Information pushing method and device and information pushing model processing method and device

Abstract

The information pushing method comprises the following steps: acquiring information browsing records of a pushing receiving object in the same time period in each historical cycle; the information browsing record comprises a historical information category to which historical information browsed by the push receiving object in a time period belongs and a browsing time record for browsing the historical information; on the basis of historical category features extracted from historical information categories in the information browsing records, predicting the pushed information category of the pushed receiving object in the time period in the planned pushing period; on the basis of historical duration characteristics extracted from browsing time records in the information browsing records, predicting predicted browsing duration of the push receiving object in the time period in the planned push period; obtaining a plurality of pieces of recommendation information of the push information category by referring to the predicted browsing duration so as to generate a recommendation information stream of which the total browsing duration is matched with the predicted browsing duration; the recommendation information flow is used for being pushed to the push receiving object in the time period of the planned push cycle. According to the method, personalized recommendation can be realized based on user browsing habits.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an information push method, apparatus, computer equipment, computer-readable storage medium and computer program product, as well as an information push model processing method, apparatus, computer equipment, computer-readable storage medium and computer program product. Background Art

[0002] With the development of computer technology, information push technology has emerged, which can push information of interest to different users. Traditional information push generally uses the user's historical interaction information through a model to predict the information that the user is interested in, and then pushes the predicted information of interest to the user.

[0003] However, users often browse information in their free time, such as on their way to and from get off work or before going to bed. Traditional push methods push a large amount of information of many types, making it difficult for users to obtain information of interest from a large amount of information within a limited time. Summary of the invention

[0004] Based on this, it is necessary to provide a personalized information push method, device, computer equipment, computer-readable storage medium and computer program product that conforms to the user's browsing habits in order to address the above technical problems.

[0005] The present application provides an information push method, the method comprising:

[0006] Acquire the information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information;

[0007] Extracting historical category features based on the historical information categories in the information browsing records, and predicting the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle;

[0008] Extracting historical duration features based on the browsing time record in the information browsing record, and predicting the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle;

[0009] Referring to the predicted browsing time, obtaining a plurality of recommended information of the push information category;

[0010] Generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

[0011] The present application also provides an information push device, the device comprising:

[0012] A record acquisition module, used to acquire information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information;

[0013] A category prediction module, used to extract historical category features based on the historical information categories in the information browsing records, and predict the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle;

[0014] A duration prediction module is used to extract historical duration features based on the browsing time records in the information browsing records, and predict the predicted browsing duration corresponding to the push receiving object based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle;

[0015] An information acquisition module, used to obtain a plurality of recommended information of the push information category with reference to the predicted browsing time;

[0016] A generation module is used to generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

[0017] The present application also provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0018] Acquire the information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information;

[0019] Extracting historical category features based on the historical information categories in the information browsing records, and predicting the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle;

[0020] Extracting historical duration features based on the browsing time record in the information browsing record, and predicting the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle;

[0021] Referring to the predicted browsing time, obtaining a plurality of recommended information of the push information category;

[0022] Generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

[0023] The present application also provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the following steps are implemented:

[0024] Acquire the information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information;

[0025] Extracting historical category features based on the historical information categories in the information browsing records, and predicting the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle;

[0026] Extracting historical duration features based on the browsing time record in the information browsing record, and predicting the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle;

[0027] Referring to the predicted browsing time, obtaining a plurality of recommended information of the push information category;

[0028] Generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

[0029] The present application also provides a computer program product, the computer program product comprising a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0030] Acquire the information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information;

[0031] Extracting historical category features based on the historical information categories in the information browsing records, and predicting the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle;

[0032] Extracting historical duration features based on the browsing time record in the information browsing record, and predicting the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle;

[0033] Referring to the predicted browsing time, obtaining a plurality of recommended information of the push information category;

[0034] Generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

[0035] The above-mentioned information push method, device, computer equipment, storage medium and computer program product obtain the information browsing records of the push receiving object in the same time period in multiple historical cycles, so as to obtain the historical information category to which the historical information browsed by the push receiving object in the time period belongs, and the browsing time record of the browsing historical information from the information browsing record. The historical category feature is extracted based on the historical information category in the information browsing record, so as to predict the push information category corresponding to the push receiving object based on the historical category feature, and the information category that the user is interested in in the same time period of the planned push cycle can be accurately predicted based on the information category habit of the user's historical browsing in a specific time period. The historical duration feature is extracted based on the browsing time record in the information browsing record, so as to predict the predicted browsing time corresponding to the push receiving object based on the historical duration feature, and the time spent by the user on information browsing in the same time period of the planned push cycle can be accurately predicted based on the user's historical browsing time habit. With reference to the predicted browsing time, multiple recommended information of the push information category is obtained, so that multiple recommended information can be selected from the information category of interest to the push receiving object in combination with the time that the user may spend on information browsing, so as to generate a recommended information stream including multiple recommended information, which is convenient for the user to browse. The total browsing time of the generated recommended information stream matches the predicted browsing time, so that the total browsing time of the generated recommended information stream can meet the browsing time habits of users in a specific time period, realizing personalized information push. The generated recommended information stream is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information stream, and enabling users to complete browsing of the recommended information stream within their usual browsing time. Moreover, the information stream includes multiple recommended information, so that users can obtain more information of interest in one information stream, which not only helps users quickly understand the relevant content of the information stream, but also helps users save browsing time and improves the comfort of browsing the information stream.

[0036] The present application provides an information push model processing method, the method comprising:

[0037] Obtaining respective sample information browsing records of sample receiving objects in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0038] Extracting sample category features based on the sample historical information category in the sample information browsing record, and predicting the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle;

[0039] Extracting sample history duration features based on the browsing time record in the sample information browsing record, and predicting the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0040] Referring to the predicted browsing time of the sample, obtaining a plurality of sample recommendation information of the sample push information category;

[0041] Generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0042] The expected recommended information flow corresponding to the sample receiving object in the sample plan push period is obtained, and model training is performed based on the sample recommended information flow and the expected recommended information flow to obtain an information push model.

[0043] The present application also provides an information push model processing device, the device comprising:

[0044] A sample record acquisition module, used to acquire the sample information browsing records of each sample receiving object in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving object in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0045] A sample category prediction module is used to extract sample category features based on the sample historical information category in the sample information browsing record, and predict the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle;

[0046] A sample duration prediction module is used to extract sample history duration features based on the browsing time record in the sample information browsing record, and predict the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0047] A sample information acquisition module, used to obtain a plurality of sample recommendation information of the sample push information category with reference to the sample predicted browsing time;

[0048] A sample generation module, configured to generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0049] The training module is used to obtain the expected recommended information flow corresponding to the sample receiving object in the sample plan push period, perform model training based on the sample recommended information flow and the expected recommended information flow, and obtain an information push model.

[0050] The present application also provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0051] Obtaining respective sample information browsing records of sample receiving objects in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0052] Extracting sample category features based on the sample historical information category in the sample information browsing record, and predicting the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle;

[0053] Extracting sample history duration features based on the browsing time record in the sample information browsing record, and predicting the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0054] Referring to the predicted browsing time of the sample, obtaining a plurality of sample recommendation information of the sample push information category;

[0055] Generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0056] The expected recommended information flow corresponding to the sample receiving object in the sample plan push period is obtained, and model training is performed based on the sample recommended information flow and the expected recommended information flow to obtain an information push model.

[0057] The present application also provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the following steps are implemented:

[0058] Obtaining respective sample information browsing records of sample receiving objects in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0059] Extracting sample category features based on the sample historical information category in the sample information browsing record, and predicting the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle;

[0060] Extracting sample history duration features based on the browsing time record in the sample information browsing record, and predicting the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0061] Referring to the predicted browsing time of the sample, obtaining a plurality of sample recommendation information of the sample push information category;

[0062] Generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0063] The expected recommended information flow corresponding to the sample receiving object in the sample plan push period is obtained, and model training is performed based on the sample recommended information flow and the expected recommended information flow to obtain an information push model.

[0064] The present application also provides a computer program product, the computer program product comprising a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0065] Obtaining respective sample information browsing records of sample receiving objects in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0066] Extracting sample category features based on the sample historical information category in the sample information browsing record, and predicting the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle;

[0067] Extracting sample history duration features based on the browsing time record in the sample information browsing record, and predicting the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0068] Referring to the predicted browsing time of the sample, obtaining a plurality of sample recommendation information of the sample push information category;

[0069] Generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0070] The expected recommended information flow corresponding to the sample receiving object in the sample plan push period is obtained, and model training is performed based on the sample recommended information flow and the expected recommended information flow to obtain an information push model.

[0071] The above-mentioned information push model processing method, device, computer equipment, storage medium and computer program product obtain the sample information browsing records of the sample receiving objects in the same sample time period in multiple sample history cycles, so as to obtain the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, and the sample browsing time record of browsing the sample history information from the sample information browsing records. Based on the sample history information category in the sample information browsing record, the sample category feature is extracted to predict the sample push information category corresponding to the sample receiving object based on the sample category feature, and the information category that the user is interested in in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing habits of information categories in a specific sample time period. Based on the browsing time record in the sample information browsing record, the sample history duration feature is extracted to predict the sample predicted browsing time corresponding to the sample receiving object based on the sample history duration feature, and the time spent by the user on information browsing in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing time habit. With reference to the sample predicted browsing time, multiple sample recommendation information of the sample push information category is obtained, so that it is possible to select multiple sample recommendation information from the predicted information category in combination with the predicted time that the user may spend on information browsing, so as to generate a predicted sample recommendation information flow, and the total browsing time of the sample recommendation information flow matches the sample predicted browsing time. The expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle represents the information flow that the expected model can output during training, and is the goal of model learning. Model training is performed based on the sample recommendation information flow generated by the model and the recommended information flow generated by the expected model, so that the recommended information flow output by the model during training gradually approaches the expected recommended information flow, thereby obtaining an information push model. The obtained information push model can automatically predict the information category and browsing time of the user in a specific time period. Based on the predicted browsing time, multiple recommended information belonging to the predicted information category is selected to generate a recommended information flow including multiple recommended information, which is convenient for the user to browse. The total browsing time of the generated recommended information flow matches the predicted browsing time, so that the total browsing time of the generated recommended information flow can meet the browsing time habits of the user in a specific time period. The generated recommended information flow is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information flow. Automatic prediction and information flow generation through the information push model can effectively improve the speed of information push. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 An application environment diagram of an information push method in an embodiment;

[0073] Figure 2 A schematic diagram of a flow chart of an information push method in an embodiment;

[0074] Figure 3 It is a flowchart of an information push method in one of the embodiments;

[0075] Figure 4 It is a flowchart of an information push method in another embodiment;

[0076] Figure 5 is a schematic diagram of format conversion in one embodiment;

[0077] Figure 6 A schematic diagram of an interface for displaying an authorization notification for information browsing records in one embodiment;

[0078] Figure 7 A schematic diagram of an interface for triggering an intelligent recommendation mode in an embodiment;

[0079] Figure 8 A schematic diagram of an interface for displaying a format indication control in an information recommendation area in an embodiment;

[0080] Fig. 9 A schematic diagram of an interface for playing audio generated by multiple pieces of recommended information in an information recommendation area in one embodiment;

[0081] Fig.10 A schematic diagram of an interface for adjusting browsing time by a user in one embodiment;

[0082] Fig.11 is a timing diagram of an information push method in an embodiment;

[0083] Fig.12 A schematic diagram of a process flow of an information push model processing method in one embodiment;

[0084] Fig.13 It is a structural block diagram of an information push device in an embodiment;

[0085] Fig.14 It is a structural block diagram of an information push model processing device in an embodiment;

[0086] Fig.15 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0088] The embodiments of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, assisted driving, etc. For example, it can be applied to the field of artificial intelligence (AI) technology, where artificial intelligence is the theory, method, technology and application system of simulating, extending and expanding human intelligence using digital computers or machines controlled by digital computers, perceiving the environment, acquiring knowledge and using knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that the machines have the functions of perception, reasoning and decision-making. The scheme provided in the embodiments of the present application relates to an information push method and an information push model processing method of artificial intelligence, which are specifically described by the following embodiments.

[0089] The information push method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other servers. Both the terminal 102 and the server 104 can independently execute the information push method provided in the embodiment of the present application. The terminal 102 and the server 104 can also be used in conjunction to execute the information push method provided in the embodiment of the present application. When the terminal 102 and the server 104 are used in conjunction to execute the information push method provided in the embodiment of the present application, the terminal 102 obtains the respective information browsing records of the push receiving object in the same time period in multiple historical cycles from the server 104; the information browsing record includes the historical information category to which the historical information browsed by the push receiving object in the time period belongs, and the browsing time record of the browsing historical information. The terminal 102 extracts the historical category features based on the historical information category in the information browsing record, and predicts the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow that needs to be pushed in the time period in the planned push cycle. The terminal 102 extracts the historical duration feature based on the browsing time record in the information browsing record, and predicts the predicted browsing time corresponding to the push receiving object based on the historical duration feature; the predicted browsing time represents the browsing time of the information flow to be pushed in the time period of the planned push cycle. The terminal 102 refers to the predicted browsing time to obtain multiple recommended information of the push information category. The terminal 102 generates a recommended information flow including multiple recommended information, and the total browsing time of the recommended information flow matches the predicted browsing time; the recommended information flow is used to push to the push receiving object in the time period of the planned push cycle.

[0090] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0091] The information push model processing method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other servers. Both the terminal 102 and the server 104 can independently execute the information push model processing method provided in the embodiment of the present application. The terminal 102 and the server 104 can also be used in conjunction to execute the information push model processing method provided in the embodiment of the present application. When the terminal 102 and the server 104 are used in conjunction to execute the information push model processing method provided in the embodiment of the present application, the terminal 102 obtains the sample information browsing records of the sample receiving object in the same sample time period in multiple sample history cycles from the server 104; the sample information browsing record includes the sample history information category to which the sample history information browsed by the sample receiving object in the sample time period belongs, and the sample browsing time record of browsing the sample history information. The terminal 102 extracts the sample category feature based on the sample history information category in the sample information browsing record, and predicts the sample push information category corresponding to the sample receiving object based on the sample category feature; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle. The terminal 102 extracts the sample history duration feature based on the browsing time record in the sample information browsing record, and predicts the sample predicted browsing time corresponding to the sample receiving object based on the sample history duration feature; the sample predicted browsing time represents the browsing time of the sample information flow to be pushed in the sample time period in the sample plan push cycle. The terminal 102 refers to the sample predicted browsing time to obtain multiple sample recommendation information of the sample push information category. The terminal 102 generates a sample recommendation information flow including multiple sample recommendation information, and the total browsing time of the sample recommendation information flow matches the sample predicted browsing time. The terminal 102 obtains the expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle, performs model training based on the sample recommendation information flow and the expected recommended information flow, and obtains the information push model.

[0092] In one embodiment, Figure 2 As shown, a method for pushing information is provided, which is applied to Figure 1 The computer device in Figure 1 The terminal or server in the example is used to illustrate, including the following steps:

[0093] Step S202, obtaining the information browsing records of the push recipients in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipients in the time period belongs, and the browsing time record of the browsing historical information.

[0094] The push receiving object refers to the object that receives the recommended information flow, for example, the user who receives the recommended information flow.

[0095] A historical period refers to a time period in the past, which can be measured in days, weeks, months, etc. For example, the past day is a historical period. A historical period can be divided into multiple time periods, such as 8:00-9:00, 9:00-12:00, etc. in the past day.

[0096] Different time periods can represent that the push recipient is in different scenarios, for example, 8:00-9:00 on the way to work, 12:00-14:00 in the lunch break scenario, etc.

[0097] Information browsing records refer to the browsing records of information by the push recipient within a certain period of time in the historical cycle. Information browsing records include the historical information category to which the historical information browsed by the push recipient within a certain period of time belongs, and the browsing time record of the historical information. The information browsing records may also include the historical information browsed by the push recipient within the time period, and the total browsing time of each historical information.

[0098] Browsing time records refer to records related to browsing time of historical information browsed by the push recipient in a certain period of time. Browsing time records include the browsing time of each historical information, and may also include the start time and end time of browsing historical information.

[0099] Specifically, the computer device determines the same time period in multiple historical cycles, and obtains the information browsing records of the push recipients in the same time period. For each information browsing record, the historical information category to which the historical information browsed by the push recipient belongs and the browsing time record of the browsing historical information are obtained from the targeted information browsing record.

[0100] Step S204, extracting historical category features based on the historical information categories in the information browsing records, and predicting the push information category corresponding to the push recipient based on the historical category features; the push information category represents the category of information in the information flow to be pushed in this time period in the planned push cycle.

[0101] The historical category feature refers to a feature that can characterize the information category of historical information. The push information category characterizes the category of information in the information flow that needs to be pushed to the push receiving object in the same time period in the planned push cycle.

[0102] Specifically, the computer device can extract historical category features based on the historical information categories in each information browsing record, and obtain the historical category features corresponding to the same time period of the push receiving object in multiple historical cycles. Further, the computer device can determine multiple historical information belonging to the historical information category, and extract historical category features based on the multiple historical information.

[0103] The computer device predicts the push information category corresponding to the same time period in the planned push cycle based on the historical category features, that is, predicts the push information category that the push recipient is interested in in the same time period in the planned push cycle.

[0104] Step S206, extracting historical duration features based on the browsing time records in the information browsing records, and predicting the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in this time period in the planned push cycle.

[0105] The historical duration feature refers to a feature that can characterize the historical browsing duration, and the historical duration feature characterizes the historical browsing duration of the push recipient in the same time period in multiple historical cycles.

[0106] The predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period of the planned push cycle, that is, the possible browsing duration of the push recipient in the time period.

[0107] Specifically, the computer device may extract features based on the browsing time records in each information browsing record to obtain the corresponding historical duration features. Based on the historical duration features, the computer device predicts the predicted browsing duration corresponding to the time period in the planned push cycle of the push recipient. That is, the predicted browsing duration of the push recipient in the same time period in the planned push cycle is predicted.

[0108] Step S208 , referring to the predicted browsing time, obtaining a plurality of recommended information of the push information category.

[0109] Specifically, there are multiple candidate information corresponding to the push information category, and these candidate information all belong to the push information category. It takes a certain information browsing time to browse each candidate information. The computer device can refer to the predicted browsing time and obtain multiple recommended information from the multiple candidate information of the push information category, so that the total browsing time of the obtained multiple recommended information matches the predicted browsing time.

[0110] In this embodiment, a plurality of recommended information may be screened out from a plurality of candidate information of the push information category according to the information browsing time and the predicted browsing time of each candidate information.

[0111] Step S210, generating a recommendation information stream including a plurality of recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to a push receiving object in a time period within a planned push cycle.

[0112] Specifically, the computer device performs media synthesis on the plurality of recommendation information to obtain a recommendation information stream, wherein the total browsing time of the recommendation information stream matches the predicted browsing time.

[0113] When the time period within the planned push cycle is reached, the computer device pushes the recommended information flow to the push recipient.

[0114] In this embodiment, when a plurality of recommendation information meets the media synthesis condition, the computer device performs media synthesis on the plurality of recommendation information to obtain a recommendation information stream.

[0115] The media synthesis condition represents a constraint condition for media synthesis of multiple recommendation information. The media synthesis condition means that multiple recommendation information can be synthesized according to the native media format. The native media format represents the presentation form of the recommendation information. The native media format includes at least one of a text format, an image format, a graphic format, an audio format, or a video format.

[0116] In this embodiment, the computer device determines the native media formats of the plurality of recommendation information, and judges whether the plurality of recommendation information meets the media synthesis condition according to the native media formats of the plurality of recommendation information. If they meet the condition, media synthesis is performed.

[0117] The native media formats of any two recommended information may be the same or different. For example, the native media format of recommended information A is text format, and the native media format of recommended information B is image and text format.

[0118] In this embodiment, the information browsing records of the push receiving object in the same time period in multiple historical cycles are obtained, so as to obtain the historical information category to which the historical information browsed by the push receiving object in the time period belongs, and the browsing time record of the browsing historical information from the information browsing record. The historical category feature is extracted based on the historical information category in the information browsing record, so as to predict the push information category corresponding to the push receiving object based on the historical category feature, and the information category that the user is interested in in the same time period of the planned push cycle can be accurately predicted based on the information category habit of the user's historical browsing in a specific time period. The historical duration feature is extracted based on the browsing time record in the information browsing record, so as to predict the predicted browsing duration corresponding to the push receiving object based on the historical duration feature, and the time spent by the user on information browsing in the same time period of the planned push cycle can be accurately predicted based on the user's historical browsing duration habit. With reference to the predicted browsing duration, multiple recommended information of the push information category is obtained, so that multiple recommended information can be selected from the information category of interest to the push receiving object in combination with the time that the user may spend on information browsing, so as to generate a recommended information flow including multiple recommended information, which is convenient for the user to browse. The total browsing time of the generated recommended information stream matches the predicted browsing time, so that the total browsing time of the generated recommended information stream can meet the browsing time habits of users in a specific time period, realizing personalized information push. The generated recommended information stream is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information stream, and enabling users to complete browsing of the recommended information stream within their usual browsing time. Moreover, the information stream includes multiple recommended information, so that users can obtain more information of interest in one information stream, which not only helps users quickly understand the relevant content of the information stream, but also helps users save browsing time and improves the comfort of browsing the information stream.

[0119] In one embodiment, Figure 3 As shown, the information browsing record also includes the historical position of the historical information browsed by the push recipient within the time period. In other embodiments, the information browsing record also includes the historical triggering time of the push recipient browsing within the time period.

[0120] The computer device can predict the push trigger position corresponding to the push recipient based on the historical position in each information browsing record; the push trigger position represents the position that triggers the push of the information flow to be pushed in the time period of the planned push cycle. That is, when the push recipient is at the push position in the time period of the planned push cycle, the recommended information flow is pushed to the push recipient.

[0121] After the recommended information flow is generated, when the push receiving object is in the push position during the time period within the planned push cycle, the recommended information flow is pushed to the push receiving object.

[0122] Alternatively, the computer device may predict the predicted trigger time corresponding to the push recipient object based on the historical trigger time in each information browsing record; the predicted trigger time represents the time when the push of the information flow to be pushed is triggered in the time period in the planned push cycle.

[0123] When the predicted trigger time in the time period within the planned push cycle is reached, the recommended information flow is pushed to the push receiving object.

[0124] The recommended information stream pushed includes at least one of a synthetic text information stream, a synthetic image, a synthetic graphic and text information stream, a synthetic audio or a synthetic video.

[0125] In this embodiment, the flow chart of the information push method is as follows: Figure 4 As shown, the information push model is used to obtain the information browsing records of each user in the same time period in multiple historical cycles. The information browsing records may include historical locations, historical browsing information, and the historical information category to which the historical information belongs. The information push model is used to predict the user's preferences in the time period based on each information browsing record, that is, to predict the push information category and predicted browsing time corresponding to the user. The information push model is used to generate a recommended information flow based on the push information category and the predicted browsing time. The recommended information flow may include synthesized audio and synthesized video. The recommended information flow is pushed to the user in the time period within the planned push cycle.

[0126] In one embodiment, based on the historical category features, predicting the push information category corresponding to the push recipient includes:

[0127] Acquire candidate category features extracted from multiple candidate information categories respectively; for each candidate information category, predict the predicted category score corresponding to the targeted candidate information category based on the candidate category features and historical category features of the targeted candidate information category; the predicted category score represents the probability that the information flow to be pushed includes information of the targeted candidate information category; from the multiple candidate information categories, filter out the push information category, and the predicted category score of the push information category meets the preset category push condition.

[0128] The predicted category score represents the probability that the information flow to be pushed to the push recipient includes information of the targeted candidate information category, that is, the probability that the push recipient browses the information under the targeted candidate information category.

[0129] The preset category push condition is a constraint condition on the predicted category score, that is, a constraint condition on the predicted category score corresponding to the push information category selected according to the preset category push condition.

[0130] Specifically, the computer device obtains multiple candidate information categories, and determines multiple candidate information corresponding to each candidate information category. For each candidate information category, feature extraction is performed based on the multiple candidate information corresponding to the candidate information category to obtain the candidate category feature of the candidate information category.

[0131] For each candidate information category, the category similarity between the candidate category feature and the historical category feature of the candidate information category is determined, and the category similarity is used as the predicted category score corresponding to the candidate information category. The category similarity may be cosine similarity.

[0132] The computer device obtains a preset category push condition, selects a prediction category score that meets the preset category push condition from each prediction category score, and determines the candidate information category corresponding to the selected prediction category score as the push information category.

[0133] In this embodiment, the preset category push condition may be to filter out the largest predicted category score, or to filter out the predicted category score that is greater than the category score threshold, or to filter out a preset number of predicted category scores in order. For example, the largest predicted category score is filtered out from each predicted category score, and the candidate information category corresponding to the largest predicted category score is determined as the pushed information category.

[0134] In this embodiment, candidate category features extracted from multiple candidate information categories are obtained, and for each candidate information category, a predicted category score corresponding to the candidate information category is predicted based on the candidate category features and historical category features of the candidate information category. The probability of a user browsing each candidate information category can be predicted, and the push information category that the user is most likely to be interested in during the time period in the planned push cycle can be screened out from the multiple candidate information categories based on these probabilities, so that the information of interest can be accurately pushed to the user.

[0135] In one embodiment, referring to the predicted browsing time, a plurality of recommended information of the push information category is obtained, including:

[0136] Acquire a candidate information set belonging to a push information category, and determine the information browsing time corresponding to each candidate information in the candidate information set; perform at least two combinations of the candidate information in the candidate information set to form at least two information groups, each information group including at least two candidate information; determine the total candidate browsing time corresponding to each information group based on the information browsing time of the candidate information in each information group; determine the information group corresponding to the candidate total browsing time that matches the predicted browsing time, and determine the candidate information included in the determined information group as multiple recommended information of the push information category.

[0137] The candidate information set refers to a set of multiple candidate information belonging to the push information category.

[0138] Specifically, the computer device obtains a candidate information set belonging to the push information category, wherein the candidate information set includes multiple candidate information. The computer device determines the information browsing time corresponding to each candidate information, and the information browsing time may be a browsing time pre-configured for the corresponding candidate information, or may be determined based on the historical browsing time of the candidate information. The determination based on the historical browsing time of the candidate information may be based on the average, median, or weighted value of the historical browsing time of the candidate information by different historical receiving objects.

[0139] The computer device performs at least two combinations of candidate information in the candidate information set to form at least two information groups, each of which includes at least two candidate information. For each information group, the computer device determines the total candidate browsing time corresponding to the targeted information group based on the sum of the information browsing time of each candidate information in the targeted information group, thereby obtaining the total candidate browsing time corresponding to each information group.

[0140] In this embodiment, for each information group, the computer device performs weighted summation of the information browsing time of each candidate information in the targeted information group to obtain the total candidate browsing time corresponding to the targeted information group.

[0141] The computer device determines the candidate browsing total time that matches the predicted browsing time among the candidate browsing total time corresponding to each information group, and determines the information group corresponding to the matching candidate browsing total time. The computer device determines each candidate information included in the determined information group as a plurality of recommended information of the push information category.

[0142] In this embodiment, a candidate information set belonging to the push information category is obtained, the information browsing time corresponding to each candidate information in the candidate information set is determined, and at least two combinations are performed on the candidate information in the candidate information set to form at least two information groups, each of which includes at least two candidate information, so that the candidate total browsing time corresponding to each information group can be accurately determined according to the information browsing time of the candidate information in each information group. An information group corresponding to the candidate total browsing time matching the predicted browsing time is determined, so that the candidate information included in the determined information group is determined as multiple recommended information of the push information category, so that the recommended information can be filtered based on the information browsing time of each candidate information, so that the total browsing time of the filtered multiple recommended information can match the predicted browsing time, so that the total browsing time of the generated recommended information flow meets the user's information browsing time habits, so that the recommended information flow of corresponding time can be generated according to the user's browsing time habits, so as to improve the user's browsing interest in the pushed information flow and help improve the conversion rate of the information flow.

[0143] In one embodiment, generating a recommendation information flow including a plurality of recommendation information includes:

[0144] Determine the native media format of each of the multiple recommendation information; determine the target media format of the recommendation information stream to be generated; perform media synthesis on the multiple recommendation information according to the native media format of each of the multiple recommendation information and the target media format to obtain the recommendation information stream.

[0145] The target media format refers to a pre-set media format, which represents the presentation form of the recommended information flow to be generated. The target media format includes at least one of a text format, an image format, a graphic format, an audio format or a video format.

[0146] Specifically, the computer device determines the native media format of each of the plurality of recommendation information, and obtains and determines the target media format of the recommendation information stream to be generated. The computer device performs media synthesis on the plurality of recommendation information according to the native media format of each of the plurality of recommendation information and the target media format to obtain the recommendation information stream.

[0147] In this embodiment, the native media format of each of the multiple recommendation information is determined, the target media format of the recommendation information stream to be generated is determined, and the multiple recommendation information are subjected to media synthesis based on the native media format of each of the multiple recommendation information and the target media format, so as to accurately generate the recommendation information stream in the target media format.

[0148] In one embodiment, according to the native media formats and target media formats of the plurality of recommendation information, the plurality of recommendation information are subjected to media synthesis to obtain a recommendation information stream, including:

[0149] Determine first category recommendation information and second category recommendation information from multiple recommendation information, wherein the native media format of the first category recommendation information and the target media format have a media synthesis condition, and the native media format of the second category recommendation information and the target media format do not have a media synthesis condition; convert the second category recommendation information into the target media format according to a format conversion method from the native media format of the second category recommendation information to the target media format; perform media synthesis on the first category recommendation information and the second category recommendation information in the target media format to obtain a recommendation information flow.

[0150] Specifically, the computer device determines the first category of recommended information and the second category of recommended information from the plurality of recommended information according to whether the native media formats of the plurality of recommended information have the media synthesis condition with the target media format. The native media format of the first category of recommended information has the media synthesis condition with the target media format, and the native media format of the second category of recommended information does not have the media synthesis condition with the target media format.

[0151] For example, the text format and the image format have the media synthesis condition, the text format and the graphic format have the media synthesis condition, the image format and the graphic format have the media synthesis condition, and the audio format and the video format have the synthesis condition. The text format and the audio format do not have the media synthesis condition, and the text format and the video format do not have the media synthesis condition.

[0152] For each native media format of the second type of recommendation information, the computer device obtains a format conversion method from the native media format to the target media format, and converts the recommendation information corresponding to the native media format into the target media format according to the format conversion method. According to the same process, all the second type of recommendation information can be converted into the target media format to obtain the second type of recommendation information in the target media format.

[0153] The computer device performs media synthesis on the first type of recommendation information and the second type of recommendation information in the target media format to obtain a recommendation information stream.

[0154] In this embodiment, the first type of recommendation information and the second type of recommendation information in the target media format are subjected to media synthesis, which may be to splice the first type of recommendation information and the second type of recommendation information in the target media format to obtain a recommendation information stream.

[0155] In this embodiment, first-category recommendation information and second-category recommendation information are determined from multiple recommendation information to determine first-category recommendation information whose native media format and target media format have media synthesis conditions, and second-category recommendation information whose native media format and target media format do not have media synthesis conditions, so as to determine which recommendation information needs to be format converted. According to the format conversion method from the native media format to the target media format of the second-category recommendation information, the second-category recommendation information can be accurately converted into the target media format, so that the two types of recommendation information can be media synthesized, thereby multiple recommendation information in different forms can be synthesized into one recommendation information flow, which is convenient for users to browse.

[0156] In one of the embodiments, the native media format includes at least one of a text format, an image format, a graphic-text format, an audio format, or a video format; the target media format includes at least one of a text format, an image format, a graphic-text format, an audio format, or a video format; and the recommended information stream includes at least one of a synthesized text information stream, a synthesized image, a synthesized graphic-text information stream, a synthesized audio, or a synthesized video.

[0157] Specifically, the original media format of each recommendation information is one of a text format, an image format, a graphic format, an audio format or a video format. The original media formats of different recommendation information may be the same or different.

[0158] The target media format includes at least one of a text format, an image format, a graphic format, an audio format or a video format.

[0159] When the native media format is one of a text format, an image format, or a graphic-text format, and the target media format includes at least one of a text format, an image format, or a graphic-text format, the native media format and the target media format have a media synthesis condition.

[0160] When the native media format is one of an audio format or a video format, and the target media format includes at least one of the audio format or the video format, the native media format and the target media format have a media synthesis condition.

[0161] When the native media format is one of a text format, an image format, or a graphic format, and the target media format includes at least one of an audio format or a video format, the native media format and the target media format do not have a media synthesis condition.

[0162] When the native media format is one of an audio format or a video format, and the target media format includes at least one of a text format, an image format, or a graphic format, the native media format and the target media format do not have a media synthesis condition.

[0163] In this embodiment, when the target media format includes a text format, the generated recommended information stream includes a synthesized text information stream; when the target media format includes an image format, the generated recommended information stream includes a synthesized image; when the target media format includes a graphic format, the generated recommended information stream includes a synthesized graphic information stream; when the target media format includes an audio format, the generated recommended information stream includes a synthesized audio; when the target media format includes a video format, the generated recommended information stream includes a synthesized video.

[0164] like Figure 5 As shown, the multiple recommended information is picture 1 and picture 2, and the native media format of picture 1 and picture 2 is picture format. The target media format includes audio format and video format, and according to the format conversion method between picture format and audio format, picture 1 and picture 2 are converted into audio format respectively to obtain audio 1 and audio 2. Audio 1 and audio 2 are synthesized to obtain synthesized audio.

[0165] According to the format conversion method between the image format and the video format, the image 1 and the image 2 are converted into the video format respectively to obtain the video 1 and the video 2. The video 1 and the video 2 are synthesized to obtain the synthesized video.

[0166] In this embodiment, format conversion includes, for example, converting text and image format to video format, converting video format to text and image format, converting text and image format to audio format, converting audio format to text and image format, converting audio format to video format, etc., but is not limited thereto.

[0167] Among them, the format conversion between images and videos is achieved by extracting sequence frames or expanding sequence frames; the format conversion between text and audio is achieved through speech synthesis (Text To Speech, TTS).

[0168] In this embodiment, text format, image format, graphic format, audio format and video format can be converted to each other, so that the recommended information that needs format conversion can be converted from the native media format to the target media format, so that multiple recommended information in different forms can be synthesized into the same recommended information stream for user browsing. In addition, the synthesized recommended information stream can be in the form of text, image, graphic, audio and video, etc., which can provide users with a variety of information browsing methods, so that the browsing methods can adapt to different application scenarios.

[0169] In one embodiment, based on the historical duration feature, predicting the predicted browsing duration corresponding to the push recipient includes:

[0170] Acquire multiple candidate browsing times; for each candidate browsing time, predict a predicted time score corresponding to the candidate browsing time based on the candidate browsing time and historical time features; filter the predicted browsing time from the multiple candidate browsing times, and the predicted time score corresponding to the predicted browsing time meets the preset browsing time condition.

[0171] The candidate browsing duration refers to the preset browsing duration. The predicted duration score represents the probability of the browsing duration that the push recipient is interested in in the same time period during the planned push cycle.

[0172] The preset browsing time condition is a constraint condition on the predicted browsing time score, that is, a constraint condition on the predicted browsing time score corresponding to the predicted browsing time selected according to the preset browsing time condition.

[0173] Specifically, the computer device obtains multiple candidate browsing durations, and for each candidate browsing duration, determines the duration similarity between the candidate browsing duration and the historical duration feature, and uses the duration similarity as the predicted duration score corresponding to the candidate browsing duration. The duration similarity may be cosine similarity.

[0174] The computer device obtains a preset browsing time condition, selects a predicted time score that meets the preset browsing time condition from each predicted time score, and determines the candidate browsing time corresponding to the selected predicted time score as the predicted browsing time.

[0175] In this embodiment, the preset browsing duration condition may be to filter out the maximum predicted duration score, or to filter out predicted duration scores greater than a duration score threshold, or to filter out a specific number of predicted duration scores in order.

[0176] In this embodiment, multiple candidate browsing times are obtained, and for each candidate browsing time, a predicted time score corresponding to the candidate browsing time is predicted based on the candidate browsing time and historical time features. This can predict the user's interest level in each candidate browsing time during this time period in the planned push cycle, thereby screening out the browsing time that the user is most interested in during this time period in the planned push cycle, so that a recommended information flow of the corresponding browsing time can be generated for the user, which is in line with the user's browsing habits, helps to increase the user's interest in the pushed information, and thus improves the conversion rate of information.

[0177] In one embodiment, generating a recommendation information flow including a plurality of recommendation information includes:

[0178] Acquire a target media format indicated by a terminal corresponding to a push receiving object; the target media format is determined in response to a triggering operation on a format indication control when the terminal displays an information recommendation area including a format indication control; generate a recommended information stream of the target media format based on multiple recommended information, the recommended information stream including multiple recommended information; the recommended information stream is used for display in the information recommendation area.

[0179] Specifically, an information recommendation area is displayed on the terminal corresponding to the push receiving object, and a format indication control is displayed in the information recommendation area. The push receiving object can trigger the format indication control, and the terminal corresponding to the push receiving object determines the target media format indicated by the format indication control and feeds back to the computer device.

[0180] The computer device obtains the target media format indicated by the terminal corresponding to the push receiving object, and generates a recommendation information stream of the target media format based on multiple recommendation information. The recommendation information stream includes multiple recommendation information.

[0181] The computer device pushes the recommended information stream to the terminal corresponding to the push receiving object, and the terminal displays the recommended information stream in the information recommendation area.

[0182] In this embodiment, the target media format indicated by the terminal corresponding to the push recipient is obtained to determine the media format of the recommended information stream to be pushed to the push recipient, thereby generating a recommended information stream of the target media format based on multiple recommended information, so that multiple recommended information are integrated into the same information stream for easy browsing by users. The recommended information stream pushed to the terminal corresponding to the push recipient is displayed in the information recommendation area, so that the recommended information stream is displayed in a specific area, effectively distinguishing other information, thereby avoiding interference from other information.

[0183] like Figure 6 As shown, an information display interface is displayed on the terminal corresponding to the push recipient, and an authorization notification for the information browsing record is displayed on the information display interface. After the user triggers the authorization for the information browsing record, the terminal can record the relevant browsing information of the push recipient in multiple time periods to form an information browsing record. Multiple time periods include but are not limited to time periods such as commuting, eating, before lunch break, and before going to bed. The information browsing record is as follows:

[0184] Browsing time: 8:00-8:30;

[0185] Browsing location: on the subway, on the bus, in the car, etc.; judging the reading location by the change of the path;

[0186] Information categories of browsed information: News 70%, Entertainment 20%, Sports 10%;

[0187] Media formats: text 50%, image 20%, video 20%, audio 10%.

[0188] Users can read content at will in the information flow of the information display interface, and the information flow background stores the user's information browsing history. Users trigger preset operations on the information display interface, and smart browsing prompts are displayed, such as Figure 7 As shown. The preset operation can be to pull down the information flow page, and continue to pull down after the refresh appears. The user chooses to turn on the smart reading function, and a prompt message is displayed on the information display interface to prompt the user to intelligently recommend information based on the user's reading behavior.

[0189] like Figure 8 As shown, after it is turned on, it intelligently determines the information categories that users want to browse in different time periods. The information recommendation area pops up, and format indicator controls, such as video playback controls and audio playback controls, are displayed in the information recommendation area. When the user clicks the video playback control, a video generated by multiple recommended information is played in the information recommendation area. For example, a 10-minute video synthesized from 50 recommended information is played.

[0190] like Fig. 9As shown, when the user clicks the audio playback control, the audio generated by multiple recommended information is played in the information recommendation area. For example, a 10-minute audio synthesized from 50 recommended information is played.

[0191] The media formats of the recommended information stream include but are not limited to video, audio, image, text, graphics, etc. Video and audio can make it more convenient for users to browse information in scenes such as driving.

[0192] The intelligently recommended information stream is usually synthesized and loaded in advance, so that users can watch or listen to it even without a network environment. For example, a recommended information stream is generated based on multiple recommended information before the user enters the subway.

[0193] The duration of smart recommendations is usually dynamically adjusted based on the user's past data. For example, if the morning commute time becomes 15 minutes, the subsequently generated recommended information flow will be 15 minutes.

[0194] In one embodiment, the method further comprises:

[0195] The total browsing time is sent to the terminal corresponding to the push receiving object; the total browsing time after the total browsing time is adjusted is obtained; with reference to the total browsing time after the time adjustment, the recommendation information included in the recommendation information flow is adjusted to obtain the adjusted recommendation information flow; the adjusted recommendation information flow is used to push to the push receiving object in the time period within the planned push cycle.

[0196] Specifically, the information recommendation area is displayed on the terminal corresponding to the push receiving object, and the push receiving object can trigger a time adjustment operation in the information recommendation area to display the time adjustment area. In response to the time adjustment operation triggered by the push receiving object, the computer device sends the total browsing time to the terminal corresponding to the push receiving object. The terminal displays the total browsing time in the time adjustment area and displays an adjustment control for the total browsing time. The adjustment control includes a time increase control and a time decrease control, such as Fig.10 For example, in the duration adjustment area, the user increases the total browsing time of the recommended information stream from 10 minutes to 15 minutes, or decreases it from 10 minutes to 8 minutes.

[0197] The push recipient can increase or decrease the total browsing time of the recommended information flow in the time adjustment area. The terminal obtains the total browsing time after the total browsing time is adjusted, and feeds it back to the computer device. The computer device adjusts the recommended information included in the recommended information flow with reference to the total browsing time after the time adjustment, and obtains the adjusted recommended information flow, so that the total browsing time of the adjusted recommended information flow matches the adjusted total browsing time.

[0198] The computer device pushes the adjusted recommended information flow to the terminal corresponding to the push recipient during the time period within the planned push cycle.

[0199] In this embodiment, the total browsing time is sent to the terminal corresponding to the push receiving object, so that the push receiving object can adjust the total browsing time. The total browsing time after the total browsing time is adjusted is obtained, and the recommended information included in the recommended information flow is adjusted with reference to the total browsing time after the time adjustment, so as to obtain the adjusted recommended information flow, thereby providing the user with the function of adjusting the browsing time of the information flow, so that the user can adjust the browsing time of the information flow according to the needs, making the information flow recommendation more flexible.

[0200] In one embodiment, an application scenario of an information push method is provided, and the specific processing is as follows: Fig.11 As shown in the timing diagram.

[0201] At 8:00 on July 20, when the user opened the information display interface of the information flow on the client, the client sent a query to the server to ask whether to display the information recommendation area on the information display interface.

[0202] The server performs category prediction and browsing time prediction: The server stores the information browsing records of the user in the time period of 8:00-9:00 on July 17-July 19. Each information browsing record includes the historical information browsed in the time period of 8:00-9:00, the historical information category to which the historical information belongs, and the historical browsing time of browsing the historical information. Based on the historical browsing time, the predicted browsing time corresponding to the user in 8:00-9:00 on July 20 is predicted, for example, the predicted browsing time is 15 minutes. According to the historical information and the historical information category to which the historical information belongs, the information category that the user wants to browse in 8:00-9:00 on July 20 is predicted, that is, the push information category. With reference to the predicted browsing time, multiple recommended information of the push information category is obtained. When multiple recommended information are all videos, the multiple recommended information are synthesized to obtain a 15-minute synthesized video. When multiple recommended information are not completely videos, the recommended information that is not a video is converted into a video, and then the videos are synthesized to obtain a synthesized video.

[0203] The server sends the synthesized video to the client, and the client displays an information recommendation area on the information display interface, and displays the total viewing time of the synthesized video in the information recommendation area, which is 15 minutes.

[0204] The user clicks the video play control in the information recommendation area, and the client plays the synthesized video in the information recommendation area.

[0205] The client records the user's browsing data between 8:00 and 9:00 on July 20, forms an information browsing record, and uploads it to the server.

[0206] The server stores the user's information browsing records from 8:00 to 9:00 on July 20 for subsequent information recommendation.

[0207] In this embodiment, based on the user's browsing behavior habits, the information categories and browsing duration that the user is most likely to browse in different time intervals are determined, the core content that the user wants to browse is intelligently extracted, and different media forms such as video, audio, pictures, and text are synthesized, and intelligently recommended to the user for browsing. The user can quickly understand the relevant content of the information flow through different media forms, helping the user save time and improve the comfort of reading the information flow. In addition, the user's needs for fast information consumption in multiple scenarios are met. For example, on the way to and from get off work, before going to bed, and at mealtime, the user can consume content information based on intelligent duration and different media types, thereby increasing the user's consumption frequency and stickiness, and increasing the scale and activity of the product.

[0208] In one embodiment, Fig.12 As shown, a method for processing an information push model is provided, and the method is applied to Figure 1 The computer device in Figure 1 The method includes:

[0209] Step S1202, obtaining the sample information browsing records of each sample receiving object in the same sample time period in multiple sample history cycles; the sample information browsing record includes the sample history information category to which the sample history information browsed by the sample receiving object in the sample time period belongs, and the sample browsing time record of browsing the sample history information.

[0210] The sample receiving object refers to the object that receives the sample recommendation information flow. The sample historical period refers to the past time period, which can be measured in days, weeks, months, etc. The sample historical period can be divided into multiple sample time periods.

[0211] The sample information browsing record refers to the browsing record of the sample information by the sample receiving object within a certain time period in the sample history cycle. The sample information browsing record includes the sample history information category to which the sample history information browsed by the sample receiving object within a certain sample time period belongs, and the sample browsing time record of browsing the sample history information. The sample information browsing record may also include the sample history information browsed by the sample receiving object within the time period, and may also include the total sample browsing time of each sample history information.

[0212] The sample browsing time record refers to the record of the browsing time of the sample historical information browsed by the sample receiving object in a certain time period. The sample browsing time record includes the sample browsing time of each sample historical information, and may also include the sample start time and sample end time of the sample browsing historical information.

[0213] Specifically, the computer device determines the same sample time period in multiple sample history cycles, and obtains the sample information browsing records of the sample receiving objects in the multiple same sample time periods. For each sample information browsing record, the sample history information category to which the sample history information browsed by the sample receiving object belongs and the sample browsing time record of browsing the sample history information are obtained from the sample information browsing record.

[0214] Step S1204, extracting sample category features based on the sample historical information categories in the sample information browsing records, and predicting the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period during the sample plan push cycle.

[0215] In this embodiment, based on the sample category features, predicting the sample push information category corresponding to the sample receiving object includes:

[0216] Obtain sample candidate category features extracted from multiple sample candidate information categories respectively; for each sample candidate information category, predict the sample prediction category score corresponding to the targeted candidate information category based on the sample candidate category features and sample historical category features of the targeted sample candidate information category; the sample prediction category score represents the probability that the sample push information flow includes the targeted sample candidate information category information; from multiple sample candidate information categories, screen out a sample push information category, the sample prediction category score of which meets the sample category push condition.

[0217] Step S1206, extracting the sample historical duration feature based on the browsing time record in the sample information browsing record, and predicting the sample predicted browsing duration corresponding to the sample receiving object based on the sample historical duration feature; the sample predicted browsing duration represents the browsing duration of the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle.

[0218] In this embodiment, based on the sample historical duration feature, predicting the sample predicted browsing duration corresponding to the sample receiving object includes:

[0219] Obtain multiple sample candidate browsing times; for each sample candidate browsing time, predict a sample predicted time score corresponding to the targeted sample candidate browsing time based on the targeted sample candidate browsing time and sample historical time features; filter the sample predicted browsing time from the multiple sample candidate browsing times, the sample predicted browsing time corresponding to which has a sample predicted time score that meets the sample browsing time condition.

[0220] Step S1208 , referring to the predicted browsing time of the sample, obtaining a plurality of sample recommendation information of the sample push information category.

[0221] In this embodiment, referring to the predicted browsing time of the sample, multiple sample recommendation information of the sample push information category is obtained, including:

[0222] A sample candidate information set belonging to a sample push information category is obtained, and the sample information browsing time corresponding to each sample candidate information in the sample candidate information set is determined; at least two combinations are performed on the sample candidate information in the sample candidate information set to form at least two sample information groups, and each sample information group includes at least two sample candidate information; according to the sample information browsing time of the sample candidate information in each sample information group, the total sample candidate browsing time corresponding to each sample information group is determined; a sample information group corresponding to a total sample candidate browsing time that matches the sample predicted browsing time is determined, and the sample candidate information included in the determined sample information group is determined as a plurality of sample recommendation information of the sample push information category.

[0223] Step S1210: Generate a sample recommendation information stream including a plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time.

[0224] In this embodiment, generating a sample recommendation information flow including a plurality of sample recommendation information includes:

[0225] Determine the native media format of each of the multiple sample recommendation information; determine the target media format of the sample recommendation information stream to be generated; perform media synthesis on the multiple sample recommendation information according to the native media format of each of the multiple sample recommendation information and the target media format to obtain the sample recommendation information stream.

[0226] In this embodiment, according to the native media formats and target media formats of the plurality of sample recommendation information, the plurality of sample recommendation information are subjected to media synthesis to obtain a sample recommendation information stream, including:

[0227] Determine, from the plurality of recommendation information, first-category sample recommendation information and second-category sample recommendation information, wherein the native media format of the first-category sample recommendation information and the target media format have a media synthesis condition, and the native media format of the second-category sample recommendation information and the target media format do not have a media synthesis condition;

[0228] Convert the second type of sample recommendation information into the target media format according to the format conversion method from the native media format of the second type of sample recommendation information to the target media format;

[0229] The first type of sample recommendation information and the second type of sample recommendation information in the target media format are subjected to media synthesis to obtain a sample recommendation information stream.

[0230] In this embodiment, the native media format includes at least one of text format, image format, graphic and text format, audio format or video format; the target media format includes at least one of text format, image format, graphic and text format, audio format or video format; the sample recommendation information flow includes at least one of synthetic text information flow, synthetic image, synthetic graphic and text information flow, synthetic audio or synthetic video.

[0231] Step S1212, obtaining the expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle, performing model training based on the sample recommended information flow and the expected recommended information flow, and obtaining an information push model.

[0232] Among them, it is expected that the recommended information flow will be used as the information flow label for model training.

[0233] Specifically, the computer device obtains the expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle, and determines the information flow loss between the sample recommended information flow and the expected recommended information flow. Model training is performed based on the information flow loss to obtain the information push model. Further, model training is performed based on the information flow loss, and when the number of model training times reaches a preset number of training times, the training stops, and the information push model is obtained.

[0234] In this embodiment, the information flow loss represents the difference between the sample recommended information flow and the expected recommended information flow. The information flow loss can be determined by the similarity between the sample recommended information flow and the expected recommended information flow; the higher the similarity, the smaller the information flow loss; the lower the similarity, the greater the information flow loss.

[0235] In this embodiment, the computer device obtains the model loss function, determines the target loss according to the model loss function and the information flow loss, performs model training based on the target loss, and obtains the information push model.

[0236] In this embodiment, when the target loss does not meet the training stop condition, the model parameters are adjusted based on the target loss and training continues. When the target loss meets the training stop condition, the training stops and the information push model is obtained.

[0237] The training stopping condition can be that the target loss is less than or equal to the loss threshold.

[0238] In this embodiment, the sample information browsing records of the sample receiving objects in the same sample time period in multiple sample history cycles are obtained to obtain the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, as well as the sample browsing time record of browsing the sample history information from the sample information browsing records. The sample category feature is extracted based on the sample history information category in the sample information browsing record to predict the sample push information category corresponding to the sample receiving object based on the sample category feature, and the information category that the user is interested in in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing habits of information categories in a specific sample time period. The sample history duration feature is extracted based on the browsing time record in the sample information browsing record to predict the sample predicted browsing time corresponding to the sample receiving object based on the sample history duration feature, and the time spent by the user on information browsing in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing time habits. With reference to the sample predicted browsing time, multiple sample recommendation information of the sample push information category is obtained, so that it is possible to select multiple sample recommendation information from the predicted information category in combination with the predicted time that the user may spend on information browsing, so as to generate a predicted sample recommendation information flow, and the total browsing time of the sample recommendation information flow matches the sample predicted browsing time. The expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle represents the information flow that the expected model can output during training, and is the goal of model learning. Model training is performed based on the sample recommendation information flow generated by the model and the recommended information flow generated by the expected model, so that the recommended information flow output by the model during training gradually approaches the expected recommended information flow, thereby obtaining an information push model. The obtained information push model can automatically predict the information category and browsing time of the user in a specific time period. Based on the predicted browsing time, multiple recommended information belonging to the predicted information category is selected to generate a recommended information flow including multiple recommended information, which is convenient for the user to browse. The total browsing time of the generated recommended information flow matches the predicted browsing time, so that the total browsing time of the generated recommended information flow can meet the browsing time habits of the user in a specific time period. The generated recommended information flow is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information flow. Automatic prediction and information flow generation through the information push model can effectively improve the speed of information push.

[0239] Moreover, it enables users to browse the recommended information stream within their usual browsing time. Moreover, the information stream includes multiple recommended information, so that users can obtain more interesting information in one information stream, which helps users quickly understand the relevant content of the information stream, saves browsing time and improves the comfort of browsing the information stream.

[0240] In this embodiment, the information push model is predicted based on logistic regression, using information category and browsing time as dimensions, and selecting n days of data as training data. If a user reads a certain information category in one day, this information category is marked as 1, otherwise it is marked as 0. If the user continues to read this information category on the xth day after n days, this information category is marked as 1, otherwise it is marked as 0. Select n days of data under an information category for each user as the independent variable, and use the result value of the same information category as the dependent variable to construct the information push model. Input the sequence of information categories read by the user to be predicted for n days, and the information push model can predict the user's category prediction score for each information category. Input the browsing time of each information category read by the user to be predicted for n days, and the information push model can predict the browsing time of the user in each time period.

[0241] In one embodiment, model training is performed based on the sample recommendation information flow and the expected recommendation information flow to obtain the information push model, including:

[0242] Determine the information flow loss between the sample recommendation information flow and the expected recommendation information flow; obtain the expected push information category corresponding to the sample receiving object in the sample plan push cycle, and determine the category loss between the sample push information category and the expected push information category; perform model training based on the information flow loss and category loss to obtain the information push model.

[0243] The expected push information category is used as the category label for model training, representing the target of the push information category predicted by the expected model.

[0244] The category loss represents the difference between the category of the sample push information and the category of the expected push information. The category loss can be determined by the similarity between the category of the sample push information and the category of the expected push information; the higher the similarity, the smaller the category loss; the lower the similarity, the greater the category loss.

[0245] Specifically, the computer device determines the information flow loss between the sample recommendation information flow and the expected recommendation information flow. The computer device obtains the expected push information category corresponding to the sample receiving object in the sample plan push cycle, calculates the similarity between the sample push information category and the expected push information category, and determines the category loss based on the similarity. For example, a similarity of 90% indicates a category loss of 10%.

[0246] The computer device determines the target loss according to the information flow loss and the category loss, performs model training based on the target loss, and obtains the information push model.

[0247] In this embodiment, a model loss function may be obtained, and the information flow loss and the category loss may be substituted into the model loss function to obtain a target loss.

[0248] In this embodiment, the information flow loss characterizes the difference between the recommended information flow predicted by the model and the information flow learned by the expected model. The category loss characterizes the difference between the push information category predicted by the model and the push information category learned by the expected model. Model training is performed based on the information flow loss and the category loss, so that the obtained information push model is more accurate in both category prediction and recommended information flow generation.

[0249] In one embodiment, model training is performed based on information flow loss and category loss to obtain an information push model, including:

[0250] Obtain the expected browsing time of the sample recipients during the sample plan push cycle, and determine the time loss between the sample predicted browsing time and the expected browsing time; perform model training based on information flow loss, category loss, and time loss to obtain an information push model.

[0251] The duration loss represents the difference between the sample predicted browsing duration and the expected browsing duration. The expected browsing duration is used as the duration label for model training, representing the target of the browsing duration predicted by the expected model.

[0252] Specifically, the computer device obtains the expected browsing time of the sample receiving object corresponding to the sample plan push period, and takes the difference or ratio between the sample predicted browsing time and the expected browsing time as the time loss.

[0253] The computer device determines the target loss according to the information flow loss, category loss and duration loss, performs model training based on the target loss and obtains the information push model.

[0254] In this embodiment, a model loss function may be obtained, and information flow loss, category loss, and duration loss may be substituted into the model loss function to obtain a target loss.

[0255] In this embodiment, the time loss represents the difference between the browsing time predicted by the model and the browsing time learned by the expected model. Model training is performed based on information flow loss, category loss and time loss, so that the obtained information push model can be more accurate in category prediction, time prediction and recommended information flow generation.

[0256] In one embodiment, an information push method is provided, which is applied to a computer device and includes:

[0257] Obtaining the sample information browsing records of each sample receiving object in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving object in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0258] Extract sample category features based on sample historical information categories in sample information browsing records, and predict the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period during the sample plan push cycle;

[0259] Extract the sample history duration feature based on the browsing time record in the sample information browsing record, and predict the sample predicted browsing time corresponding to the sample receiving object based on the sample history duration feature; the sample predicted browsing time represents the browsing time of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0260] Refer to the sample prediction browsing time to obtain multiple sample recommendation information of the sample push information category;

[0261] Generate a sample recommendation information stream including multiple sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0262] Obtain the expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle, and determine the information flow loss between the sample recommended information flow and the expected recommended information flow;

[0263] Obtain the expected push information category corresponding to the sample receiving object in the sample plan push cycle, and determine the category loss between the sample push information category and the expected push information category;

[0264] Obtain the expected browsing time of the sample recipient corresponding to the sample plan push cycle, and determine the time loss between the sample predicted browsing time and the expected browsing time;

[0265] Model training is performed based on information flow loss, category loss, and duration loss to obtain an information push model.

[0266] Obtaining the information browsing records of the push recipients in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipients in the time period belongs, and the browsing time record of the browsing historical information;

[0267] Through the information push model, historical category features are extracted based on historical information categories in the information browsing records, and candidate category features extracted from multiple candidate information categories are obtained;

[0268] Through the information push model, for each candidate information category, based on the candidate category features and historical category features of the candidate information category, the predicted category score corresponding to the candidate information category is predicted; the predicted category score represents the probability that the information flow to be pushed includes information of the candidate information category;

[0269] Through the information push model, a push information category is screened out from a variety of candidate information categories, and the predicted category score of the push information category meets the preset category push conditions; the push information category represents the category of information in the information flow that needs to be pushed in the time period of the planned push cycle.

[0270] Through the information push model, historical duration features are extracted based on the browsing time records in the information browsing records to obtain multiple candidate browsing times; for each candidate browsing time, a predicted duration score corresponding to the candidate browsing time is predicted based on the candidate browsing time and historical duration features; from the multiple candidate browsing times, the predicted browsing time is screened, and the predicted duration score corresponding to the predicted browsing time meets the preset browsing time condition; the predicted browsing time represents the browsing time of the information flow to be pushed in the time period in the planned push cycle.

[0271] Obtain a candidate information set belonging to the push information category through the information push model, and determine the information browsing time corresponding to each candidate information in the candidate information set;

[0272] By using the information push model, at least two combinations are performed on the candidate information in the candidate information set to form at least two information groups, each of which includes at least two candidate information;

[0273] Through the information push model, according to the information browsing time of the candidate information in each information group, the total candidate browsing time corresponding to each information group is determined;

[0274] An information group corresponding to a candidate total browsing time that matches the predicted browsing time is determined through an information push model, and the candidate information included in the determined information group is determined as a plurality of recommended information of a push information category.

[0275] Determine the native media format of each of the plurality of recommended information through the information push model; determine the target media format of the recommended information flow to be generated through the information push model;

[0276] Determine, from the plurality of recommendation information, first category recommendation information and second category recommendation information through the information push model, wherein the native media format of the first category recommendation information and the target media format have a media synthesis condition, and the native media format of the second category recommendation information and the target media format do not have a media synthesis condition;

[0277] By using the information push model, according to the format conversion method from the native media format of the second type of recommended information to the target media format, the second type of recommended information is converted into the target media format;

[0278] Through the information push model, the first type of recommended information and the second type of recommended information in the target media format are synthesized to obtain a recommended information flow, and the total browsing time of the recommended information flow matches the predicted browsing time; the recommended information flow is used to push to the push recipients within the time period of the planned push cycle.

[0279] In this embodiment, the sample information browsing records of the sample receiving objects in the same sample time period in multiple sample history cycles are obtained to obtain the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, as well as the sample browsing time record of browsing the sample history information from the sample information browsing records. The sample category feature is extracted based on the sample history information category in the sample information browsing record to predict the sample push information category corresponding to the sample receiving object based on the sample category feature, and the information category that the user is interested in in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing habits of information categories in a specific sample time period. The sample history duration feature is extracted based on the browsing time record in the sample information browsing record to predict the sample predicted browsing time corresponding to the sample receiving object based on the sample history duration feature, and the time spent by the user on information browsing in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing time habits. Referring to the sample predicted browsing time, multiple sample recommendation information of the sample push information category is obtained, so that it is possible to select multiple sample recommendation information from the predicted information category in combination with the predicted time that the user may spend on information browsing, so as to generate a predicted sample recommendation information flow, and the total browsing time of the sample recommendation information flow matches the sample predicted browsing time. The expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle represents the information flow that the expected model can output during training, and is the goal of model learning. The duration loss represents the difference between the browsing time predicted by the model and the browsing time learned by the expected model. Model training is performed based on information flow loss, category loss and duration loss, so that the obtained information push model can be more accurate in category prediction, duration prediction and recommendation information flow generation.

[0280] The information push model obtained can automatically predict the information category and browsing time of users in a specific time period. Based on the predicted browsing time, multiple recommended information belonging to the predicted information category is selected to generate a recommended information stream including multiple recommended information for user browsing. The total browsing time of the generated recommended information stream matches the predicted browsing time, so that the total browsing time of the generated recommended information stream can meet the browsing time habits of users in a specific time period. The generated recommended information stream is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information stream. Automatic prediction and information stream generation through the information push model can effectively improve the speed of information push.

[0281] Moreover, it enables users to browse the recommended information stream within their usual browsing time. Moreover, the information stream includes multiple recommended information, so that users can obtain more interesting information in one information stream, which helps users quickly understand the relevant content of the information stream, saves browsing time and improves the comfort of browsing the information stream.

[0282] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0283] Based on the same inventive concept, the embodiment of the present application also provides an information push device for implementing the information push method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more information push device embodiments provided below can refer to the limitations on the information push method above, and will not be repeated here.

[0284] In one embodiment, Fig.13 As shown, an information push device 1300 is provided, comprising: a record acquisition module 1302, a category prediction module 1304, a duration prediction module 1306, an information acquisition module 1308 and a generation module 1310, wherein:

[0285] The record acquisition module 1302 is used to acquire the information browsing records of the push recipients in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipients in the time period belongs, and the browsing time record of the browsing historical information.

[0286] The category prediction module 1304 is used to extract historical category features based on the historical information categories in the information browsing records, and predict the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period of the planned push cycle.

[0287] The duration prediction module 1306 is used to extract historical duration features based on the browsing time records in the information browsing records, and predict the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle.

[0288] The information acquisition module 1308 is used to obtain multiple recommended information of the push information category with reference to the predicted browsing time.

[0289] The generation module 1310 is used to generate a recommendation information stream including multiple recommendation information, and the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object within the time period of the planned push cycle.

[0290] In this embodiment, the information browsing records of the push receiving object in the same time period in multiple historical cycles are obtained, so as to obtain the historical information category to which the historical information browsed by the push receiving object in the time period belongs, and the browsing time record of the browsing historical information from the information browsing record. The historical category feature is extracted based on the historical information category in the information browsing record, so as to predict the push information category corresponding to the push receiving object based on the historical category feature, and the information category that the user is interested in in the same time period of the planned push cycle can be accurately predicted based on the information category habit of the user's historical browsing in a specific time period. The historical duration feature is extracted based on the browsing time record in the information browsing record, so as to predict the predicted browsing duration corresponding to the push receiving object based on the historical duration feature, and the time spent by the user on information browsing in the same time period of the planned push cycle can be accurately predicted based on the user's historical browsing duration habit. With reference to the predicted browsing duration, multiple recommended information of the push information category is obtained, so that multiple recommended information can be selected from the information category of interest to the push receiving object in combination with the time that the user may spend on information browsing, so as to generate a recommended information flow including multiple recommended information, which is convenient for the user to browse. The total browsing time of the generated recommended information stream matches the predicted browsing time, so that the total browsing time of the generated recommended information stream can meet the browsing time habits of users in a specific time period. The generated recommended information stream is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information streams, and enabling users to complete browsing of the recommended information stream within their accustomed browsing time. Moreover, the information stream includes multiple recommended information, so that users can obtain more information of interest in one information stream, which not only helps users quickly understand the relevant content of the information stream, but also helps users save browsing time and improves the comfort of browsing the information stream.

[0291] In one of the embodiments, the category prediction module 1304 is also used to obtain candidate category features extracted from multiple candidate information categories; for each candidate information category, based on the candidate category features and historical category features of the targeted candidate information category, predict the predicted category score corresponding to the targeted candidate information category; the predicted category score represents the probability that the information flow to be pushed includes information of the targeted candidate information category; from the multiple candidate information categories, filter out the push information category, and the predicted category score of the push information category meets the preset category push condition.

[0292] In this embodiment, candidate category features extracted from multiple candidate information categories are obtained, and for each candidate information category, a predicted category score corresponding to the candidate information category is predicted based on the candidate category features and historical category features of the candidate information category. The probability of a user browsing each candidate information category can be predicted, and the push information category that the user is most likely to be interested in during the time period in the planned push cycle can be screened out from the multiple candidate information categories based on these probabilities, so that the information of interest can be accurately pushed to the user.

[0293] In one of the embodiments, the information acquisition module 1308 is also used to obtain a candidate information set belonging to the push information category, determine the information browsing time corresponding to each candidate information in the candidate information set; perform at least two combinations of the candidate information in the candidate information set to form at least two information groups, each information group including at least two candidate information; determine the total candidate browsing time corresponding to each information group based on the information browsing time of the candidate information in each information; determine the information group corresponding to the candidate total browsing time that matches the predicted browsing time, and determine the candidate information included in the determined information group as multiple recommended information of the push information category.

[0294] In this embodiment, a candidate information set belonging to the push information category is obtained, the information browsing time corresponding to each candidate information in the candidate information set is determined, and at least two combinations are performed on the candidate information in the candidate information set to form at least two information groups, each of which includes at least two candidate information, so that the candidate total browsing time corresponding to each information group can be accurately determined according to the information browsing time of the candidate information in each information group. An information group corresponding to the candidate total browsing time matching the predicted browsing time is determined, so that the candidate information included in the determined information group is determined as multiple recommended information of the push information category, so that the recommended information can be filtered based on the information browsing time of each candidate information, so that the total browsing time of the filtered multiple recommended information can match the predicted browsing time, so that the total browsing time of the generated recommended information flow meets the user's information browsing time habits, so that the recommended information flow of corresponding time can be generated according to the user's browsing time habits, so as to improve the user's browsing interest in the pushed information flow and help improve the conversion rate of the information flow.

[0295] In one embodiment, the generation module 1310 is further used to determine the native media format of each of the multiple recommendation information; determine the target media format of the recommendation information stream to be generated; and perform media synthesis on the multiple recommendation information according to the native media format of each of the multiple recommendation information and the target media format to obtain the recommendation information stream.

[0296] In this embodiment, the native media format of each of the multiple recommendation information is determined, the target media format of the recommendation information stream to be generated is determined, and the multiple recommendation information are subjected to media synthesis based on the native media format of each of the multiple recommendation information and the target media format, so as to accurately generate the recommendation information stream in the target media format.

[0297] In one of the embodiments, the generation module 1310 is further used to determine first-category recommendation information and second-category recommendation information from multiple recommendation information, the native media format of the first-category recommendation information and the target media format having a media synthesis condition, and the native media format of the second-category recommendation information and the target media format not having a media synthesis condition; converting the second-category recommendation information into the target media format according to a format conversion method from the native media format of the second-category recommendation information to the target media format; and performing media synthesis on the first-category recommendation information and the second-category recommendation information in the target media format to obtain a recommendation information stream.

[0298] In this embodiment, first-category recommendation information and second-category recommendation information are determined from multiple recommendation information to determine first-category recommendation information whose native media format and target media format have media synthesis conditions, and second-category recommendation information whose native media format and target media format do not have media synthesis conditions, so as to determine which recommendation information needs to be format converted. According to the format conversion method from the native media format to the target media format of the second-category recommendation information, the second-category recommendation information can be accurately converted into the target media format, so that the two types of recommendation information can be media synthesized, thereby multiple recommendation information in different forms can be synthesized into one recommendation information flow, which is convenient for users to browse.

[0299] In one of the embodiments, the native media format includes at least one of a text format, an image format, a graphic-text format, an audio format, or a video format; the target media format includes at least one of a text format, an image format, a graphic-text format, an audio format, or a video format; and the recommended information stream includes at least one of a synthesized text information stream, a synthesized image, a synthesized graphic-text information stream, a synthesized audio, or a synthesized video.

[0300] In this embodiment, text format, image format, graphic format, audio format and video format can be converted to each other, so that the recommended information that needs format conversion can be converted from the native media format to the target media format, so that multiple recommended information in different forms can be synthesized into the same recommended information stream for user browsing. In addition, the synthesized recommended information stream can be in the form of text, image, graphic, audio and video, etc., which can provide users with a variety of information browsing methods, so that the browsing methods can adapt to different application scenarios.

[0301] In one of the embodiments, the duration prediction module 1306 is also used to obtain multiple candidate browsing durations; for each candidate browsing duration, based on the candidate browsing duration and historical duration characteristics, predict the predicted duration score corresponding to the candidate browsing duration; from the multiple candidate browsing durations, filter the predicted browsing duration, and the predicted duration score corresponding to the predicted browsing duration meets the preset browsing duration condition.

[0302] In this embodiment, multiple candidate browsing times are obtained, and for each candidate browsing time, a predicted time score corresponding to the candidate browsing time is predicted based on the candidate browsing time and historical time features. This can predict the user's interest level in each candidate browsing time during this time period in the planned push cycle, thereby screening out the browsing time that the user is most interested in during this time period in the planned push cycle, so that a recommended information flow of the corresponding browsing time can be generated for the user, which is in line with the user's browsing habits, helps to increase the user's interest in the pushed information, and thus improves the conversion rate of information.

[0303] In one embodiment, the generating module 1310 is further used to obtain a target media format indicated by a terminal corresponding to the push receiving object; the target media format is determined in response to a triggering operation on the format indication control when the terminal displays an information recommendation area including a format indication control;

[0304] Based on the multiple recommendation information, a recommendation information stream in a target media format is generated, the recommendation information stream includes the multiple recommendation information; the recommendation information stream is used to be displayed in the information recommendation area.

[0305] In this embodiment, the target media format indicated by the terminal corresponding to the push recipient is obtained to determine the media format of the recommended information stream to be pushed to the push recipient, thereby generating a recommended information stream of the target media format based on multiple recommended information, so that multiple recommended information are integrated into the same information stream for easy browsing by users. The recommended information stream pushed to the terminal corresponding to the push recipient is displayed in the information recommendation area, so that the recommended information stream is displayed in a specific area, effectively distinguishing other information, thereby avoiding interference from other information.

[0306] In one embodiment, the device further comprises:

[0307] The adjustment module is used to send the total browsing time to the terminal corresponding to the push receiving object; obtain the total browsing time after the total browsing time is adjusted; refer to the total browsing time after the time adjustment, adjust the recommended information included in the recommended information flow, and obtain the adjusted recommended information flow; the adjusted recommended information flow is used to push to the push receiving object in the time period within the planned push cycle.

[0308] In this embodiment, the total browsing time is sent to the terminal corresponding to the push receiving object, so that the push receiving object can adjust the total browsing time. The total browsing time after the total browsing time is adjusted is obtained, and the recommended information included in the recommended information flow is adjusted with reference to the total browsing time after the time adjustment, so as to obtain the adjusted recommended information flow, thereby providing the user with the function of adjusting the browsing time of the information flow, so that the user can adjust the browsing time of the information flow according to the needs, making the information flow recommendation more flexible.

[0309] Based on the same inventive concept, the embodiment of the present application also provides an information push model processing device for implementing the information push model processing method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more information push model processing device embodiments provided below can refer to the limitations on the information push model processing method above, and will not be repeated here.

[0310] In one embodiment, Fig.14As shown, an information push model processing device 1400 is provided, including: a sample record acquisition module 1402, a sample category prediction module 1404, a sample duration prediction module 1406, a sample information acquisition module 1408, a sample generation module 1410 and a training module 1412, wherein:

[0311] The sample record acquisition module 1402 is used to acquire the sample information browsing records of each sample receiving object in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving object in the sample time period belongs, and the sample browsing time record of browsing the sample history information;

[0312] The sample category prediction module 1404 is used to extract sample category features based on the sample historical information categories in the sample information browsing records, and predict the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle;

[0313] The sample duration prediction module 1406 is used to extract the sample history duration feature based on the browsing time record in the sample information browsing record, and predict the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration feature; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle;

[0314] The sample information acquisition module 1408 is used to obtain multiple sample recommendation information of the sample push information category with reference to the sample predicted browsing time;

[0315] The sample generation module 1410 is used to generate a sample recommendation information stream including a plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time;

[0316] The training module 1412 is used to obtain the expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle, perform model training based on the sample recommended information flow and the expected recommended information flow, and obtain an information push model.

[0317] In this embodiment, the sample information browsing records of the sample receiving objects in the same sample time period in multiple sample history cycles are obtained to obtain the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, as well as the sample browsing time record of browsing the sample history information from the sample information browsing records. The sample category feature is extracted based on the sample history information category in the sample information browsing record to predict the sample push information category corresponding to the sample receiving object based on the sample category feature, and the information category that the user is interested in in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing habits of information categories in a specific sample time period. The sample history duration feature is extracted based on the browsing time record in the sample information browsing record to predict the sample predicted browsing time corresponding to the sample receiving object based on the sample history duration feature, and the time spent by the user on information browsing in the same time period of the sample plan push cycle can be accurately predicted based on the user's historical browsing time habits. With reference to the sample predicted browsing time, multiple sample recommendation information of the sample push information category is obtained, so that it is possible to select multiple sample recommendation information from the predicted information category in combination with the predicted time that the user may spend on information browsing, so as to generate a predicted sample recommendation information flow, and the total browsing time of the sample recommendation information flow matches the sample predicted browsing time. The expected recommended information flow corresponding to the sample receiving object in the sample plan push cycle represents the information flow that the expected model can output during training, and is the goal of model learning. Model training is performed based on the sample recommendation information flow generated by the model and the recommended information flow generated by the expected model, so that the recommended information flow output by the model during training gradually approaches the expected recommended information flow, thereby obtaining an information push model. The obtained information push model can automatically predict the information category and browsing time of the user in a specific time period. Based on the predicted browsing time, multiple recommended information belonging to the predicted information category is selected to generate a recommended information flow including multiple recommended information, which is convenient for the user to browse. The total browsing time of the generated recommended information flow matches the predicted browsing time, so that the total browsing time of the generated recommended information flow can meet the browsing time habits of the user in a specific time period. The generated recommended information flow is pushed to users in the same time period within the planned push cycle, realizing intelligent push of information flow. Automatic prediction and information flow generation through the information push model can effectively improve the speed of information push.

[0318] Moreover, it enables users to browse the recommended information stream within their usual browsing time. Moreover, the information stream includes multiple recommended information, so that users can obtain more interesting information in one information stream, which helps users quickly understand the relevant content of the information stream, saves browsing time and improves the comfort of browsing the information stream.

[0319] In one of the embodiments, the training module 1412 is also used to determine the information flow loss between the sample recommendation information flow and the expected recommendation information flow; obtain the expected push information category corresponding to the sample receiving object in the sample plan push cycle, and determine the category loss between the sample push information category and the expected push information category; perform model training based on the information flow loss and category loss to obtain an information push model.

[0320] In this embodiment, the information flow loss characterizes the difference between the recommended information flow predicted by the model and the information flow learned by the expected model. The category loss characterizes the difference between the push information category predicted by the model and the push information category learned by the expected model. Model training is performed based on the information flow loss and the category loss, so that the obtained information push model is more accurate in both category prediction and recommended information flow generation.

[0321] In one of the embodiments, the training module 1412 is also used to obtain the expected browsing time of the sample receiving object corresponding to the sample plan push cycle, determine the time loss between the sample predicted browsing time and the expected browsing time; perform model training based on information flow loss, category loss and time loss to obtain an information push model.

[0322] In this embodiment, the time loss represents the difference between the browsing time predicted by the model and the browsing time learned by the expected model. Model training is performed based on information flow loss, category loss and time loss, so that the obtained information push model can be more accurate in category prediction, time prediction and recommended information flow generation.

[0323] Each module in the above-mentioned information push device and information push model processing device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0324] In one embodiment, a computer device is provided, which may be a server or a terminal. Taking a server as an example, its internal structure diagram may be as follows: Fig.15As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store relevant data of the information push method and relevant processing of the information push model processing method. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it can implement an information push method and can also implement an information push model processing method.

[0325] Those skilled in the art will understand that Fig.15 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0326] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0327] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0328] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0329] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions. In addition, users can refuse or can easily refuse to push information.

[0330] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processors involved in each embodiment provided in this application may be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc., but are not limited thereto.

[0331] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0332] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. An information push method, characterized in that: The method comprises: Acquire the information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information; Extracting historical category features based on the historical information categories in the information browsing records, and predicting the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle; Extracting historical duration features based on the browsing time record in the information browsing record, and predicting the predicted browsing duration corresponding to the push recipient based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle; Referring to the predicted browsing time, obtaining a plurality of recommended information of the push information category; Generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

2. The method according to claim 1, characterized in that The predicting, based on the historical category feature, the push information category corresponding to the push recipient object includes: Obtaining candidate category features extracted from a plurality of candidate information categories; For each candidate information category, based on the candidate category features of the candidate information category and the historical category features, predict a predicted category score corresponding to the candidate information category; the predicted category score represents the probability that the information flow to be pushed includes information of the candidate information category; From the multiple candidate information categories, a push information category is screened out, and the predicted category score of the push information category meets the preset category push condition.

3. The method according to claim 1, characterized in that The obtaining of a plurality of recommended information of the push information category by referring to the predicted browsing time includes: Acquire a candidate information set belonging to the push information category, and determine the information browsing time corresponding to each candidate information in the candidate information set; Perform at least two combinations on the candidate information in the candidate information set to form at least two information groups, each information group including at least two of the candidate information; Determine the total candidate browsing time corresponding to each information group according to the information browsing time of the candidate information in each information group; An information group corresponding to a candidate total browsing duration that matches the predicted browsing duration is determined, and the candidate information included in the determined information group is determined as a plurality of recommended information of the push information category.

4. The method according to claim 1, characterized in that: The generating of the recommendation information flow including the plurality of recommendation information includes: Determining a native media format of each of the plurality of recommendation information; Determining a target media format for the recommendation information stream to be generated; The plurality of recommendation information are subjected to media synthesis according to respective native media formats of the plurality of recommendation information and the target media format to obtain a recommendation information stream.

5. The method according to claim 4, characterized in that The step of performing media synthesis on the plurality of recommendation information according to respective native media formats of the plurality of recommendation information and the target media format to obtain a recommendation information stream includes: Determine, from the plurality of recommendation information, first category recommendation information and second category recommendation information, wherein a native media format of the first category recommendation information and the target media format have a media synthesis condition, and a native media format of the second category recommendation information and the target media format do not have a media synthesis condition; converting the second type of recommendation information into the target media format according to a format conversion method from a native media format of the second type of recommendation information to the target media format; The first type of recommendation information and the second type of recommendation information in the target media format are subjected to media synthesis to obtain a recommendation information stream.

6. The method according to claim 5, characterized in that The native media format includes at least one of a text format, an image format, a graphic-text format, an audio format, or a video format; the target media format includes at least one of a text format, an image format, a graphic-text format, an audio format, or a video format; the recommended information stream includes at least one of a synthesized text information stream, a synthesized image, a synthesized graphic-text information stream, a synthesized audio, or a synthesized video.

7. The method according to claim 1, characterized in that The predicting the predicted browsing duration corresponding to the push receiving object based on the historical duration feature includes: Get multiple candidate browsing durations; For each of the candidate browsing durations, based on the candidate browsing duration and the historical duration features, predict a predicted duration score corresponding to the candidate browsing duration; From the multiple candidate browsing durations, a predicted browsing duration is screened, and a predicted duration score corresponding to the predicted browsing duration satisfies a preset browsing duration condition.

8. The method according to any one of claims 1 to 7, characterized in that: The generating of the recommendation information flow including the plurality of recommendation information includes: Acquiring a target media format indicated by a terminal corresponding to the push receiving object; the target media format is determined in response to a triggering operation on the format indication control when the terminal displays an information recommendation area including a format indication control; Based on the multiple recommendation information, a recommendation information stream in the target media format is generated, the recommendation information stream includes the multiple recommendation information; the recommendation information stream is used to be displayed in the information recommendation area.

9. The method according to any one of claims 1 to 7, characterized in that: The method further comprises: Sending the total browsing time to the terminal corresponding to the push receiving object; Obtaining the total browsing time after adjusting the total browsing time; Referring to the total browsing time after the time adjustment, the recommendation information included in the recommendation information flow is adjusted to obtain the adjusted recommendation information flow; the adjusted recommendation information flow is used to push to the push receiving object in the time period within the planned push cycle.

10. A method for processing an information push model, characterized in that: The method comprises: Obtaining respective sample information browsing records of sample receiving objects in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving objects in the sample time period belongs, and the sample browsing time record of browsing the sample history information; Extracting sample category features based on the sample historical information categories in the sample information browsing records, and predicting the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle; Extracting sample history duration features based on the browsing time record in the sample information browsing record, and predicting the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle; Referring to the predicted browsing time of the sample, obtaining a plurality of sample recommendation information of the sample push information category; Generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time; The expected recommended information flow corresponding to the sample receiving object in the sample plan push period is obtained, and model training is performed based on the sample recommended information flow and the expected recommended information flow to obtain an information push model.

11. The method according to claim 10, characterized in that The performing model training based on the sample recommendation information flow and the expected recommendation information flow to obtain the information push model includes: Determining an information flow loss between the sample recommendation information flow and the expected recommendation information flow; Obtain the expected push information category of the sample receiving object corresponding to the sample plan push period, and determine the category loss between the sample push information category and the expected push information category; Model training is performed based on the information flow loss and the category loss to obtain an information push model.

12. The method according to claim 11, characterized in that The performing model training based on the information flow loss and the category loss to obtain the information push model includes: Obtain the expected browsing time of the sample receiving object corresponding to the sample plan push period, and determine the time loss between the sample predicted browsing time and the expected browsing time; Model training is performed based on the information flow loss, the category loss and the duration loss to obtain an information push model.

13. An information push device, characterized in that: The device comprises: A record acquisition module, used to acquire information browsing records of each push recipient in the same time period in multiple historical cycles; the information browsing records include the historical information category to which the historical information browsed by the push recipient in the time period belongs, and the browsing time record of browsing the historical information; A category prediction module, used to extract historical category features based on the historical information categories in the information browsing records, and predict the push information category corresponding to the push receiving object based on the historical category features; the push information category represents the category of information in the information flow to be pushed in the time period in the planned push cycle; A duration prediction module is used to extract historical duration features based on the browsing time records in the information browsing records, and predict the predicted browsing duration corresponding to the push receiving object based on the historical duration features; the predicted browsing duration represents the browsing duration of the information flow to be pushed in the time period in the planned push cycle; An information acquisition module, used to obtain a plurality of recommended information of the push information category with reference to the predicted browsing time; A generation module is used to generate a recommendation information stream including the multiple recommendation information, wherein the total browsing time of the recommendation information stream matches the predicted browsing time; the recommendation information stream is used to push to the push receiving object in the time period within the planned push cycle.

14. An information push model processing device, characterized in that: The device comprises: A sample record acquisition module, used to acquire the sample information browsing records of each sample receiving object in the same sample time period in multiple sample history cycles; the sample information browsing records include the sample history information category to which the sample history information browsed by the sample receiving object in the sample time period belongs, and the sample browsing time record of browsing the sample history information; A sample category prediction module is used to extract sample category features based on the sample historical information category in the sample information browsing record, and predict the sample push information category corresponding to the sample receiving object based on the sample category features; the sample push information category represents the category of information in the sample information flow that needs to be pushed in the sample time period in the sample plan push cycle; A sample duration prediction module is used to extract sample history duration features based on the browsing time record in the sample information browsing record, and predict the sample predicted browsing duration corresponding to the sample receiving object based on the sample history duration features; the sample predicted browsing duration represents the browsing duration of the sample information flow to be pushed in the sample time period in the sample plan push cycle; A sample information acquisition module, used to obtain a plurality of sample recommendation information of the sample push information category with reference to the sample predicted browsing time; A sample generation module, configured to generate a sample recommendation information stream including the plurality of sample recommendation information, wherein the total browsing time of the sample recommendation information stream matches the sample predicted browsing time; The training module is used to obtain the expected recommended information flow corresponding to the sample receiving object in the sample plan push period, perform model training based on the sample recommended information flow and the expected recommended information flow, and obtain an information push model.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 12 are implemented.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

Images