense touches, swipes, and gestures on the Touch Panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation.

[0139] The aforementioned power supply components provide power to various components within the device in which they reside. These power supply components may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power to the device in which they reside.

[0140] The aforementioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals can be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0141] Accordingly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to implement the steps in the above-described method embodiments. The computer-readable storage medium includes volatile or non-volatile components, or a combination thereof, and can be removable or non-removable. Examples of computer-readable storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technologies, CD-ROM, Digital Video Disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium. Accordingly, this application also provides a computer program product, which includes a computer program or instructions. When the computer program or instructions are executed by a processor, the processor is able to implement the steps in the above method embodiments. It should be understood that each step or combination of steps in the above method flow can be implemented by the computer program or instructions. In addition, these computer programs or instructions can be applied to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device, so that the processor of the general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device can be implemented as a means to implement the corresponding functions in the above method embodiments.

[0142] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0143] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. An information push method, characterized in that, include: Obtain interest attribute information corresponding to the object to be pushed, wherein the interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information; Based on the interest attribute information, a previously pushed object is determined that corresponds to the object to be pushed, and the similarity between the previously pushed object and the object to be pushed is greater than or equal to a preset threshold. Based on the interested groups corresponding to the already pushed objects, determine the target push group corresponding to the objects to be pushed; Information is pushed to the target audience based on the object to be pushed to.

2. The method according to claim 1, characterized in that, Obtain the interest attribute information corresponding to the object to be pushed, including: Obtain the object identifier information of the object to be pushed; Based on the object identification information, determine the object attribute information corresponding to the object to be pushed; The object attribute information is understood using a multimodal model to determine the interest attribute information.

3. The method according to claim 2, characterized in that, Using a multimodal model to understand the object's attribute information and determine the interest attribute information includes: The object attribute information is generalized to obtain multiple associated attribute information, and the degree of association between the associated attribute information and the object attribute information is greater than or equal to a preset threshold. The interest attribute information is determined by understanding the object attribute information and the multiple associated attribute information using the multimodal model.

4. The method according to claim 3, characterized in that, The object attribute information is generalized to obtain multiple associated attribute information, including: A hotspot knowledge graph corresponding to the object attribute information is determined. The hotspot knowledge graph includes multiple nodes and edges between two nodes. The nodes are used to represent the current hot information, and the edges are used to represent the correlation between hot information. Based on the hotspot knowledge graph, the object attribute information is generalized to obtain associated hotspot information corresponding to the object attribute information; Based on the associated hotspot information, the multiple associated attribute information is determined.

5. The method according to claim 4, characterized in that, Based on the aforementioned hotspot information, the multiple associated attribute information is determined, including: Determine the search keywords corresponding to the object attribute information; Based on the search keywords, a search is conducted on a preset platform to obtain search interest information associated with the object attribute information; Based on the associated hotspot information and the search interest information, the multiple associated attribute information is determined.

6. The method according to claim 1, characterized in that, Based on the interested audience corresponding to the already pushed objects, determine the target audience corresponding to the objects to be pushed, including: When there are multiple interest attribute information items, the priority of each of the multiple interest attribute information items is determined; Based on multiple interest attribute information and the priority of each interest attribute information, a target push audience corresponding to the object to be pushed is determined from the interested audience.

7. The method according to claim 6, characterized in that, Based on multiple interest attribute information and the priority of each interest attribute information, a target push audience corresponding to the object to be pushed is determined from the interested audience, including: Among the interested groups, determine the recommended groups corresponding to each of the multiple interest attribute information; Based on the recommended audience corresponding to each of the multiple interest attribute information and the priority corresponding to each of the multiple interest attribute information, the target push audience corresponding to the object to be pushed is determined.

8. The method according to any one of claims 1-7, characterized in that, Based on the interested audience corresponding to the already pushed objects, determine the target audience corresponding to the objects to be pushed, including: Determine the audience filtering strategy corresponding to the target audience; Based on the audience filtering strategy and the audience of interest corresponding to the already pushed objects, the target audience corresponding to the objects to be pushed is determined.

9. The method according to claim 8, characterized in that, Based on the audience filtering strategy and the interested audiences corresponding to the already pushed objects, the target audience corresponding to the objects to be pushed is determined, including: When the audience screening strategy includes an interest recommendation screening strategy, the recommendation degree of the pushed object to different users in the interested audience is determined, and the recommendation degree is used to characterize the degree of interest of users in the interested audience to the pushed object; Based on the recommendation level, the target push audience corresponding to the target object is determined by filtering the interested group. The recommendation level of users in the target push audience is greater than or equal to a preset threshold.

10. The method according to claim 9, characterized in that, Based on the recommendation level, the target audience corresponding to the object to be pushed to is determined by filtering among the interested population, including: If the audience filtering strategy also includes a tag filtering strategy, the first push audience is determined by filtering the interested audience based on the recommendation level. Determine user behavior tags associated with the pushed objects; Based on the user behavior tags, a second push group is determined by filtering the interested population, and the push users in the second push group correspond to the user behavior tags; The target audience is determined based on the first and second push audiences.

11. An information push device, characterized in that, include: The first acquisition module is used to acquire interest attribute information corresponding to the object to be pushed, and the interest attribute information includes at least one of the following: associated image information, object brand information, object function information, and object style information; The first determining module is used to determine, based on the interest attribute information, the already pushed object corresponding to the object to be pushed, wherein the similarity between the already pushed object and the object to be pushed is greater than or equal to a preset threshold. The first determining module is further configured to determine the target push audience corresponding to the object to be pushed based on the interest audience corresponding to the already pushed object; The first processing module is used to push information to the target audience based on the object to be pushed to.

12. An electronic device, characterized in that, include: A memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the method of any one of claims 1-10.

13. A computer storage medium, characterized in that, Used to store a computer program that, when executed by a computer, implements the method of any one of claims 1-10.

14. A computer program product, characterized in that, include: A computer program, when executed by a processor of an electronic device, causes the processor to perform the steps of the method of any one of claims 1-10.

Images