Data processing method and apparatus, and device, computer-readable storage medium and computer program product

Abstract

Disclosed in the embodiments of the present application are a data processing method and apparatus, and a device and a computer-readable storage medium. The method comprises: on the basis of a pre-loading request, acquiring a first pre-loading usage probability corresponding to a terminal device, wherein the first pre-loading usage probability is used for indicating a probability that a target object interacts with pre-loaded content; and if the first pre-loading usage probability is equal to or greater than an online recommendation threshold, generating a set of online recommended content, and on the basis of the set of online recommended content, generating a list of target pre-loaded content, wherein a generation timestamp of the set of online recommended content is later than a generation timestamp of the pre-loading request. By means of the present application, not only can computing resources be reduced, but the utilization rate of online recommended content can also be increased.

Description

A data processing method, device, apparatus, computer-readable storage medium, and computer program product

[0001] Cross-reference to Related Applications

[0002] The present application is based on Chinese Patent Application No. 202410897362.9, filed on July 5, 2024, and claims priority to the above Chinese Patent Application, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present application relates to the technical field of the Internet, and in particular to a data processing method, device, apparatus, computer-readable storage medium, and computer program product. BACKGROUND

[0004] With the continuous development of Internet technology, various applications have sprung up like mushrooms. People can watch a variety of content through various applications, such as browsing information, watching videos, and browsing novels, etc.

[0005] In order to achieve a high-quality experience of information flow (feed flow, also known as content flow), in the prior art, a terminal device initiates a preloading request before a user consumes content, so that the application background performs online content recommendation according to the preloading request, and then the terminal device can perform online content preloading, that is, online content preloading. However, preloading online content requires a complete recommendation process, which consumes a lot of computing resources. In addition, since the preloading request is initiated before the user browses the content, not every online recommended content corresponding to the preloading request will be used by the user, so there are a large number of useless preloading requests. Obviously, using the prior art not only reduces the use rate of online recommended content and the timeliness of recommended content, thereby reducing the freshness and attractiveness of the content recommended to the user, resulting in poor user stickiness, but also wastes a lot of computing resources and storage resources of the application background. SUMMARY

[0006] The embodiments of the present application provide a data processing method, device, apparatus, computer-readable storage medium, and computer program product, which can not only reduce the occupation of computing resources, but also improve the timeliness of content recommendation.

[0007] The embodiments of the present application provide a data processing method, applied to a computer device, comprising:

[0008] obtaining a preloading request sent by a terminal device, and obtaining a first preloading use probability corresponding to the terminal device according to the preloading request; the first preloading use probability is used to indicate a probability that a target object corresponding to the terminal device interacts with content provided by a preloading function; the preloading function is used to respond to the preloading request;

[0009] If the first preloading usage probability is equal to or greater than the online recommendation threshold, an online recommendation content set for the terminal device is generated, and a target preloading content list is generated according to the online recommendation content set; a generation timestamp of the online recommendation content set is later than a generation timestamp of the preloading request.

[0010] Embodiments of the present application provide a data processing apparatus, comprising:

[0011] The acquisition module is configured to acquire a preloading request sent by a terminal device, and acquire a first preloading usage probability corresponding to the terminal device according to the preloading request; the first preloading usage probability is used to indicate a probability of a target object corresponding to the terminal device interacting with content provided by a preloading function; the preloading function is used to respond to the preloading request;

[0012] The generation module is further configured to, if the first preloading usage probability is equal to or greater than an online recommendation threshold, generate an online recommendation content set for the terminal device, and generate a target preloading content list according to the online recommendation content set; a generation timestamp of the online recommendation content set is later than a generation timestamp of the preloading request.

[0013] Embodiments of the present application provide a computer device, comprising: a processor, a memory, and a network interface;

[0014] The processor is connected with the memory and the network interface, wherein the network interface is used to provide a data communication function, the memory is used to store a computer program, and the processor is used to call the computer program, so that the computer device executes the method in the embodiments of the present application.

[0015] In an aspect, the embodiments of the present application provide a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the method in the embodiments of the present application.

[0016] The embodiments of the present application provide a computer program product, which comprises a computer program stored in a computer readable storage medium; a processor of a computer device reads the computer program from the computer readable storage medium, and the processor executes the computer program, so that the computer device executes the method in the embodiments of the present application.

[0017] It can be known from the above that the embodiment of the application proposes a resource allocation method based on a first preloading use probability, and for a terminal device corresponding to the first preloading use probability equal to or higher than an online recommendation threshold, a set of online recommendation content generated in real time is provided, and then a preloading content list generated in real time can be provided. That is, by comparing the first preloading use probability and the online recommendation threshold, a target object with a high use probability can be allocated high-quality preloading content (i.e., recommendation content generated in real time), which can ensure the timeliness and content recommendation quality of the content recommended to the target object, thereby improving the attractiveness of the recommended content to the target object, promoting the target object to participate in content creation and interaction, improving the satisfaction and stickiness of the target object to the recommendation platform, and also improving the use rate of online recommendation content, increasing content diversity, enabling the target object to access real-time high-quality content, and recommending content in real time according to the first preloading use probability of the target object, which can make more rational use of computing resources, storage resources and network bandwidth, thereby improving resource utilization. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] FIG. 1 is a schematic diagram of a data processing system architecture provided by an embodiment of the present application;

[0020] FIG. 2 is a schematic diagram of a data processing scenario provided by an embodiment of the present application;

[0021] FIG. 3 is a schematic diagram of a data processing method provided by an embodiment of the present application;

[0022] FIG. 4 is a schematic diagram of a data processing scenario provided by an embodiment of the present application;

[0023] FIG. 5 is a schematic diagram of a data processing scenario provided by an embodiment of the present application;

[0024] FIG. 6 is a schematic diagram of a data processing method provided by an embodiment of the present application;

[0025] FIG. 7 is a schematic diagram of a data processing method provided by an embodiment of the present application;

[0026] FIG. 8 is an example graph of accumulation of preloading online recommendation content quota provided by an embodiment of the present application;

[0027] FIG. 9 is a schematic diagram of a data processing method provided by an embodiment of the present application;

[0028] FIG. 10 is a classification example of loading content provided by an embodiment of the present application;

[0029] FIG. 11 is a schematic diagram of a probability of using online recommended content randomly distributed provided by an embodiment of the present application;

[0030] FIG. 12 is a schematic diagram of a probability of using online recommended content distributed according to probability provided by an embodiment of the present application;

[0031] FIG. 13 is a structural schematic diagram of a data processing apparatus provided by an embodiment of the present application;

[0032] FIG. 14 is a structural schematic diagram of a computer device provided by an embodiment of the present application.

[0033] It should be noted that the above "first" and "second" are only used to distinguish different schemes, and do not represent the degree of superiority or priority in the implementation process. DETAILED DESCRIPTION

[0034] In order to clearly and completely describe the technical solutions in the embodiments of the present application, it should be apparent that the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of the present application.

[0035] For ease of understanding, first, some nouns are simply explained as follows:

[0036] 1) Preloaded content, before the user actively requests to play a video, the server uses the idle network bandwidth and processing capacity of the terminal device to transmit the video content from the server to the local cache of the terminal device in advance.

[0037] 2) Associated storage, the process of binding and persistently storing two or more data items (such as "first preloaded use probability" and "device identification") in a certain logical relationship. The core goal is to establish a searchable association between data, so as to facilitate subsequent query, analysis and application.

[0038] Please refer to FIG. 1, which is a schematic diagram of a data processing system architecture provided by an embodiment of the present application. As shown in FIG. 1, the system can include a business server 100 and a terminal device cluster, which can include: a terminal device 200a, a terminal device 200b, a terminal device 200c,..., and a terminal device 200n. It can be understood that the above system can include one or more terminal devices, and the present application does not limit the number of terminal devices.

[0039] Among them, there can be a communication connection between the terminal device clusters, for example, there is a communication connection between the terminal device 200a and the terminal device 200b, and there is a communication connection between the terminal device 200a and the terminal device 200c. At the same time, any terminal device in the terminal device cluster can have a communication connection with the service server 100, for example, the terminal device 200a and the service server 100 have a communication connection, wherein the above communication connection is not limited to the connection mode, which can be connected directly or indirectly through wired communication mode, or directly or indirectly through wireless communication mode, or through other ways, which is not limited in this application.

[0040] It should be understood that each terminal device in the terminal device cluster as shown in FIG. 1 can be installed with an application client, which can respectively interact with the service server 100 shown in FIG. 1 when running in each terminal device, that is, the above communication connection. Among them, the application client can be a video application, a social application, an instant messaging application, a game application, a music application, a shopping application, a novel application, a browser and other application clients with content preloading function. Among them, the application client can be an independent client, or an embedded sub-client integrated in a certain client (such as a social client, an education client and a multimedia client, etc.), which is not limited here.

[0041] Taking the video application as an example, the service server 100 can be a collection of multiple servers including the background server corresponding to the video application, data processing server, etc., so that each terminal device can transmit data with the service server 100 through the application client corresponding to the video application. For example, the terminal device 200a can send a preloading request to the service server 100 through the application client of the video application, and the service server 100 can generate a preloading content list for the terminal device 200a according to the preloading request, and return the preloading content list to the terminal device 200a.

[0042] It can be understood that in the specific embodiments of the present application, the user information (such as preloading request and pre-recommended content pool) and other related data are involved, when the embodiments in the present application are applied to specific products or technologies, the user's permission or consent is required, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards in relevant regions.

[0043] For the convenience of subsequent understanding and description, the embodiments of the present application can select one terminal device in the terminal device cluster shown in FIG. 1 as an example for description, for example, taking the terminal device 200a as an example for description. When starting the application client, and before outputting the content, the terminal device 200a can generate a preloading request in the application client, and send the preloading request to the service server 100 through the application client. It can be understood that in order to realize the high-quality experience of the information flow (feed flow, which can also be referred to as content flow), the application client can perform content preloading before starting and displaying the content output page, that is, the content is loaded before the target object corresponding to the terminal device 200a browses the content, so that the application client does not appear to be stuck or slow when the target object browses the content, and thus the viewing experience of the target object can be prompted.

[0044] Among them, preloading, that is, preloading, prepares the content before the target object corresponding to the terminal device really triggers the network request to pull the content from the background, and directly displays the preloaded content to the target object when the target object really triggers the loading. Preloading can make the target object experience the effect of instant loading.

[0045] As can be seen from the above, the preloading request described in the embodiments of the present application occurs before the content output, and is used to request the service server 100 to pre-load the content. The embodiments of the present application do not limit the type of content, which can be set according to the actual application scene, including but not limited to text content (such as novels, news), image content, video content and audio content.

[0046] The service server 100 obtains the preloading request sent by the terminal device 200a, and obtains the first preloading use probability corresponding to the terminal device 200a according to the preloading request, wherein the first preloading use probability is used to indicate the probability that the target object corresponding to the terminal device 200a interacts with the content provided by the preloading function; the preloading function is used to respond to the preloading request. The first preloading use probability can indicate the real probability that the target object interacts with the historical content provided by the preloading function, or can be used as a predicted probability that the target object interacts with the preloaded content. The determination process of the first preloading use probability is not described here, please refer to the description in the embodiments corresponding to FIG. 3 and FIG. 6. The preloading function can provide a preloaded content list for the terminal device 200a.

[0047] If the first preloading usage probability is less than the online recommendation threshold, the service server 100 acquires a pre-recommendation content pool for the terminal device 200a, where a generation timestamp of the pre-recommendation content pool is earlier than a generation timestamp of the preloading request, i.e., the pre-recommendation content pool is generated in advance, and the generation process of the pre-recommendation content pool is not described here, please refer to the description in the embodiment corresponding to FIG. 9 below. In some embodiments, the service server 100 can generate a target preloading content list through the pre-recommendation content pool for the terminal device. It can be understood that, since the pre-recommendation content pool is generated in advance, the service server 100 can avoid real-time recommendation of online content for the terminal device 200a, and the pre-computed content is used as preloading content, thereby saving the computing resources of online content (because the first preloading usage probability is less than the online recommendation threshold).

[0048] If the first preloading usage probability is equal to or greater than the online recommendation threshold, the service server 100 generates an online recommendation content set for the terminal device, where a generation timestamp of the online recommendation content set is later than a generation timestamp of the preloading request, i.e., the service server recommends online content in real time, and the online content is used as preloading content; and the service server 100 generates a target preloading content list according to the online recommendation content set. It can be understood that, compared with pre-recommendation content (i.e., online content not recommended in real time), online content recommended in real time has the characteristic of high quality, i.e., more matching the browsing characteristics of the target object, so as to improve the interaction probability of the target object and the preloading content, and thereby improve the utilization rate (i.e., output rate) of the preloading content (at this time, the online recommendation content).

[0049] As known above, the embodiments of the present application propose a resource allocation method based on the first preloading usage probability, for the terminal device corresponding to the first preloading usage probability equal to or higher than the online recommendation threshold, an online recommendation content set generated in real time is provided, and thereby a preloading content list generated in real time can be provided; and for the terminal device corresponding to the first preloading usage probability lower than the online recommendation threshold, preloading content is allocated from a pre-recommendation content pool (i.e., a recommendation content pool generated in advance). Therefore, through the pre-recommendation content pool, the resource cost of real-time computation can be reduced; and by comparing the first preloading usage probability and the online recommendation threshold, high-quality preloading content (i.e., recommendation content generated in real time) can be allocated to the target object with high usage probability, so as to not only improve the content recommendation quality, but also improve the usage rate of online recommendation content.

[0050] It should be noted that the above business server 100, terminal device 200a, terminal device 200b, terminal device 200c,..., and terminal device 200n can all be blockchain nodes in a blockchain network, and the data described throughout the text (such as the preloading request and the pre-recommended content pool) can be stored. The storage method can be that the blockchain nodes generate blocks according to the data and add the blocks to the blockchain for storage.

[0051] A blockchain is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm, mainly used for organizing data in chronological order and encrypting it into a ledger, making it tamper-proof and counterfeit-proof, while allowing data verification, storage, and updating. A blockchain is essentially a decentralized database, and each node in the database stores a same blockchain. The blockchain network can distinguish nodes into core nodes, data nodes, and light nodes. The core nodes, data nodes, and light nodes collectively form blockchain nodes.

[0052] The core nodes are responsible for the consensus of the entire blockchain network, that is, the core nodes are consensus nodes in the blockchain network. The process of writing transaction data in the blockchain network into a ledger can be that a data node or a light node in the blockchain network obtains transaction data, transmits the transaction data in the blockchain network (that is, the nodes transmit in the manner of a baton), until a consensus node receives the transaction data. The consensus node then packages the transaction data into a block, performs consensus on the block, and writes the transaction data into the ledger after the consensus is completed. Here, the preloading request and the pre-recommended content pool are used as example transaction data. After the business server 100 (blockchain node) performs consensus on the transaction data, it generates a block according to the transaction data, stores the block in the blockchain network; and for reading of the transaction data (i.e., the preloading request and the pre-recommended content pool), the blockchain node can obtain the block containing the transaction data in the blockchain network, and obtain the transaction data in the block.

[0053] It can be understood that the method provided by the embodiments of the present application can be executed by a computer device, which includes but is not limited to a terminal device or a business server. The business server can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud database, cloud service, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, and big data and artificial intelligence platform. The terminal device includes but is not limited to a mobile phone, a computer, a smart voice interaction device, a smart home appliance, a vehicle-mounted terminal, an aircraft, etc. The terminal device and the business server can be directly or indirectly connected through wired or wireless means, which is not limited in the embodiments of the present application.

[0054] Please refer to FIG. 2, which is a scene diagram of data processing according to an embodiment of the present application. The embodiments of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, intelligent transportation, assisted driving, audio and video, etc. The embodiments of the present application can be applied to business scenarios such as query scenarios, recommendation scenarios, and distribution scenarios of multimedia content, and here the specific business scenarios will not be listed one by one. The implementation process of the data processing scenario can be performed in a business server, or can be performed in a terminal device and a business server in interaction, which is not limited here. In order to facilitate description and understanding, the embodiments of the present application will be described by taking the terminal device and the business server as computer devices, that is, the terminal device and the business server as examples. The terminal device can be any one of the terminal device cluster in the embodiment corresponding to FIG. 1, and the business server can be the business server 100 in the embodiment corresponding to FIG. 1.

[0055] As shown in FIG. 2, the target object 20a has a binding relationship with the terminal device 20b, which can be that the target object 20a is a user using the terminal device 20b. The terminal device 20b is installed with an application client, and in order to facilitate understanding, the application client is exemplified as a multimedia application in FIG. 2. The terminal device 20b can display a non-content output page 201b and a system time (i.e. current time) in the multimedia application. The non-content output page 201b does not include a content detail page (also referred to as a content bottom page), for example, the content is video content, and the non-content output page 201b does not include a video playing page, that is, the non-content output page 201b can include a page in the multimedia application other than the video playing page; for example, the content is text content, and the non-content output page 201b does not include a text detail page, that is, the non-content output page 201b can include a page in the multimedia application other than the text detail page. It should be emphasized that the embodiments of the present application do not limit the type of content, which can include one or more of the following: text content, video content, image content, and audio content.

[0056] For ease of understanding, the content is exemplified as video content, i.e., the content requested by the preloading request is video content. Please refer to Fig. 2 again. When the non-content output page 201b is displayed, i.e., before entering the content (e.g., video, novel, information) playing page, the terminal device 20b generates a preloading request 20c at 2024-1-1 12:00. The preloading request 20c can include a device identifier of the terminal device 20b, such as an Internet Protocol Address (IP address), an application account logged in by a multimedia application, and the like. The type of the device identifier is not limited in the embodiments of the present application, and can be set according to actual application scenarios. For ease of description, the device identifier is exemplified as device 1 in Fig. 2.

[0057] In some embodiments, the terminal device 20b sends the preloading request 20c to the service server 100. After obtaining the preloading request 20c sent by the terminal device 20b, the service server 100 can obtain an index table 20d according to the preloading request 20c. The index table 20d can include an index key of a device identifier and an index value of a preloading use probability (abbreviated as probability in Fig. 2). Fig. 2 exemplifies three index key-value pairs in the index table 20d. The first index key-value pair includes an index key generated by the device 1 and an index value generated by 0.4. The meaning of the first index key-value pair is that the current preloading use probability of the terminal device indicated by the device 1 is 0.4. The second index key-value pair includes an index key generated by the device 2 and an index value generated by 0.3. The meaning of the second index key-value pair is that the current preloading use probability of the terminal device indicated by the device 2 is 0.3. The third index key-value pair includes an index key generated by the device 3 and an index value generated by 0.5. The meaning of the third index key-value pair is that the current preloading use probability of the terminal device indicated by the device 3 is 0.5.

[0058] It should be emphasized that the above description is the current preloading use probability, not the first preloading use probability. The reason is that the preloading use probability is periodically updated, i.e., the service server 100 periodically updates the preloading use probability corresponding to each terminal device. Specifically, the preloading use probability of a terminal device is updated at a first update period (e.g., half an hour or one hour, etc.). Please refer to the descriptions in the embodiments of Figs. 3 and 6.

[0059] As shown in FIG. 2, the service server 100 can search in the index table 20d, determine the first preloading usage probability 20e corresponding to the terminal device 20b as the index value (i.e. 0.4) indicating that the device 1 of the terminal device 20b has an index relationship. The first preloading usage probability 20e is used to indicate the probability of the target object 20a interacting with the content provided by the preloading function. It can be understood that the content provided by the preloading function can include historical content, i.e. the content pulled in the historical period (earlier than January 1, 2024 12:00 in the example of FIG. 2), at this time, the first preloading usage probability 20e can be regarded as the real probability of the target object 20a interacting with the historical content; the content provided by the preloading function can also include the content to be pulled (i.e. the preloading content), at this time, the first preloading usage probability 20e can be regarded as the predicted probability of the target object 20a interacting with the preloading content (to be generated).

[0060] In some embodiments, the service server 100 compares the first preloading usage probability 20e with an online recommendation threshold. If the first preloading usage probability 20e (0.4 in the example of FIG. 2) is less than the online recommendation threshold, for example, the online recommendation threshold is equal to 0.45, the service server 100 can determine that the probability of the terminal device 20b outputting the preloading content is low, i.e. the probability of the target object 20a triggering the preloading content is low. In a popular way, the probability of the target object 20a watching the preloading content is low. At this time, the service server 100 obtains the pre-recommendation content pool 20h corresponding to the terminal device 20b. The pre-recommendation content pool 20h is generated in advance, i.e. the generation timestamp 201h of the pre-recommendation content pool 20h is earlier than the generation timestamp of the preloading request 20c. As shown in the example of FIG. 2, the generation timestamp 201h of the pre-recommendation content pool 20h is January 1, 2024 00:01, and the generation timestamp of the preloading request 20c is January 1, 2024 12:00. The generation process of the pre-recommendation content pool 20h is not described in the embodiments of the present application at present, please refer to the description in the embodiment corresponding to FIG. 9 below.

[0061] FIG. 2 shows that the pre-recommendation content pool 20h includes video 1, video 2, video 3, and video 4. Through the pre-recommendation content pool 20h, the service server 100 can generate a target pre-loading content list 20i, and FIG. 2 shows that the pre-loading content list 20i includes video 1, video 2, and video 3. Embodiments of the present application do not limit the number of pre-loading contents in the pre-loading content list 20i, and the number can be set according to actual application scenarios. In some embodiments, the service server 100 can return the pre-loading content list 20i to the terminal device 20b, and subsequently, if the target object 20a triggers content output, for example, triggers video playback, the terminal device 20b does not need to initiate a content acquisition request to the service server 100, and can output the pre-loading contents in the pre-loading content list 20i in sequence, such as video 1, video 2, and video 3 shown in FIG. 2.

[0062] The above describes a scenario in which the first pre-loading use probability 20e is less than the online recommendation threshold. If the first pre-loading use probability 20e (0.4 shown in FIG. 2) is equal to or greater than the online recommendation threshold, for example, the online recommendation threshold is equal to 0.36, the service server 100 can determine that the probability of the terminal device 20b outputting pre-loading content is high, that is, the probability of the target object 20a triggering pre-loading content is high, in simple terms, the probability of the target object 20a watching pre-loading content is high. At this time, the service server 100 generates an online recommendation content set 20f for the terminal device 20b, wherein, unlike the pre-recommendation content pool 20h, the generation timestamp 201f of the online recommendation content set 20f is later than the generation timestamp of the pre-loading request 20c, for example, the generation timestamp 201f is January 1, 2024, 12:01, that is, the service server 100 generates the online recommendation content set 20f in real time for the pre-loading request 20c sent by the terminal device 20b. Embodiments of the present application do not describe the generation process of the online recommendation content set 20f for the time being, please refer to the description in the embodiment corresponding to FIG. 3 below.

[0063] FIG. 2 shows that the online recommendation content set 20f includes video a, video b, video c, and video d, and according to the online recommendation content set 20f, the service server 100 can generate a target pre-loading content list 20g, as shown in FIG. 2, the target pre-loading content list 20g can include video a, video b, and video c. It can be understood that the target pre-loading content list 20g and the target pre-loading content list 20i correspond to the same function, and therefore the subsequent processing process of the target pre-loading content list 20g is not described here, and can be referred to the description of the target pre-loading content list 20i above.

[0064] It can be known from the above that the embodiment of the application provides the resource allocation method based on the first preloading use probability, provides the online recommendation content set generated in real time for the terminal device corresponding to the first preloading use probability equal to or higher than the online recommendation threshold, and further provides the preloading content list generated in real time; and allocates the preloading content from the pre-recommendation content pool (i.e., the recommendation content pool generated in advance) for the terminal device corresponding to the first preloading use probability lower than the online recommendation threshold. Therefore, the resource cost of real-time calculation can be reduced through the pre-recommendation content pool; and the target object with high use probability can be allocated high-quality preloading content (i.e., the recommendation content generated in real time) by comparing the first preloading use probability and the online recommendation threshold, so that the content recommendation quality can be improved, and the use rate of online recommendation content can be improved.

[0065] Please refer to FIG. 3, which is a flow diagram of a data processing method according to an embodiment of the application. The implementation process of the data processing method can be performed in a business server, or can be performed in interaction between a terminal device and a business server, which is not limited here. For the convenience of description and understanding, the embodiment of the application is described by taking the business server as an example, that is, the business server is taken as a computer device. The business server can be the business server 100 in the embodiment corresponding to FIG. 1. As shown in FIG. 3, the data processing method can at least include the following steps S101-S104.

[0066] In step S101, a preloading request sent by a terminal device is acquired, and a first preloading use probability corresponding to the terminal device is determined according to the preloading request.

[0067] The first preloading use probability is used to indicate the probability of interaction between a target object corresponding to the terminal device and preloading content; the preloading content can be content provided by a preloading function, and the preloading function is a function started in response to the preloading request.

[0068] In some embodiments, the preloading request sent by the terminal device is acquired at a first time; and the first preloading use probability corresponding to the terminal device is determined by determining the number of application start times and the number of preloading use times of the terminal device in a first time period, the latest timestamp of the first time period being earlier than the first time; determining the ratio of the number of application start times to the number of preloading use times, and determining the first preloading use probability according to the ratio. In some embodiments, the first preloading use probability and a device identifier used to indicate the terminal device can be stored in a database in association.

[0069] In some embodiments, a determination is made of an acquisition timestamp of the preloading request, if the acquisition timestamp is in a first type period, an online recommended content set for the terminal device is generated, and according to the online recommended content set, a target preloading content list is generated; if the acquisition timestamp is in a second type period, a first preloading use probability corresponding to the terminal device is acquired; the total number of output contents corresponding to the second type period is greater than the total number of output contents corresponding to the first type period. The first type period can be a content output valley period, for example, the first type period can be from 1 a.m. to 7 a.m. every day.

[0070] The preloading use probability corresponding to the terminal device has timeliness, and the embodiment of the application refers to the preloading use probability corresponding to the terminal device in the current period as a first preloading use probability. The service server sets an update period of the preloading use probability corresponding to the terminal device as a first update period T1, and then updates the preloading use probability corresponding to the terminal device every T1 hours. Assuming that T1 is 0.5, the service server updates the preloading use probability corresponding to the terminal device every half an hour, that is, generates a new preloading use probability. Assuming that the current preloading use probability (which can be understood as the first preloading use probability) of the terminal device is generated by the service server at 12:00 on May 2, 2024, the service server regenerates the preloading use probability of the terminal device at 12:30 on May 2, 2024. The preloading use probability corresponding to the next update period of the current preloading use probability is referred to as a second preloading use probability in the embodiment of the application.

[0071] The data processing method provided by the embodiment of the application can be applied to various multimedia data and product applications including but not limited to short videos, long videos, texts, etc. It can be understood that each application focuses on different data dimensions, for example, application 1 focuses on the duration of the target object browsing the content playback page (for example, the video playback page), application 2 focuses on the number of times of triggering the content playback page by the target object, and application 3 focuses on the content type output by the content playback page, for example, whether the output content is online recommended data or pre-recommended data. Since each application focuses on different data dimensions, the method for determining the preloading use probability corresponding to the terminal device by the service server is not limited in the embodiment of the application, and can be set according to the actual application scenario.

[0072] The embodiment of the present application introduces a simple method for determining preloading use probability. When updating the preloading use probability, the service server counts the application startup times and preloading use times of the terminal device in a first time period. The first time period refers to the time period corresponding to the first update period, for example, if the first update period is half an hour, the first time period corresponds to a time length of half an hour, for example, the service server generates the preloading use probability corresponding to the terminal device at 12 o'clock, and generates the preloading use probability corresponding to the new terminal device at 12:30, and the first time period can include 12 o'clock-12:30. The application startup times of the terminal device refer to the number of times the application client is started, and the preloading use times refer to the number of times the terminal device enters the content output page, that is, the application client is not only started, but also enters the content output page.

[0073] In some embodiments, the service server determines the times ratio between the application startup times and the preloading use times, and can determine the times ratio as the first preloading use probability, or further determines the first preloading use probability in combination with other dimension data of the application client. In combination with the above example, if the generation timestamp of the first preloading use probability is 12:30, the preloading use probability corresponding to the terminal device in the time period from 12:30 to 13:00 is the first preloading use probability, and the service server regenerates the new preloading use probability at 13:00. The service server stores the first preloading use probability and the device identifier used to indicate the terminal device in association in the database, such as the index table 20d shown in the above FIG. 2.

[0074] Please refer to FIG. 4, which is a scene diagram two of data processing provided by the embodiment of the present application. As shown in FIG. 4, the service server 100 acquires the preloading request 20c, which carries the device identifier used to indicate the terminal device. FIG. 4 takes device 1 as an example of the device identifier. The service server 100 determines the acquisition timestamp of the preloading request 20c, that is, the timestamp of acquiring the preloading request 20c. If the acquisition timestamp is in the first type time period 30a, FIG. 4 takes 1:00-9:00 of the day as an example of the first type time period 30a, for example, the acquisition timestamp is 2:00 a.m., and the service server 100 generates the online recommendation content set 20f for the terminal device, wherein the generation process of the online recommendation content set 20f is described in the following step S103, which is not described here.

[0075] According to the online recommendation content set 20f, the service server 100 generates the target preloading content list; it can be understood that the specific process of generating the target preloading content list by the service server according to the online recommendation content set 20f is the same as the specific process of generating the target preloading content list by the service server according to the pre-recommendation content pool, so please refer to the specific generation process of the preloading content list in the following step S102.

[0076] If the acquisition timestamp is in the second type period 30b, the second type period 30b is exemplified as 10:11 of the current day to 1:00 of the next day in FIG. 4, for example, the acquisition timestamp is 8:00 pm, the service server 100 acquires the first preloading use probability 20e corresponding to the terminal device, and the first preloading use probability 20e is exemplified as 0.4 in FIG. 4. Wherein, for the output content of all terminal devices, the total number of output content corresponding to the second type period is greater than the total number of output content corresponding to the first type period, so the above-mentioned first type period 30a can be understood as a content output low peak period, that is, a smaller target object watches content (such as video, novel, etc.) through the terminal device, and the above-mentioned second type period 30b can be understood as a content output peak period, that is, a larger target object watches content through the terminal device.

[0077] It can be understood that, since the total number of output content corresponding to the content output low peak period is small, the system resources of the service server are in an idle state, that is, the utilization rate of the system resources is low, in order to maximize the utilization of the system resources, high-quality online recommended content is provided for the terminal device in the content output low peak period, that is, the online recommended content is generated in real time, and the online recommended content is pushed to the terminal device as preloading content. It can be known from the above that, by adopting the embodiment of the present application, not only can the system resources of the background be reasonably utilized, but also higher-quality preloading content (generated from the online recommended content) can be provided for the target object.

[0078] Correspondingly, since the total number of output content corresponding to the content output peak period is large, that is, a large number of target objects watch content (such as video, information, novel, etc.) through the terminal device, the system resources of the service server are in a busy state, at this time, the utilization rate of the system resources is high, in order to reasonably allocate the system resources, it is necessary to judge whether the target object will play the preloading content, so the service server acquires the first preloading use probability, and makes further judgment based on the first preloading use probability, for details, please refer to steps S102 and S103 below.

[0079] Step 102, judging whether the first preloading use probability is less than the online recommended threshold value.

[0080] The online recommendation threshold refers to a probability threshold for online content recommendation. The online recommendation threshold is determined in association with the current preloading usage probability of the terminal device and in association with the real-time second online quota. The generation process of the online recommendation threshold is not described in detail here. Please refer to the description in the embodiment corresponding to FIG. 7. The service server compares the first preloading usage probability of the terminal device with the online recommendation threshold. If the first preloading usage probability is less than the online recommendation threshold, step 103 is performed. If the first preloading usage probability is equal to or greater than the online recommendation threshold, step 104 is performed.

[0081] In step S103, the target preloading content list is generated by using the pre-recommendation content pool of the terminal device. The generation timestamp of the pre-recommendation content pool is earlier than the generation timestamp of the preloading request.

[0082] In some embodiments, in the pre-recommendation content pool of the terminal device, F pre-recommendation contents in an effective state are determined. F is a positive integer. The recommendation scores corresponding to the F pre-recommendation contents are determined. G recommendation scores are obtained from the F recommendation scores. The G recommendation scores are greater than or equal to H recommendation scores. The H recommendation scores include the recommendation scores other than the G recommendation scores in the F recommendation scores. G is a positive integer less than or equal to F. H is a positive integer less than F. G pre-recommendation contents corresponding to the G recommendation scores are obtained from the F pre-recommendation contents. The target preloading content list is generated according to the G pre-recommendation contents.

[0083] In some embodiments, in the pre-recommendation content pool of the terminal device, F pre-recommendation contents in an effective state are determined. F is a positive integer. The recommendation scores corresponding to the F pre-recommendation contents are determined. G recommendation scores are obtained from the F recommendation scores. The G recommendation scores are greater than or equal to H recommendation scores. The H recommendation scores include the recommendation scores other than the G recommendation scores in the F recommendation scores. G is a positive integer less than or equal to F. H is a positive integer less than F. G pre-recommendation contents corresponding to the G recommendation scores are obtained from the F pre-recommendation contents. The target preloading content list is generated according to the G pre-recommendation contents.

[0084] The first pre-recommended content and the second pre-recommended content are any two of the G pre-recommended contents, that is, the similarity of any two of the G pre-recommended contents needs to be calculated to determine the content similarity between the two pre-recommended contents. For ease of understanding, the preloaded content is exemplified as a video in the embodiments of the present application. Exemplarily, G is 4, which are video 1, video 2, video 3 and video 4 respectively, and then the content similarity between video 1 and video 2, video 2 and video 3, video 3 and video 4, video 1 and video 3, video 1 and video 4, and video 2 and video 4 needs to be calculated.

[0085] In some embodiments, when determining the content similarity between the first pre-recommended content and the second pre-recommended content, multiple modal information such as vision, audio, text, etc. needs to be integrated for determination. In implementation, a first key frame can be determined from the first pre-recommended content, a second key frame can be determined from the second pre-recommended content, image feature vectors of the first key frame and the second key frame can be extracted, and an image similarity of the first key frame and the second key frame can be determined through the two image feature vectors. When there are multiple key frames, all key frame features of the first pre-recommended content and the second pre-recommended content need to be pairwise compared (such as cosine similarity), and the average or maximum value is taken. Then when determining the audio similarity, a first audio feature of the first pre-recommended content and a second audio feature of the second pre-recommended content can be extracted, and the audio similarity can be determined based on the first audio feature and the second audio feature; when determining the text similarity, first text information in the first pre-recommended content can be processed by word embedding to obtain a first text vector, second text information in the second pre-recommended content can be processed by word embedding to obtain a second text vector, and the text similarity of the first pre-recommended content and the second pre-recommended content can be determined through the first text vector and the second text vector. Finally, the image similarity, the audio similarity and the text similarity are weighted to obtain the content similarity of the first pre-recommended content and the second pre-recommended content.

[0086] After obtaining the preloading request, the business server needs to perform a complete recommendation process to determine the online recommended content, which consumes a lot of resources. In addition, since the preloading needs to be initiated before the target object browses the content (such as video, novel, information), but not every content requested by the preloading request will be used by the target object, so a lot of unnecessary requests will be increased, which causes great request and cost pressure on the background service. Based on this, the embodiments of the present application will determine the terminal device initiating the preloading request to determine which type of preloaded content to return. The type of preloaded content can be divided into two categories, the first category is pre-recommended content, which comes from a pre-generated pre-recommended content pool, and the second category is online recommended content, which comes from a real-time generated online recommended content set.

[0087] Please refer to FIG. 4 again, if the first preloading usage probability 20e is less than the online recommendation threshold, for example, the online recommendation threshold is 0.45, the service server 100 acquires the pre-recommendation content pool 20h corresponding to the terminal device, and generates the target preloading content list by targeting the pre-recommendation content pool 20h of the terminal device.

[0088] The process of generating the target preloading content list by targeting the pre-recommendation content pool of the terminal device is described below. For the convenience of understanding, the preloaded content is exemplified as a video in the embodiments of the present application. Please refer to FIG. 5, which is a scene diagram three of data processing provided by the embodiments of the present application. FIG. 5 exemplifies the generation timestamp 30e of the pre-recommendation content pool 30f corresponding to the terminal device as January 1, 2024, 00:01, and the pre-recommendation content pool 30f includes video 1, video 2, video 3, video 4, video 5, video 6, …, and video 7. The service server determines the video in the invalid state, referred to as invalid video, in the pre-recommendation content pool 30f. The reasons why the video is in the invalid state are not limited in the embodiments of the present application, and can be set according to the actual application scene, including but not limited to the played, deleted by the publisher, and blacklisted publisher, etc. as exemplified in FIG. 5.

[0089] As exemplified in FIG. 5, video 1 belongs to invalid video because video 1 has been played. For example, the service server determines that the acquisition timestamp of the preloading request currently received is January 1, 2024, 12:00, and the terminal device has initiated a historical preloading request during January 1, 2024, 00:01- January 1, 2024, 12:00, for example, a preloading request has been sent to the service server at January 1, 2024, 11:00. The preloading request can be referred to as a historical preloading request compared with the system time (i.e. the current time). The service server returns video 1 in the pre-recommendation content pool 30f as the preloading content to the terminal device in response to the historical preloading request, i.e. generates the preloading content list including video 1 and returns it to the terminal device. The terminal device acquires the preloading content list including video 1, and plays video 1 when the target object triggers the content playing, so video 1 belongs to the played video, and the played video belongs to the invalid video.

[0090] As exemplified in FIG. 5, the publisher corresponding to video 5 deletes video 5, so the service platform does not have the playing right for video 5, and video 5 belongs to invalid video. The target object blacklists the publisher (for example, publisher 300) of video 6, i.e. sets not to watch the video published by the publisher 300, so video 6 belongs to invalid video.

[0091] FIG. 5 illustrates three types of invalid videos, and other invalid types can be included in actual applications. In addition, FIG. 5 illustrates the preloaded content as a video, and if the preloaded content is audio or text, the invalid types can include, but are not limited to, output, deleted, blacklisted, and the like.

[0092] The service server deletes the videos in the invalid state in the pre-recommendation content pool 30f (i.e., the pre-recommendation content), and obtains the pre-recommendation content in the valid state, as illustrated in the pre-recommendation content 30g in FIG. 5, including video 2, video 3, video 4, and video 7, i.e., F is illustrated as 4 in FIG. 5. In some embodiments, the service server obtains the recommendation score corresponding to each pre-recommendation content, and the recommendation score of video 2 is illustrated as 0.87, the recommendation score of video 3 is illustrated as 0.85, the recommendation score of video 4 is illustrated as 0.83, and the recommendation score of video 7 is illustrated as 0.8 in FIG. 5. For ease of understanding, G is illustrated as 3 in FIG. 5, and the service server obtains the pre-recommendation content with the top 3 recommendation scores in the pre-recommendation content 30g, i.e., video 2, video 3, and video 4, in the descending order of the obtained recommendation scores.

[0093] In some embodiments, the service server can generate a first preloaded content list according to video 2, video 3, and video 4, and obtain the content similarity between each two pre-recommendation contents in the first preloaded content list, as illustrated in FIG. 5, the similarity between video 2 and video 3 is 0.7, the similarity between video 2 and video 4 is 0.3, and the similarity between video 3 and video 4 is 0.2; wherein video 2, video 3, or video 4 can be the first pre-recommendation content, and the second pre-recommendation content can be one of the remaining two videos. The service server compares each content similarity with the similarity threshold, and if the three similarities illustrated in FIG. 5 are all less than the similarity threshold, for example, the similarity threshold is 0.8, the service server can determine the first preloaded content list as the preloaded content list 30i, i.e., including video 2, video 3, and video 4, and return the preloaded content list 30i to the terminal device.

[0094] If the content similarity is greater than or equal to the similarity threshold, for example, the similarity threshold is 0.5, the video 2 in the example of FIG. 5 and the similarity between the video 2 is 0.7 which is greater than 0.5, the service server deletes the video with a smaller recommendation score in the two videos, that is, deletes the video 3 in the first preloaded content list, and obtains a second preloaded content list. The second preloaded content list after deleting the video 3 includes two pre-recommended contents, that is, the video 2 and the video 4, so the service server obtains the pre-recommended content with the largest recommendation score from the pre-recommended content 30g except the second pre-recommended content list, which is the video 7 in the example of FIG. 5, and adds the video 7 to the second preloaded content list after deleting the video 3, and the service server can obtain a third preloaded content list 30h.

[0095] It can be understood that the service server processes the third preloaded content list 30h in the same way as the candidate preloaded content list including the video 2, the video 3 and the video 4, and thus details are not described here. If the similarity between two pre-recommended contents in the third preloaded content list 30h is equal to or greater than the similarity threshold, the service server continues to obtain the pre-recommended content from the pre-recommended content 30g. If the pre-recommended content 30g does not include the pre-recommended content except the third preloaded content list 30h, the service server can determine the third preloaded content list 30h as the target preloaded content list.

[0096] A feasible implementation manner is as follows: when the preloading request sent by the terminal device is obtained, if the service server determines that the preloading use probability corresponding to the terminal device is less than the online recommendation threshold, and the total number of pre-recommended contents contained in the pre-recommended content pool corresponding to the terminal device is less than G, the service server can generate an online recommendation content set for the terminal device in real time, and then generate a target preloaded content list based on the online recommendation content set.

[0097] The preloaded content list of the embodiment of the application is called a feed stream, that is, a feed stream. The feed stream is an information stream that is continuously updated and presented to a target object content, and through personalized recommendation, the target object can efficiently obtain the demand for specific content.

[0098] In step S104, an online recommendation content set for the terminal device is generated, and a target preloaded content list is generated according to the online recommendation content set. The generation time stamp of the online recommendation content set is later than the generation time stamp of the preloading request.

[0099] In some embodiments, a total number of preloading usage probabilities equal to or greater than the online recommendation threshold is determined; if the total number is less than or equal to the first online recommendation quota, an online recommendation content set for the terminal device is generated according to the output content features of the terminal device and the candidate content pool corresponding to the system time; if the total number is greater than the first online recommendation quota, a target preloading content list is generated through the pre-recommendation content pool.

[0100] When the first preloading usage probability is equal to or greater than the online recommendation threshold, the service server determines a first online recommendation quota corresponding to the current time, which is a real-time changing parameter determined according to a real-time changing second online recommendation quota and an online recommendation total quota with fixed properties, and the relationship among the three can be represented by the following formula (1):

[0101] Online recommendation total quota≥second online recommendation quota+first online recommendation quota (1)

[0102] The online recommendation total quota is pre-set and fixed, determined according to the system resources of the service server, the second online recommendation quota refers to the total number of non-preloading online recommendation contents, the non-preloading online recommendation contents refer to the contents pulled by the terminal device after entering the content playing page, i.e., the second total number of contents pulled by the terminal device after entering the content playing page, and the preloading online recommendation contents refer to the contents pulled by the terminal device before entering the content playing page, i.e., the first total number of contents pulled by the terminal device before entering the content playing page. The embodiment of the present application sets to determine the real-time changing first online recommendation quota according to the real-time changing second online recommendation quota and the fixed online recommendation total quota, which can specifically collect the real second online recommendation quota of the historical period by the service server as the second online recommendation quota of the current time, for example, the real second online recommendation quota of the last second is determined as the second online recommendation quota of the current time, so that the first online recommendation quota of the current time can be obtained by performing difference processing on the online recommendation total quota and the second online recommendation quota of the last second.

[0103] The service server counts the total number of preloading usage probabilities equal to or greater than the online recommendation threshold, and FIG. 5 illustrates the total number 30c as 1.5 million, i.e., there are more than or equal to 1.5 million terminal devices that initiate preloading requests to the service server in the current period (e.g., 1 second, half a minute, etc.), and there are 150 terminal devices corresponding to preloading usage probabilities (including the first preloading usage probability described above) equal to or greater than the online recommendation threshold.

[0104] The business server compares the total number 30c with the first online recommendation quota, which mainly ensures that the business server does not overload and maximizes the use of system resources. Therefore, if the total number 30c is less than or equal to the first online recommendation quota, for example, the first online recommendation quota is 1.51 million, the business server obtains the candidate content pool 30d corresponding to the output content features of the terminal device and the system time. Among them, the output content features of the terminal device refer to the features of the content (including preloaded content and non-preloaded content) output by the terminal device in the historical period. One feasible way to obtain the output content features is as follows: the business server inputs the content output by the terminal device in the historical period, for example, the videos output in a week, into the video recognition model, and obtains the features corresponding to the videos output in a week by recognizing and integrating the videos output in a week through the video recognition model. For example, the videos output by the terminal device in a week have the features of being funny and cute. Among them, the video recognition model is a trained AI model, and the model type and structure of the video recognition model are not limited in the embodiments of the present application, and can be set according to the needs of the actual application scene.

[0105] One way to generate the candidate content pool 30d is as follows: generating the candidate content that has output permission at the current time of the business server to generate the candidate content pool 30d. In order to narrow the scope, another way to generate the candidate content pool 30d is as follows: generating the candidate content associated with the terminal device to generate the candidate content pool 30d, for example, the terminal device is bound with an application account, the content published by the friend account having a friend relationship with the application account, the content output by the friend account, and the content interacted (such as comments, likes, rewards, and attention) by the friend account can be added to the candidate content pool 30d, in addition, the content published by the content publisher interacted by the application client installed in the terminal device can be added to the candidate content pool 30d.

[0106] The service server identifies the features of the candidate content in the candidate content pool 30d. As shown in FIG. 4, the candidate content pool 30d includes video 1, video 2, video 3, video 4, video 5, …, video a, video b, video c, video d, and video e. The service server can identify video 1 by using the video recognition model, obtain the video features of video 1, and in some embodiments, identify the feature similarity 1 between the video features of video 1 and the output content features. If the feature similarity 1 is greater than or equal to the similarity threshold z (to distinguish from the similarity threshold described above, the similarity threshold is referred to as the similarity threshold z), the service server can determine video 1 as the online recommendation content. If the feature similarity 1 is less than the similarity threshold z, video 1 is filtered out, and the next video (for example, video 2) is processed as video 1. Until all candidate content in the candidate content pool 30d is identified, all online recommendation content is determined as the online recommendation content set. As shown in FIG. 4, the online recommendation content set 20f includes video a, video b, video c, …, and video d.

[0107] As shown in FIG. 5, each online recommendation content in the online recommendation content set corresponds to a recommendation score. The service server can determine the target preloaded content list in the online recommendation content set according to the recommendation score of each online recommendation content. It can be understood that the process of generating the target preloaded content list by the service server according to the online recommendation content set is the same as the process of generating the target preloaded content list by the service server according to the pre-recommendation content pool. Therefore, the description is not repeated here. Please refer to the description in step S102 above.

[0108] In another scenario, please refer to FIG. 4 again. If the total number 30c is greater than the first online recommendation quota, for example, the first online recommendation quota is 1.4 million, the service server generates the target preloaded content list by using the pre-recommendation content pool 20h.

[0109] As described above, the embodiments of the present application propose a resource allocation method based on the first preloaded use probability. For the terminal device corresponding to the first preloaded use probability equal to or higher than the online recommendation threshold, the online recommendation content set generated in real time is provided, and then the preloaded content list generated in real time can be provided. For the terminal device corresponding to the first preloaded use probability lower than the online recommendation threshold, the preloaded content is allocated from the pre-recommendation content pool (i.e., the pre-generated recommendation content pool). Therefore, by using the pre-recommendation content pool, the resource cost of real-time calculation can be reduced. By comparing the first preloaded use probability and the online recommendation threshold, the target object with high use probability can be allocated high-quality preloaded content (i.e., real-time generated recommendation content). Therefore, not only the content recommendation quality can be improved, but also the use rate of online recommendation content can be improved.

[0110] Please refer to FIG. 6, which is a flow diagram of a data processing method according to an embodiment of the present application. The data processing method can be executed by a service server (for example, the service server 100 shown in FIG. 1) or can be executed by interaction between a service server and a terminal device (for example, the terminal device 200a shown in FIG. 1). For ease of understanding, the embodiment of the present application takes the method executed by the service server as an example, that is, the service server as a computer device. As shown in FIG. 6, the data processing method can at least include the following steps S201-S207.

[0111] In step S201, a preloading request sent by a terminal device is acquired, and according to the preloading request, a first preloading use probability corresponding to the terminal device is acquired. The first preloading use probability is used to indicate a probability of a target object corresponding to the terminal device interacting with preloading content.

[0112] In some embodiments, in the database, B pieces of historical preloading content lists generated for the terminal device in a first time period are acquired; B is a positive integer; the first time period is earlier than the second time period, and the first time period is adjacent to the second time period; in the database, historical list scores corresponding to the B pieces of historical preloading content lists are acquired, and the B historical list scores are summed to obtain a first score; among the B pieces of historical preloading content lists, C pieces of historical preloading content lists that have been output by the terminal device are determined; C is a natural number less than or equal to B; according to the first score and the C pieces of historical preloading content lists, the first preloading use probability is determined.

[0113] In some embodiments, in the database, B pieces of historical preloading content lists generated for the terminal device in a first time period are acquired; B is a positive integer; the first time period is earlier than the second time period, and the first time period is adjacent to the second time period; in the database, historical list scores corresponding to the B pieces of historical preloading content lists are acquired, and the B historical list scores are summed to obtain a first score; among the B pieces of historical preloading content lists, C pieces of historical preloading content lists that have been output by the terminal device are determined; C is a natural number less than or equal to B; according to the first score and the C pieces of historical preloading content lists, the first preloading use probability is determined.

[0114] As described in step S101 of the embodiment corresponding to FIG. 3, if the terminal device sends a preloading request to the service server in a second type time period, that is, a content output peak time period, the service server acquires the first preloading use probability corresponding to the terminal device. The method proposed in the embodiment of the present application can be applied to multiple item and product applications including short videos, long videos, and graphic texts. Since different product applications focus on different dimension data, the determination method of the first preloading use probability is not limited in the embodiment of the present application, and can be set according to the actual application scenario.

[0115] The embodiment of the application describes a feasible method for determining preloading use probability. When updating the preloading use probability corresponding to the terminal device, the service server obtains B historical preloading content lists generated for the terminal device in a first time period in the database. As described in step S101 of the embodiment corresponding to FIG. 3, the preloading use probability corresponding to the terminal device has timeliness, that is, the service server updates the preloading use probability of the terminal device according to a first update period. The first time period can be understood as a time period corresponding to the first update period. For example, the first update period is half an hour, so the service server updates the preloading use probability of the terminal device every half an hour. For example, the service server generates the preloading use probability corresponding to the terminal device at 9:30. At 10:00, the service server obtains B historical preloading content lists generated for the terminal device in the first time period (9:30-10:00), that is, the terminal device sends B preloading requests to the service server in the first time period.

[0116] In some embodiments, the service server obtains historical list scores corresponding to the B historical preloading content lists from the database, sums the B historical list scores to obtain a first score, determines C historical preloading content lists output by the terminal device from the B historical preloading content lists, wherein the historical preloading content list output by the terminal device can also be understood as a historical preloading content list with an output attribute, that is, a historical preloading content list that interacts with a target object, that is, after a historical preloading content list is returned to the terminal device, the target object corresponding to the terminal device triggers the output of the preloading content in the historical preloading content list, for example, the target object watches the preloading content in the historical preloading content list through the terminal device.

[0117] The service server sums historical list scores corresponding to the C historical preloading content lists to obtain a second score, determines a score ratio between the first score and the second score, and the service server can determine the score ratio as the current preloading use probability corresponding to the terminal device, that is, the first preloading use probability; or further combines other dimension data of the application client, for example, the times ratio described in step S101 of the embodiment corresponding to FIG. 3, for example, performs weighted sum processing on the times ratio and the score ratio to determine the first preloading use probability, wherein the sum of the weights corresponding to the times ratio and the score ratio is equal to 1.

[0118] In step S202, if the first preloading use probability is less than an online recommendation threshold, a target preloading content list is generated through a pre-recommendation content pool of the terminal device. The generation timestamp of the pre-recommendation content pool is earlier than the generation timestamp of the preloading request.

[0119] Step S203, if the first preloading usage probability is equal to or greater than the online recommendation threshold, generating an online recommendation content set for the terminal device, and generating a target preloading content list according to the online recommendation content set; the generation time stamp of the online recommendation content set is later than the generation time stamp of the preloading request.

[0120] In some embodiments, in combination with steps S202-S203, in an application scenario (for example, short video browsing), the embodiments of the present application can effectively reduce the time consumption of waiting for playing when the terminal device enters the short video stream through preloading content. Since the content preloaded to the client may not be used by the target object, that is, the target object may not trigger the content to play the content, if the preloaded content is not used, the resources consumed by this preloading process are wasted. Based on this, the embodiments of the present application provide real-time recommended resources (that is, online recommendation content) for target objects with high preloading usage probability, and provide pre-recommended resources (that is, pre-recommended content) for target objects with low usage probability. The above resource allocation method can guarantee the experience of the target object and the recommendation quality.

[0121] As can be seen from the above, by constructing a pre-recommended content pool, high-quality content with low cost can be provided, and system availability can be effectively guaranteed; based on efficient real-time recommendation and pre-recommended resource allocation, resource utilization maximization can be achieved.

[0122] Step S204, determining the content types corresponding to the A preloading contents in the preloading content list; A is a positive integer.

[0123] In some embodiments, the embodiments of the present application do not limit the application scenario and do not limit the content type, so the content type is not limited and can be determined according to the actual application scenario. The content type can include but is not limited to the following: a type similar to the output content feature, a friend publishing type, a friend interaction type, and an account interaction type. The preloading content belonging to the type similar to the output content feature refers to the similarity between the preloading content and the output content feature being greater than a similarity threshold z. The preloading content belonging to the friend publishing type refers to the preloading content belonging to the content published by the friend account. The preloading content belonging to the friend interaction type refers to the preloading content belonging to the content interacted (for example, liked, commented, forwarded, rewarded, etc.) by the friend account. The preloading content belonging to the account interaction type refers to the preloading content belonging to the content interacted by the application account logged in the application client.

[0124] Step S205, for each preloading content in the A preloading contents, determining a content score corresponding to the preloading content according to the content type corresponding to the preloading content.

[0125] In some embodiments, the business server can set a content score for each content type, in combination with the content types shown in step S204, for example, setting the content score of the type of content similar to the output content feature to 2, setting the content score of the friend publishing type to 3, setting the content score of the friend interaction type to 1, and setting the content score of the account interaction type to 2.

[0126] The business server can determine the content score of each preloaded content according to the content type to which each preloaded content belongs, for example, the target preloaded content list 20i shown in the example of FIG. 2 includes video 1, video 2, and video 3, wherein video 1 belongs to the type similar to the output content feature, so the content score of video 1 is 2; video 2 belongs to the friend publishing type, so the content score of video 2 is 3, and video 3 belongs to the friend publishing type, so the content score of video 3 is 3.

[0127] Step S206: Summing up the content scores corresponding to the A preloaded contents respectively to obtain a list score corresponding to the preloaded content list; the list score is used to determine a second preloading usage probability corresponding to a second time period.

[0128] In some embodiments, in combination with step S205, the business server determines the content scores corresponding to the three videos in the target preloaded content list 20i shown in FIG. 2, i.e., 2, 3, and 3, and sums up the three content scores to obtain a list score corresponding to the target preloaded content list 20i, i.e., 2+3+3=8. It can be understood that the determination method of the historical list score corresponding to the historical preloaded content list is the same as the determination method of the list score corresponding to the target preloaded content list.

[0129] Step S207: Storing the list score and the device identifier used to indicate the terminal device in association in a database.

[0130] In some embodiments, the beneficial effects brought by the embodiments of the present application can be summarized as follows:

[0131] 1. Using very low cost to improve the experience and availability of the feed stream (such as short video) and maintain the high quality of recommended content.

[0132] 2. The generation method and application of the pre-recommended content pool are proposed, which processes a large number of preloading requests at low cost and guarantees the high-quality experience of the feed stream.

[0133] 3. The resource allocation method based on preloading usage probability is proposed, which enables the target object with high usage probability to obtain higher quality resources, and further improves the recommendation quality under the condition of unchanged cost.

[0134] 4. A method of using a pre-recommended content pool to back up in real time when recommending faults is proposed, which guarantees the availability of the system.

[0135] As known above, the embodiment of the application proposes a resource allocation method based on a first preloaded usage probability. For a terminal device corresponding to a first preloaded usage probability equal to or higher than an online recommendation threshold, an online recommended content set generated in real time is provided, and a preloaded content list generated in real time can be further provided. For a terminal device corresponding to a first preloaded usage probability lower than the online recommendation threshold, a preloaded content is allocated from a pre-recommended content pool (i.e., a recommended content pool generated in advance). Therefore, by using the pre-recommended content pool, the resource cost of real-time calculation can be reduced. By comparing the first preloaded usage probability and the online recommendation threshold, high-quality preloaded content (i.e., recommended content generated in real time) can be allocated to a target object with a high usage probability, so that the content recommendation quality can be improved, and the usage rate of online recommended content can be improved.

[0136] Please refer to FIG. 7, which is a flowchart of a data processing method according to an embodiment of the application. The method can be executed by a business server (for example, the business server 100 shown in FIG. 1), or can be executed by interaction between a business server and a terminal device (for example, the terminal device 200a shown in FIG. 1). For ease of understanding, the embodiment of the application takes the method executed by the business server as an example, that is, the business server as a computer device. As shown in FIG. 7, the method can at least include steps S301-S306.

[0137] In step S301, D preloaded usage probabilities of all terminal devices in a first time period are obtained from a database; all terminal devices include terminal devices; and the D preloaded usage probabilities include a first preloaded usage probability.

[0138] In some embodiments, all terminal devices refer to all terminal devices that consume content in the first time period. The first time period can be understood as a time period corresponding to a first update period. The business server updates the preloaded usage probability corresponding to each terminal device in all terminal devices according to the first update period. For example, if the first update period is half an hour, the business server updates the preloaded usage probability corresponding to each terminal device in all terminal devices every half an hour. For example, if the current preloaded usage probability of the terminal device is generated at 9:30, the business server updates the preloaded usage probability of the terminal device at 10:00, so the first time period mentioned above can be from 9:30 to 10:00.

[0139] It can be understood that the total number of all terminal devices is equal to D.

[0140] In step S302, the D preloaded usage probabilities are sorted to obtain D preloaded usage probabilities after sorting.

[0141] In some embodiments, the business server can sort the D preloaded use probabilities in descending order, and obtain the sorted D preloaded use probabilities, i.e., the preloaded use probability in the front sequence is greater than or equal to the preloaded use probability in the rear sequence.

[0142] In step S303, the first online recommendation quota corresponding to the system time is determined, and the online recommendation threshold is determined from the sorted D preloaded use probabilities according to the first online recommendation quota.

[0143] In some embodiments, the second online recommendation quota corresponding to the system time is determined, and the first online recommendation quota is obtained by subtracting the second online recommendation quota from the preset total online recommendation quota. E preloaded use probabilities are obtained from the sorted D preloaded use probabilities, and the number of the E preloaded use probabilities is equal to the first online recommendation quota. E is a positive integer less than or equal to D. The minimum preloaded use probability of the E preloaded use probabilities is determined as the online recommendation threshold.

[0144] The embodiments of the present application realize dynamic allocation of real-time recommended resources (i.e., online recommended content), and maximize the use of real-time recommended resources. In some embodiments, a total real-time recommendation quota, i.e., a fixed total online recommendation quota, is determined, and the business server determines the second online recommendation quota according to a second update period, for example, one second or one minute. That is, the business server determines the total number of non-preloaded requests obtained in the current second (or the current minute) as the second online recommendation quota of the next second (or the next minute). The non-preloaded request is generated when the terminal device enters the content output page and there is a content output request. The content returned by the business server for the non-preloaded request is called non-preloaded content, which is online recommended content, and the embodiments of the present application are also called non-preloaded online recommended content. The second online recommendation quota refers to the total number of non-preloaded online recommended content, and the non-preloaded online recommended content refers to the content pulled after the terminal device enters the content playing page. The first online recommended content refers to the content pulled before the terminal device enters the content playing page.

[0145] The second online recommendation quota and the first online recommendation quota need to satisfy the above formula (1). Please refer to FIG. 8, which is an example diagram of the accumulation broken line of the quota of the first online recommended content provided by the embodiment of the present application. The sum of the second online recommendation quota and the first online recommendation quota is less than or equal to the online recommendation total quota. Specifically, in the content output low peak period (also referred to as the first type period in the embodiment of the present application), because the sum of the number of preloading requests and the number of non-preloading requests is small, the sum of the second online recommendation quota and the first online recommendation quota can be less than the online recommendation total quota. Therefore, the system resource utilization rate of the service server is low. Therefore, in the content output low peak period, the service server returns the online recommended content as the preloading content to the terminal device in response to the preloading request, that is, the pre-recommended content is not taken (the preloading pre-recommendation is represented in FIG. 8).

[0146] In the content output peak period (also referred to as the second type period in the embodiment of the present application), the number of online target objects is large, and the sum of the number of preloading requests and the number of non-preloading requests is greater than the online recommendation total quota. Therefore, the service server sets the order of magnitude of the real-time recommendation of the preloading request to be dynamically adjusted with the order of magnitude of the real-time recommendation of the non-preloading request. Therefore, the real-time recommendation resource can be maximally utilized. As shown in FIG. 8, in the content output peak period, the sum of the first online recommendation quota and the second online recommendation quota is equal to the fixed online recommendation total quota.

[0147] In step S304, the preloading request sent by the terminal device is acquired, and the first preloading use probability corresponding to the terminal device is acquired according to the preloading request. The first preloading use probability is used to indicate the probability of the target object corresponding to the terminal device interacting with the preloading content.

[0148] In step S305, if the first preloading use probability is less than the online recommendation threshold, a target preloading content list is generated by the pre-recommended content pool of the terminal device. The generation timestamp of the pre-recommended content pool is earlier than the generation timestamp of the preloading request.

[0149] In step S306, if the first preloading use probability is equal to or greater than the online recommendation threshold, an online recommended content set for the terminal device is generated, and a target preloading content list is generated according to the online recommended content set. The generation timestamp of the online recommended content set is later than the generation timestamp of the preloading request.

[0150] The specific implementation process of steps S304-S306 can be referred to steps S101-S10 in the embodiment corresponding to FIG. 3.

[0151] It can be known from the above that the embodiment of the application proposes a resource allocation method based on a first preloading use probability. For a terminal device corresponding to a first preloading use probability equal to or higher than an online recommendation threshold, a set of online recommendation content generated in real time is provided, and then a preloading content list generated in real time can be provided. For a terminal device corresponding to a first preloading use probability lower than the online recommendation threshold, preloading content is allocated from a pre-recommendation content pool (i.e., a recommendation content pool generated in advance). Therefore, by using the pre-recommendation content pool, the resource cost of real-time calculation can be reduced. By comparing the first preloading use probability and the online recommendation threshold, high-quality preloading content (i.e., recommendation content generated in real time) can be allocated to a target object with a high use probability, so that the content recommendation quality can be improved, and the use rate of online recommendation content can be improved. In addition, the embodiment of the application proposes a method of jointly dynamically allocating an online recommendation total quota by a preloading request and a non-preloading request, so that the utilization rate of online recommendation resources can be improved.

[0152] Please refer to FIG. 9, which is a flow diagram of a data processing method provided by an embodiment of the application. The method can be executed by a business server (for example, the business server 100 shown in FIG. 1), or can be executed by interaction between a business server and a terminal device (for example, the terminal device 200a shown in FIG. 1). For ease of understanding, the embodiment of the application takes the method executed by the business server as an example for description, that is, the business server is taken as a computer device. As shown in FIG. 9, the method can at least include the following steps.

[0153] In step S401, a set of historical online recommendation content generated for a terminal device is acquired, and a first historical online recommendation content with an output attribute is determined in the set of historical online recommendation content.

[0154] In some embodiments, the first historical online recommendation content with the output attribute refers to the first historical online recommendation content output by the terminal device. In order to achieve a high-quality experience of the feed stream, various multimedia applications will perform preloading before a target object consumes multimedia content, including preloading a recommendation feed list (also referred to as a preloading content list in the application), pre-downloading content headers, pre-rendering players, and the like. Loading the content before the target object watches the content can make the application client not appear to be stuck or slow when the target object consumes the content, thereby improving the consumption experience of the target object.

[0155] The embodiment of the present application proposes a method for improving the experience, availability of feed fluid, and maintaining high recommendation quality at low cost. The historical online recommendation content set refers to the online recommendation content set generated in response to the preloading request (hereinafter referred to as historical preloading request) obtained in the historical period. The total number of the historical online recommendation content set is not limited in the embodiment of the present application, and can be set according to the actual application scenario. An optional implementation is as follows: the historical online recommendation content set can include the online recommendation content set generated in response to the preloading request obtained in the historical period, and can also include the online recommendation content set generated in response to the non-preloading request obtained in the historical period.

[0156] For example, the current system time is May 6, 2024, 12:00, and the online recommendation content set generated for the terminal device within one week (or one day or three days) before May 6, 2024, 12:00 is referred to as the historical online recommendation content set.

[0157] The business server determines the first historical online recommendation content with the output attribute in the historical online recommendation content set, wherein the first historical online recommendation content with the output attribute refers to the online recommendation content that has been consumed by the target object, for example, the online recommendation content is returned to the terminal device as preloading content (which can also be non-preloading content), and the target object consumes the online recommendation content in the terminal device, so the online recommendation content has the output attribute.

[0158] In step S402, the second historical online recommendation content is added to the initial pre-recommendation content pool; the second historical online recommendation content is the recommendation content in the historical online recommendation content set except the first historical online recommendation content.

[0159] In some embodiments, the second historical online recommendation content refers to the online recommendation content that is not consumed by the target object, and has the to-be-output attribute, corresponding to the first historical online recommendation content.

[0160] In step S403, the pre-recommendation content pool is generated according to the initial pre-recommendation content pool.

[0161] In some embodiments, in the first type period, the offline recommendation processing is performed for the terminal device, and the offline recommendation content is obtained; the offline recommendation content is added to the initial pre-recommendation content pool, and the initial pre-recommendation content pool to which the offline recommendation content is added is determined as the pre-recommendation content pool.

[0162] In the present application, the pre-recommended content is constructed by two parts of content: online computing content triggered by pre-loading request, and offline pre-computed content. The online computing content includes the set of online recommended content generated in response to the pre-loading request, and can also include the set of online recommended content generated in response to the non-pre-loading request; the offline pre-computed content refers to the offline recommended content generated by the business server through content output low peak period for the terminal device.

[0163] As can be seen from steps S401-S403, the present application embodiment saves the feed list produced by the pulling process (which can include the pre-loading process and the non-pre-loading process) as the initial pre-recommended content pool, and supplements the initial pre-recommended content pool with the feed produced by offline computation during the content output low peak period, so as to ensure that abundant and high-quality personalized pre-recommended content is obtained at low cost.

[0164] As can be seen from the above, the present application embodiment implements a pre-loading mechanism, constructs a pre-recommended content pool, and according to the probability of using the pre-loading request, allocates real-time recommended content or pre-recommended content, so that the target object can maximize the consumption of real-time recommended content, and the timeliness and content recommendation quality of the content recommended to the target object can be guaranteed, thereby improving the attractiveness of the recommended content to the target object, promoting the target object to participate in content creation and interaction, and improving the satisfaction of the target object to the recommended platform and user stickiness.

[0165] Please also refer to FIG. 10, which is a classification example diagram of loading content provided by the present application embodiment. The loading content can be divided into pre-loading content and non-pre-loading content, wherein the pre-loading content refers to the content pulled in advance by the terminal device before entering the content consumption page, and the non-pre-loading content refers to the content pulled in real time by the terminal device when the content consumption page is entered and the content is being consumed, so the non-pre-loading content is all online recommended content. The pre-loading content can be divided into two categories: one is pre-recommended content, and the other is online recommended content. When the pre-recommended content is used as pre-loading content, and when the online recommended content is used as pre-loading content, please refer to the description of the embodiment corresponding to FIG. 3. The pre-recommended content can also be divided into two categories: one is historical online recommended content, and the other is offline recommended content. The meanings and generation of the above two types of content are described in the steps above.

[0166] In step S404, the pre-loading request sent by the terminal device is obtained, and according to the pre-loading request, the first pre-loading use probability corresponding to the terminal device is obtained; the first pre-loading use probability is used to indicate the probability of interaction between the target object corresponding to the terminal device and the pre-loading content.

[0167] The specific implementation process of step S404 can be found in step S101 in the embodiment corresponding to FIG. 3.

[0168] In step S405, if the first preloading use probability is less than the online recommendation threshold, a target preloading content list is generated by using a pre-recommendation content pool for the terminal device; and the generation time stamp of the pre-recommendation content pool is earlier than the generation time stamp of the preloading request.

[0169] In some embodiments, the pre-recommendation content pool of the high-quality feed is used for preloading in the front-end scene, which can improve the preloading coverage and reduce the browsing waiting time of the target object. According to the probability of using preloading data (i.e., the preloading use probability) of the target object, the online recommendation resource (i.e., the online recommendation content) is allocated to the target object with a high use probability, and the pre-recommendation content is allocated to the target object with a low use probability, so as to maximize the use of system resources and maximize the business indicators. In addition, when the recommendation system fails, the pre-recommendation content pool is used to replace the real-time recommendation content, which fully guarantees the availability of the feed stream and ensures the high recommendation quality.

[0170] In step S406, if the first preloading use probability is equal to or greater than the online recommendation threshold, an online recommendation content set for the terminal device is generated, and a target preloading content list is generated according to the online recommendation content set; and the generation time stamp of the online recommendation content set is later than the generation time stamp of the preloading request.

[0171] In some embodiments, please refer to FIG. 11 and FIG. 12 together. FIG. 11 is a use probability diagram of randomly allocating online recommendation content according to an embodiment of the present application, and FIG. 12 is a use probability diagram of allocating online recommendation content according to probability according to an embodiment of the present application. As shown in FIG. 11, for the preloading request, the online recommendation content is used with a probability of 30%, and the pre-recommendation content is used with a probability of 70%. For example, 100 preloading requests are obtained, which are respectively sent by 100 terminal devices. The business server randomly selects 30 preloading requests from the 100 preloading requests, and returns the preloading content generated by the online recommendation content for the 30 preloading requests, and returns the preloading content generated by the pre-recommendation content for the 70 preloading requests. Since the preloading requests are randomly allocated online recommendation content, in FIG. 11, only 14% of the preloading content generated by the online recommendation content is used (i.e., consumed), and 86% of the preloading content generated by the online recommendation content is not used.

[0172] Please refer to Fig. 12 again, according to the embodiment of the present application, for the preloading request, 30% probability is still used to adopt the online recommended content, 70% probability is used to adopt the pre-recommended content, but the preloading use probability corresponding to the terminal device respectively sending the preloading request is acquired, and these preloading use probabilities are sorted. For example, 100 terminal devices respectively send preloading requests, that is, 100 preloading requests are acquired, the service server acquires the preloading use probability corresponding to the 100 terminal devices respectively, and sorts the 100 preloading use probabilities from large to small. For the preloading use probabilities ranked in the front 30, the preloading content generated by the online recommended content is responded to the terminal devices corresponding to these preloading use probabilities, and for the preloading use probabilities ranked in the back 70, the preloading content generated by the pre-recommended content is responded to the terminal devices corresponding to these preloading use probabilities.

[0173] Since the online recommended content is distributed according to the preloading use probability, in Fig. 12, 25% of the preloading content generated by the online recommended content is used, that is, consumed, and 75% of the preloading content generated by the online recommended content is not used. Compared with Fig. 11, only 14% of the preloading content generated by the online recommended content is used, Fig. 12 improves by 11%, that is, the embodiment of the present application can make the computing resources, storage resources and network bandwidth be more reasonably utilized, thereby avoiding resource waste and improving resource utilization and content usage.

[0174] As known from the above, the embodiment of the present application proposes a resource allocation method based on the first preloading use probability, for the terminal device corresponding to the first preloading use probability equal to or higher than the online recommended threshold, an online recommended content set generated in real time is provided, and then a preloading content list generated in real time can be provided; for the terminal device corresponding to the first preloading use probability lower than the online recommended threshold, preloading content is allocated from the pre-recommended content pool (that is, the recommended content pool generated in advance). Therefore, through the pre-recommended content pool, the resource cost of real-time calculation can be reduced; by comparing the first preloading use probability and the online recommended threshold, high-quality preloading content (that is, recommended content generated in real time) can be allocated to the target object with high use probability, so that not only the content recommendation quality can be improved, but also the usage rate of the online recommended content can be improved.

[0175] Please refer to Fig. 13, which is a structural schematic diagram of a data processing apparatus provided by the embodiment of the present application. The data processing apparatus 1 can be used to execute the corresponding steps in the method provided by the embodiment of the present application. As shown in Fig. 13, the data processing apparatus 1 can include an acquisition module 11 and a generation module 12.

[0176] The acquisition module 11 is configured to acquire a preloading request sent by a terminal device, and acquire a first preloading usage probability corresponding to the terminal device according to the preloading request. The first preloading usage probability is used to indicate a probability of a target object corresponding to the terminal device interacting with preloading content. The generation module 12 is configured to generate an online recommendation content set for the terminal device if the first preloading usage probability is equal to or greater than an online recommendation threshold, and generate a target preloading content list according to the online recommendation content set. A generation timestamp of the online recommendation content set is later than a generation timestamp of the preloading request.

[0177] In some embodiments, the generation module 12 is further configured to generate the target preloading content list by using a pre-recommendation content pool for the terminal device if the first preloading usage probability is less than the online recommendation threshold. A generation timestamp of the pre-recommendation content pool is earlier than the generation timestamp of the preloading request.

[0178] In some embodiments, the acquisition module 11 is further configured to acquire the preloading request sent by the terminal device at a first time point, determine an application startup number and a preloading usage number of the terminal device within a first time period, and determine a number ratio between the application startup number and the preloading usage number, and determine the first preloading usage probability according to the number ratio.

[0179] In some embodiments, the data processing apparatus 1 further includes an association storage module configured to store the first preloading usage probability and a device identifier used to indicate the terminal device in association in a database.

[0180] In some embodiments, the acquisition module 11 is further configured to perform the following operations: determining A content types respectively corresponding to A preloading contents in the preloading content list, A being a positive integer; determining a content score corresponding to each preloading content in the A preloading contents according to a content type corresponding to the preloading content; summing the content scores respectively corresponding to the A preloading contents to obtain a list score corresponding to the preloading content list; the list score is used to determine a second preloading usage probability corresponding to a second time period; and storing the list score and the device identifier used to indicate the terminal device in association in the database.

[0181] In some embodiments, the obtaining module 11 is further configured to perform the following operations: obtaining, in the database, B historical preloaded content lists generated for the terminal device in a first time period; B is a positive integer; the first time period is earlier than the second time period, and the first time period is adjacent to the second time period; obtaining, in the database, historical list scores corresponding to the B historical preloaded content lists respectively, and performing summation processing on the B historical list scores to obtain a first score; determining, in the B historical preloaded content lists, C historical preloaded content lists output by the terminal; C is a natural number less than or equal to B; and determining the first preloaded use probability according to the first score and the C historical preloaded content lists.

[0182] In some embodiments, the obtaining module 11 determines the first preloaded use probability according to the first score and the C historical preloaded content lists, and is configured to perform the following operations: performing summation processing on historical list scores corresponding to the C historical preloaded content lists respectively to obtain a second score; determining a score ratio between the first score and the second score, and determining the first preloaded use probability according to the score ratio.

[0183] In some embodiments, the obtaining module 11 is further configured to perform the following operations: obtaining, in the database, D preloaded use probabilities of all terminal devices in the first time period; all terminal devices include the terminal device; the D preloaded use probabilities include the first preloaded use probability; performing sorting processing on the D preloaded use probabilities to obtain sorted D preloaded use probabilities; determining a first online recommendation quota corresponding to the system time, and determining an online recommendation threshold in the sorted D preloaded use probabilities according to the first online recommendation quota; the first online recommendation quota is a first total number of contents pulled by the terminal device before entering the content playing page.

[0184] In some embodiments, the obtaining module 11 determines the first online recommendation quota corresponding to the system time, and is configured to perform the following operations: determining a second online recommendation quota corresponding to the system time, and performing difference processing on a preset total online recommendation quota and the second online recommendation quota to obtain the first online recommendation quota; the obtaining module determines the online recommendation threshold in the sorted D preloaded use probabilities according to the first online recommendation quota, and is configured to perform the following operations: obtaining, in the sorted D preloaded use probabilities, E preloaded use probabilities with a number equal to the preloaded online recommendation quota; E is a positive integer less than or equal to D; and determining the minimum preloaded use probability in the E preloaded use probabilities as the online recommendation threshold.

[0185] In some embodiments, the obtaining module 11 is further configured to perform the following operations: obtaining a historical online recommendation content set generated for the terminal device, determining a first historical online recommendation content output by the terminal device in the historical online recommendation content set; adding a second historical online recommendation content to the initial pre-recommendation content pool; the second historical online recommendation content is a recommendation content in the historical online recommendation content set other than the first historical online recommendation content; generating the pre-recommendation content pool according to the initial pre-recommendation content pool.

[0186] In some embodiments, the generating module 12 generates the pre-recommendation content pool according to the initial pre-recommendation content pool, and is configured to perform the following operations: performing offline recommendation processing for the terminal device to obtain offline recommendation content in a first type period; adding the offline recommendation content to the initial pre-recommendation content pool, and determining the initial pre-recommendation content pool added with the offline recommendation content as the pre-recommendation content pool.

[0187] In some embodiments, the generating module 12 generates the target preload content list through the pre-recommendation content pool for the terminal device, and is configured to perform the following operations: determining F pre-recommendation contents in an active state in the pre-recommendation content pool for the terminal device; F is a positive integer; determining recommendation scores corresponding to the F pre-recommendation contents respectively, obtaining G recommendation scores in the F recommendation scores; G recommendation scores are greater than or equal to H recommendation scores; the H recommendation scores include recommendation scores other than the G recommendation scores in the F recommendation scores; G is a positive integer less than or equal to F; H is a positive integer less than F; obtaining pre-recommendation contents corresponding to the G recommendation scores in the F pre-recommendation contents, and generating the target preload content list according to the G pre-recommendation contents.

[0188] In some embodiments, the generating module 12 generates the target preload content list according to the G pre-recommendation contents, and is configured to perform the following operations: generating a first preload content list according to the G pre-recommendation contents; the G pre-recommendation contents include a first pre-recommendation content and a second pre-recommendation content; determining a content similarity between the first pre-recommendation content and the second pre-recommendation content; if the content similarity is equal to or greater than a similarity threshold, deleting the first pre-recommendation content in the first preload content list to obtain a second preload content list; obtaining a maximum recommendation score in the H recommendation scores, and obtaining a third pre-recommendation content corresponding to the maximum recommendation score in the F pre-recommendation contents; adding the third pre-recommendation content to the second preload content list to obtain a third preload content list; generating the target preload content list according to the third preload content list; the content similarity between each two pre-load contents in the target preload content list is less than the similarity threshold.

[0189] In some embodiments, the obtaining module 11 obtains the first preloading usage probability corresponding to the terminal device according to the preloading request, and is configured to perform the following operations: determining an obtaining timestamp of the preloading request, generating an online recommendation content set for the terminal device if the obtaining timestamp is in a first type period, and generating a target preloading content list according to the online recommendation content set; if the obtaining timestamp is in a second type period, obtaining the first preloading usage probability corresponding to the terminal device; the total number of output contents corresponding to the second type period is greater than the total number of output contents corresponding to the first type period.

[0190] In some embodiments, the generating module 12 generates an online recommendation content set for the terminal device, and is configured to perform the following operations: determining a total number of preloading usage probabilities equal to or greater than an online recommendation threshold; if the total number is less than or equal to a preloading online recommendation quota, generating an online recommendation content set for the terminal device according to the output content features of the terminal device and a candidate content pool corresponding to the system time; if the total number is greater than the preloading online recommendation quota, generating a target preloading content list through a pre-recommendation content pool.

[0191] As can be seen from the above, the embodiments of the present application propose a resource allocation method based on the first preloading usage probability, for the terminal device corresponding to the first preloading usage probability equal to or higher than the online recommendation threshold, an online recommendation content set generated in real time is provided, and then a preloading content list generated in real time can be provided; for the terminal device corresponding to the first preloading usage probability lower than the online recommendation threshold, preloading content is allocated from a pre-recommendation content pool (i.e. a recommendation content pool generated in advance). Therefore, through the pre-recommendation content pool, the resource cost of real-time calculation can be reduced; by comparing the first preloading usage probability and the online recommendation threshold, high-quality preloading content (i.e. recommendation content generated in real time) can be allocated to the target object with high usage probability, so not only the content recommendation quality can be improved, but also the usage rate of online recommendation content can be improved.

[0192] Please refer to FIG. 14, which is a structural schematic diagram of a computer device provided in an embodiment of the present application. As shown in FIG. 14, the computer device 1000 can include at least one processor 1001 (for example, a CPU), at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002. The communication bus 1002 is configured to realize the connection and communication among the components. In some embodiments, the user interface 1003 can include a display screen (Display) and a keyboard (Keyboard), and the network interface 1004 can optionally include a standard wired interface and a wireless interface (for example, a WI-FI interface). The memory 1005 can be a high-speed RAM memory or a non-volatile memory (for example, at least one disk memory). The memory 1005 can also be at least one storage device located away from the aforementioned processor 1001. As shown in FIG. 14, the memory 1005 as a computer storage medium can include an operating system, a network communication module, a user interface module, and a device control application.

[0193] In the computer device 1000 shown in FIG. 14, the network interface 1004 can provide network communication functions; the user interface 1003 is mainly configured to provide an interface for user input; and the processor 1001 can be configured to invoke the device control application stored in the memory 1005 to realize the following functions.

[0194] obtaining a preloading request sent by a terminal device, and obtaining a first preloading use probability corresponding to the terminal device according to the preloading request; the first preloading use probability is used to indicate a probability that a target object corresponding to the terminal device interacts with preloading content;

[0195] if the first preloading use probability is equal to or greater than an online recommendation threshold, generating an online recommendation content set for the terminal device, generating a target preloading content list according to the online recommendation content set, and generating a time stamp of the online recommendation content set later than a time stamp of the preloading request.

[0196] It should be understood that the computer device 1000 described in the embodiments of the present application can execute the description of the data processing method or device in the foregoing embodiments, which will not be described herein again. In addition, the beneficial effects of using the same method will not be described again.

[0197] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the description of the data processing method or device in the foregoing embodiments is realized, which will not be described herein again. In addition, the beneficial effects of using the same method will not be described again.

[0198] The computer readable storage medium can be an internal storage unit of the data processing apparatus or the computer device, for example, a hard disk or a memory of the computer device. The computer readable storage medium can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. The computer readable storage medium can also include both the internal storage unit and the external storage device of the computer device. The computer readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer readable storage medium can also be used to temporarily store data that has been output or will be output.

[0199] The computer program product includes a computer program stored in a computer readable storage medium. The processor of the computer device reads the computer program from the computer readable storage medium. The processor executes the computer program, so that the computer device can execute the description of the data processing method or apparatus in the foregoing embodiments. Here, the description will not be repeated. In addition, the beneficial effects of using the same method will not be repeated.

[0200] The terms "first", "second", and the like in the specification and claims of the present application and the accompanying drawings are used to distinguish different objects, and are not used to describe a specific order. In addition, the term "comprising" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product, or equipment including a series of steps or units is not limited to the listed steps or modules, but can optionally include steps or modules that are not listed, or can optionally include other steps or units inherent to the process, method, device, product, or equipment.

[0201] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, computer software, or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0202] The above descriptions are only the preferred embodiment of the application, of course, cannot be used to limit the scope of the application, thus the equivalent variations made by the claims of the application, still belongs to the scope of the application covered.

Claims

1. A data processing method applied to a computer device, comprising: obtaining a preloading request sent by a terminal device, and determining a first preloading usage probability corresponding to the terminal device according to the preloading request; the first preloading usage probability is used to indicate a probability that a target object corresponding to the terminal device interacts with preloading content; if the first preloading usage probability is equal to or greater than an online recommendation threshold, generating an online recommendation content set for the terminal device; generating a target preloading content list according to the online recommendation content set; a generation time stamp of the online recommendation content set is later than a generation time stamp of the preloading request.

2. The method of claim 1, wherein, The method further comprises: if the first preloading usage probability is less than the online recommendation threshold, generating a target preloading content list through a pre-recommendation content pool for the terminal device; a generation time stamp of the pre-recommendation content pool is earlier than the generation time stamp of the preloading request.

3. The method of claim 1 or 2, wherein, The method further comprises: obtaining a preloading request sent by a terminal device at a first time; determining the number of application start times and the number of preloading usage times of the terminal device within a first time period; the latest time stamp of the first time period is earlier than the first time; determining a ratio of the number of application start times to the number of preloading usage times, and determining the first preloading usage probability according to the ratio; The method further comprises: storing the first preloading usage probability and a device identifier used to indicate the terminal device in a database.

4. The method according to any one of claims 1 to 3, wherein, The method further comprises: determining A content types corresponding to A preloading contents in the preloading content list; A is a positive integer; for each preloading content in the A preloading contents, determining a content score corresponding to the preloading content according to the content type corresponding to the preloading content; summing the content scores corresponding to the A preloading contents to obtain a list score corresponding to the preloading content list; the list score is used to determine a second preloading usage probability corresponding to a second time period; storing the list score and a device identifier used to indicate the terminal device in a database.

5. The method according to any one of claims 1 to 4, wherein, The method further comprises: in the database, obtaining B historical preloading content lists generated for the terminal device within a first time period; B is a positive integer; the first time period is earlier than the second time period, and the first time period is adjacent to the second time period; in the database, obtaining historical list scores corresponding to the B historical preloading content lists, and summing the B historical list scores to obtain a first score; in the B historical preloading content lists, determining C historical preloading content lists output by the terminal device; C is a natural number less than or equal to B; determining the first preloading usage probability according to the first score and the C historical preloading content lists.

6. The method according to any one of claims 1 to 5, wherein, The first preloading use probability is determined according to the first score and the C historical preloading content lists. The second score is obtained by summing the historical list scores corresponding to the C historical preloading content lists. The first preloading use probability is determined according to a score ratio between the first score and the second score.

7. The method according to any one of claims 1 to 6, wherein, The method further comprises: D preloading use probabilities of all terminal devices in a first time period are obtained from a database; the all terminal devices include the terminal device; the D preloading use probabilities include the first preloading use probability; The D preloading use probabilities are sorted to obtain sorted D preloading use probabilities; A first online recommendation quota corresponding to a system time is determined, and the online recommendation threshold is determined from the sorted D preloading use probabilities according to the first online recommendation quota; the first online recommendation quota is a first total number of contents pulled by the terminal device before entering a content playing page.

8. The method according to any one of claims 1 to 7, wherein, The first online recommendation quota corresponding to the system time is determined, comprising: A second online recommendation quota corresponding to the system time is determined, and a first online recommendation quota is obtained by subtracting the second online recommendation quota from a preset total online recommendation quota; the second online recommendation quota is a second total number of contents pulled by the terminal device after entering the content playing page; The online recommendation threshold is determined from the sorted D preloading use probabilities according to the first online recommendation quota, comprising: E preloading use probabilities with a number equal to the first online recommendation quota are obtained from the sorted D preloading use probabilities; E is a positive integer less than or equal to D; The smallest preloading use probability in the E preloading use probabilities is determined as the online recommendation threshold.

9. The method according to any one of claims 1 to 8, wherein, The method further comprises: A historical online recommendation content set generated for the terminal device is obtained, and a first historical online recommendation content output by the terminal device is determined in the historical online recommendation content set; A second historical online recommendation content is added to an initial pre-recommendation content pool; the second historical online recommendation content is a recommendation content in the historical online recommendation content set except the first historical online recommendation content; The pre-recommendation content pool is generated according to the initial pre-recommendation content pool.

10. The method according to any one of claims 1 to 9, wherein, The pre-recommendation content pool is generated according to the initial pre-recommendation content pool, comprising: Offline recommendation processing is performed on the terminal device to obtain offline recommendation contents in a first type period; The offline recommendation contents are added to the initial pre-recommendation content pool, and the initial pre-recommendation content pool to which the offline recommendation contents are added is determined as the pre-recommendation content pool.

11. The method according to any one of claims 1 to 10, wherein, The target preloading content list is generated from the pre-recommendation content pool for the terminal device, comprising: F pre-recommendation contents in an effective state are determined in the pre-recommendation content pool for the terminal device; F is a positive integer. determining recommendation scores corresponding to the F pieces of pre-recommended content respectively, obtaining G pieces of recommendation scores from the F pieces of recommendation scores; the G pieces of recommendation scores are greater than or equal to H pieces of recommendation scores; the H pieces of recommendation scores include recommendation scores other than the G pieces of recommendation scores from the F pieces of recommendation scores; G is a positive integer less than or equal to F; H is a positive integer less than F; from the F pieces of pre-recommended content, obtaining pre-recommended content corresponding to the G pieces of recommendation scores respectively, and generating the target pre-loaded content list according to the G pieces of pre-recommended content.

12. The method according to any one of claims 1 to 11, wherein, The generating the target pre-loaded content list according to the G pieces of pre-recommended content includes: generating a first pre-loaded content list according to the G pieces of pre-recommended content; the G pieces of pre-recommended content include a first pre-recommended content and a second pre-recommended content; determining a content similarity between the first pre-recommended content and the second pre-recommended content; if the content similarity is equal to or greater than a similarity threshold, deleting the first pre-recommended content from the first pre-loaded content list to obtain a second pre-loaded content list; obtaining a maximum recommendation score from the H pieces of recommendation scores, and obtaining third pre-recommended content corresponding to the maximum recommendation score from the F pieces of pre-recommended content; adding the third pre-recommended content to the second pre-loaded content list to obtain a third pre-loaded content list; generating the target pre-loaded content list according to the third pre-loaded content list; a content similarity between each two pre-loaded contents in the target pre-loaded content list is less than the similarity threshold.

13. The method of any one of claims 1 to 12, wherein, The obtaining the first pre-loaded use probability corresponding to the terminal device according to the pre-loaded request includes: determining an acquisition timestamp of the pre-loaded request, if the acquisition timestamp is in a first type period, generating an online recommended content set for the terminal device, and generating a target pre-loaded content list according to the online recommended content set; if the acquisition timestamp is in a second type period, obtaining the first pre-loaded use probability corresponding to the terminal device; a total number of output contents corresponding to the second type period is greater than a total number of output contents corresponding to the first type period.

14. The method of any one of claims 1 to 13, wherein, The generating the online recommended content set for the terminal device includes: determining a total number of pre-loaded use probabilities equal to or greater than the online recommendation threshold; if the total number is less than or equal to a first online recommendation quota, generating an online recommended content set for the terminal device according to output content features of the terminal device and a candidate content pool corresponding to a system time; if the total number is greater than the first online recommendation quota, generating a target pre-loaded content list through the pre-recommended content pool.

15. A data processing apparatus, comprising: an obtaining module configured to obtain a pre-loaded request sent by a terminal device, and obtain a first pre-loaded use probability corresponding to the terminal device according to the pre-loaded request; the first pre-loaded use probability is used to indicate a probability of a target object corresponding to the terminal device interacting with pre-loaded content; The generation module is further configured to generate an online recommendation content set for the terminal device if the first preloading use probability is equal to or greater than an online recommendation threshold, and generate a target preloading content list according to the online recommendation content set, wherein a generation timestamp of the online recommendation content set is later than a generation timestamp of the preloading request.

16. A computer device comprising: a processor, a memory, and a network interface; The processor is connected with the memory and the network interface, wherein the network interface is configured to provide a data communication function, the memory is configured to store a computer program, and the processor is configured to call the computer program to enable the computer device to execute the method in any one of claims 1 to 14. 17.A computer readable storage medium, having stored therein a computer program, the computer program being adapted to be loaded and executed by a processor to enable a computer device having the processor to execute the method in any one of claims 1 to 14. 18.A computer program product, comprising a computer program stored in a computer readable storage medium, the computer program being adapted to be read and executed by a processor to enable a computer device having the processor to execute the method in any one of claims 1 to 14.

Images