An account level adjustment method, device, equipment and medium

Abstract

The embodiments of the present application provide a method, device, equipment, and medium for adjusting an account level. The method includes: when receiving target sample content uploaded by a first client through a first user account, obtaining sample metadata information and posting transaction information; obtaining an account level model for grading the first user account; inputting account statistical information and content metadata information into a prior level model, and obtaining prior level information from the prior level model; adding the target sample content to a content recommendation pool associated with a content database based on the prior level information; before distributing the target sample content through the content recommendation pool, inputting account contribution information and content metadata information into a posterior level model, outputting posterior level information by the posterior level model, and correcting the account level corresponding to the prior level information through the account level corresponding to the posterior level information. The present application can achieve dynamic adjustment of the account level, thereby improving the efficiency of level adjustment.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to an account level adjustment method, device, equipment and medium. Background Art

[0002] At present, information flow platforms mainly recommend information flow content (e.g., pictures, texts, videos, etc.) based on the level of accounts (e.g., account Q), but the account level here is often set manually. For example, after receiving reports and complaints from copyright holders or other stakeholders, the information flow platform can periodically adjust the level of account Q through manual review (e.g., based on account Q's popularity in the industry, its performance on other information flow platforms, and personal account experience, etc.).

[0003] However, the inventors have found in practice that once the posting behavior of the content producer through the account Q on the information flow platform is abnormal, the aforementioned manual review method will be difficult to update the level of the account Q with reduced content quality and activity in the first place, so that it is impossible to adjust the level of the account Q in real time and intelligently. In addition, in the face of the massive accounts on the information flow platform, the manual review method often consumes a long manual processing time, which reduces the efficiency of adjusting the levels of these accounts. Summary of the invention

[0004] The embodiments of the present application provide an account level adjustment method, apparatus, device, and medium, which can realize dynamic adjustment of account levels, thereby improving the efficiency of level adjustment.

[0005] On the one hand, an embodiment of the present application provides a method for adjusting an account level, including:

[0006] Upon receiving the target sample content uploaded by the first client through the first user account, obtaining sample metadata information and posting flow information of the target sample content, adding the sample metadata information and the target sample content to a content database to be distributed, and adding the posting flow information to a statistical database;

[0007] Acquire an account level model for classifying the first user account; the account level model includes a priori level model and a posteriori level model;

[0008] Obtain account statistical information associated with the first user account from the statistical database, and obtain content metadata information associated with the first user account from the content database. Input the account statistical information and the content metadata information into the prior level model, and the prior level model identifies the account level of the first user account. Use the identified account level of the first user account as the prior level information of the first user account; the content metadata information includes sample metadata information; the account statistical information includes post flow information;

[0009] Based on the prior level information, add the target sample content to the content recommendation pool associated with the content database;

[0010] Before distributing the target sample content through the content recommendation pool, obtain account contribution information associated with the first user account from the statistical database. Input the account contribution information and the content metadata information into the posterior level model, and the posterior level model outputs the posterior level information of the first user account. Use the account level indicated by the posterior level information to correct the account level indicated by the prior level information.

[0011] One aspect of the embodiments of the present application provides an account level adjustment device, including:

[0012] A content receiving module, configured to obtain the sample metadata information and post flow information of the target sample content when receiving the target sample content uploaded by the first client through the first user account, add the sample metadata information and the target sample content to the content database to be distributed, and add the post flow information to the statistical database;

[0013] A model obtaining module, configured to obtain an account level model for grading the first user account; the account level model includes a prior level model and a posterior level model;

[0014] A first identification module, configured to obtain account statistical information associated with the first user account from the statistical database, and obtain content metadata information associated with the first user account from the content database. Input the account statistical information and the content metadata information into the prior level model, and the prior level model identifies the account level of the first user account. Use the identified account level of the first user account as the prior level information of the first user account; the content metadata information includes sample metadata information; the account statistical information includes post flow information;

[0015] A content adding module, configured to add the target sample content to the content recommendation pool associated with the content database based on the prior level information;

[0016] A second recognition module, configured to obtain account contribution information associated with the first user account from a statistical database before the target sample content is sent through the content recommendation pool, input the account contribution information and content metadata information into a posteriori level model, output the posteriori level information of the first user account by the posteriori level model, and correct the account level indicated by the prior level information according to the account level indicated by the posteriori level information.

[0017] The prior level model includes a first prior sub-model, a second prior sub-model, a third prior sub-model, and a fourth prior sub-model.

[0018] The first recognition module includes:

[0019] A first acquisition unit, configured to obtain an account influence parameter associated with the first prior sub-model, a content smoothness parameter associated with the second prior sub-model, a content verticality parameter associated with the third prior sub-model, and a content matching parameter associated with the fourth prior sub-model from the account statistical information and content metadata information.

[0020] A first analysis unit, configured to input the account influence parameter into the first prior sub-model, and the first prior sub-model performs an influence analysis on the first user account based on the account influence parameter to obtain the account influence degree of the first user account.

[0021] A second analysis unit, configured to input the content smoothness parameter into the second prior sub-model, and the second prior sub-model performs a smoothness analysis on the first user account based on the content smoothness parameter to obtain the content smoothness of the first user account.

[0022] A third analysis unit, configured to input the content verticality parameter into the third prior sub-model, and the third prior sub-model performs a verticality analysis on the first user account based on the content verticality parameter to obtain the content verticality of the first user account.

[0023] A fourth analysis unit, configured to input the content matching parameter into the fourth prior sub-model, and the fourth prior sub-model performs a matching degree analysis on the first user account based on the content matching parameter to obtain the content matching degree of the first user account.

[0024] A first fusion unit, configured to perform prior data fusion on the account influence degree, content smoothness, content verticality, and content matching degree to obtain a prior data fusion result of the first user account, and based on the prior data fusion result, use the recognized account level of the first user account as the prior level information of the first user account.

[0025] The account influence parameter includes a first account set and a second account set corresponding to the first user account; the first account set is the set of accounts that follow the first user account; the second account set is the set of accounts that unfollow the first user account.

[0026] The first analysis unit includes:

[0027] A first determination subunit, configured to obtain the number of followed accounts and the number of followers corresponding to the first account in the first account set, and determine a first weight associated with the first account based on the number of followed accounts and the number of followers;

[0028] A first averaging subunit, configured to determine a first average weight corresponding to each first account based on the first weight associated with each first account in the first account set and the first timestamp when each first account follows the first user account;

[0029] A second determination subunit, configured to obtain the number of unfollowed accounts and the number of accounts that have been unfollowed corresponding to the second account in the second account set, and determine a second weight associated with the second account based on the number of unfollowed accounts and the number of accounts that have been unfollowed;

[0030] A second averaging subunit, configured to determine a second average weight corresponding to each second account based on the second weight associated with each second account in the second account set and the second timestamp when each second account unfollows the first user account;

[0031] A first output subunit, configured to determine the number of accounts of the first account as the number of auxiliary accounts, and the first prior sub-model obtains the account influence degree of the first user account based on the number of auxiliary accounts, the first average weight corresponding to each first account, and the second average weight corresponding to each second account.

[0032] Wherein, the content stability parameter includes the number of samples of the sample content of the first user account in the first period;

[0033] The second analysis unit includes:

[0034] A mean processing subunit, configured to perform mean processing on the number of samples to obtain the sample mean of the first user account in the first period;

[0035] A variance processing subunit, configured to perform variance processing on the number of samples according to the sample mean to obtain the sample variance of the first user account in the first period;

[0036] A second output subunit, configured to obtain the content stability of the first user account by the second prior sub-model based on the sample mean and the sample variance.

[0037] Wherein, the content vertical parameter includes the classification information of the sample content of the first user account in the first period;

[0038] The third analysis unit includes:

[0039] A probability determination subunit, configured to determine the number of vertical categories of the classified information, and determine the vertical category feature probability corresponding to each vertical category information in the first period based on the number of vertical categories corresponding to each vertical category information in the number of vertical categories of the classified information and the number of vertical categories of the classified information;

[0040] A third output subunit, configured to obtain the content verticality of the first user account by a third prior sub-model based on the vertical category feature probability corresponding to each vertical category information.

[0041] Wherein, the content matching parameters include the label information of the sample content of the first user account in the first period, the post timestamp of the sample content, the account profile of the first user account, and the account name of the first user account;

[0042] The fourth analysis unit includes:

[0043] A word segmentation processing subunit, configured to perform word segmentation processing on the account profile and the account name to obtain the text word segmentation of the first user account, and perform string matching on the label information and the text word segmentation to obtain a string matching result;

[0044] A first matching subunit, configured to, if the string matching result indicates that there is a text word segmentation in the text word segmentation that matches the label information, determine the number of text word segments that match the label information as the matching number corresponding to the sample content;

[0045] A second matching subunit, configured to, if the string matching result indicates that there is no text word segmentation in the text word segmentation that matches the label information, determine the matching number corresponding to the sample content based on the fact that there is no text word segmentation that matches the label information;

[0046] A weight determination subunit, configured to determine the first weight information corresponding to each sample content in the sample content according to the post timestamp;

[0047] A fourth output subunit, configured to determine the number of labels of the label information of each sample content, and obtain the content matching degree of the first user account by a fourth prior sub-model based on the number of labels corresponding to each sample content, the first weight information corresponding to each sample content, and the matching number corresponding to each sample content.

[0048] Wherein, the posterior account level model includes a first posterior sub-model, a second posterior sub-model, and a third posterior sub-model;

[0049] The second recognition module includes:

[0050] A second acquisition unit, configured to acquire a platform contribution parameter associated with the first posterior sub-model, an account interaction parameter associated with the second posterior sub-model, and a content difference parameter associated with the third posterior sub-model from the account contribution information and the content metadata information;

[0051] A fifth analysis unit for inputting the platform contribution parameter into the first posterior sub-model, and the first posterior sub-model analyzes the contribution degree of the first user account based on the platform contribution parameter to obtain the platform contribution degree of the first user account;

[0052] A sixth analysis unit for inputting the account interaction parameter into the second posterior sub-model, and the second posterior sub-model analyzes the interaction degree of the first user account based on the account interaction parameter to obtain the account interaction degree of the first user account;

[0053] A seventh analysis unit for inputting the content difference parameter into the third posterior sub-model, and the third posterior sub-model analyzes the difference of the first user account based on the content difference parameter to obtain the content difference degree of the first user account;

[0054] A second fusion unit for performing posterior data fusion on the platform contribution degree, the account interaction degree and the content difference degree to obtain the posterior data fusion result of the first user account, and based on the posterior data fusion result, obtaining the posterior level information of the first user account, and correcting the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0055] Among them, the platform contribution parameter includes the posted content of the first user account and the viewing behavior corresponding to the posted content; the viewing behavior includes the viewing duration and viewing time of the viewing user viewing the posted content;

[0056] The fifth analysis unit includes:

[0057] A content acquisition sub-unit for acquiring the posted content within the validity period from the posted content and determining the acquired posted content as the valid posted content;

[0058] A quantity determination sub-unit for performing validity analysis on the viewing behavior corresponding to the valid posted content to obtain the valid viewing behavior of the viewing user viewing the valid posted content within the second period, and determining the quantity of the viewing users corresponding to the valid viewing behavior as the valid viewing quantity;

[0059] A data acquisition sub-unit for acquiring the valid viewing duration of the valid posted content within the second period from the viewing duration, and determining the second weight information corresponding to the valid viewing duration according to the valid viewing time corresponding to the valid viewing duration; the valid viewing time is acquired from the viewing time;

[0060] A duration acquisition sub-unit for obtaining the consumption duration corresponding to the valid posted content based on the valid viewing duration and the second weight information, and acquiring the target consumption duration that meets the duration acquisition condition from the consumption duration;

[0061] A fifth output subunit, configured to obtain the platform contribution degree of the first user account based on the effective viewing quantity and the target consumption duration by the first posterior sub-model.

[0062] Among them, the second period includes the T i th day; the account interaction parameter includes the viewing information of the effective posted content on the T i th day, the first interaction information on the T i th day, and the second interaction information on the T i th day;

[0063] The sixth analysis unit includes:

[0064] A weighted summation subunit, configured to perform weighted summation on the first interaction information and the second interaction information to obtain the interaction quantity corresponding to the T i th day, and based on the interaction quantity and the viewing information, obtain the average interaction quantity of the first user account on the T i th day;

[0065] An accumulation processing subunit, configured to perform accumulation processing on the interaction quantities of each day of the first user account within the second period to obtain the total interaction quantity of the first user account within the second period;

[0066] A sixth output subunit, configured to obtain the account interaction degree of the first user account based on the total interaction quantity and the average interaction quantity by the second posterior sub-model.

[0067] Among them, the content difference parameter includes the label information of the sample content of the first user account within the second period; the sample content includes sample content S i and sample content S i+1 ; sample content S i+1 is the next sample content of sample content S i ;

[0068] The seventh analysis unit includes:

[0069] A label determination subunit, configured to determine the label information of sample content S i as the first label information, and determine the label information of sample content S i+1 as the second label information;

[0070] An intersection processing subunit, configured to perform intersection processing on the first label information and the second label information to obtain intersection label information, and determine the label quantity corresponding to the intersection label information as the first label quantity;

[0071] A union processing subunit, configured to perform union processing on the first label information and the second label information to obtain union label information, and determine the label quantity corresponding to the union label information as the second label quantity;

[0072] A seventh output subunit, configured to determine a sample content S based on a first tag quantity and a second tag quantity i and the sample content S i+1 The content similarity of is obtained by a third posterior sub-model based on the content similarity, and the content difference degree of the first user account is obtained

[0073] Wherein, the apparatus further includes:

[0074] A first calling module, configured to call a prior level model through an account grading service to obtain prior level information indicated by the prior level model

[0075] A system review module, configured to call an audit system through an account grading service, perform a system review on the prior level information through the audit system, and write the prior level information that passes the system review into a level database

[0076] A second calling module, configured to call a posterior level model through an account grading service to obtain posterior level information indicated by the posterior level model, and write the posterior level information into a level database

[0077] Wherein, the apparatus further includes:

[0078] A data acquisition module, configured to obtain a network environment, exposure data, and interaction behaviors of a second user account corresponding to a second client for a target sample content through a statistical reporting service; the network environment is determined by a caching time for the second client to cache the target sample content; the exposure data includes a target viewing time and a target viewing duration of the second user account for the target sample content; the interaction behaviors include a first interaction operation of the second user account for the target sample content and a second interaction operation for the first user account

[0079] A data addition module, configured to determine the network environment, exposure data, and interaction behaviors as end statistical information associated with the second client, and add the end statistical information to a statistical database

[0080] An embodiment of the present application provides a computer device on the one hand, including: a processor and a memory

[0081] The processor is connected to the memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the computer device executes the method provided by the embodiment of the present application

[0082] An embodiment of the present application provides a computer-readable storage medium on the one hand. The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded and executed by a processor so that a computer device having the processor executes the method provided by the embodiment of the present application

[0083] One aspect of the embodiments of the present application provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided by the embodiments of the present application.

[0084] In the embodiments of the present application, when a computer device receives a certain sample content (for example, the target sample content uploaded by the first client through the first user account), it can obtain the sample metadata information and the publication flow information of the target sample content, add the sample metadata information and the target sample content to the content database to be data-distributed, and add the publication flow information to the statistical database. Further, the computer device can obtain an account level model for grading the first user account, where the account level model can include a prior level model and a posterior level model. Further, the computer device can obtain the account statistical information associated with the first user account from the statistical database, and obtain the content metadata information associated with the first user account from the content database, input the account statistical information and the content metadata information into the prior level model, and the prior level model identifies the account level of the first user account, and uses the identified account level of the first user account as the prior level information of the first user account. Among them, the content metadata information may include the sample metadata information, and the account statistical information may include the publication flow information. Further, the computer device can add the target sample content to the content recommendation pool associated with the content database based on the prior level information. Further, before the computer device distributes the target sample content through the content recommendation pool, it can obtain the account contribution information associated with the first user account from the statistical database, input the account contribution information and the content metadata information into the posterior level model, and the posterior level model outputs the posterior level information of the first user account, and corrects the account level indicated by the prior level information through the account level indicated by the posterior level information. It can be seen that the embodiments of the present application can quickly and accurately identify the prior level information and the posterior level information of the first user account through the prior level model and the posterior level model with level recognition functions. The prior level information can be used to describe the prior account level of the first user account (that is, the account level indicated by the prior level information), and the posterior level information can be used to describe the posterior account level of the first user account (that is, the account level indicated by the posterior level information). The prior account level can be used to grade the account level of the first user account, and the posterior account level can be used to dynamically update the account level of the first user account. In other words, the posterior account level can be used to correct the prior account level. Based on this, the embodiments of the present application can improve the efficiency of level recognition and achieve dynamic adjustment of the account level based on the prior level model and the posterior level model, thereby improving the efficiency of level adjustment. Description of the Drawings

[0085] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0086] Figure 1 is a schematic structural diagram of a network architecture provided by an embodiment of the present application;

[0087] Figure 2 is a schematic diagram of a scenario for data interaction provided by an embodiment of the present application;

[0088] Figure 3 is a schematic flowchart of a method for adjusting account levels provided by an embodiment of the present application;

[0089] Figure 4 is a schematic flowchart of a priori level model provided by an embodiment of the present application;

[0090] Figure 5 is a schematic flowchart of a process for label matching provided by an embodiment of the present application;

[0091] Figure 6 is a schematic flowchart of a posteriori level model provided by an embodiment of the present application;

[0092] Figure 7 is a schematic flowchart of a method for adjusting account levels provided by an embodiment of the present application;

[0093] Figure 8 is a system flowchart of a dynamic grading of self-media account levels provided by an embodiment of the present application;

[0094] Figure 9 is a schematic structural diagram of a device for adjusting account levels provided by an embodiment of the present application;

[0095] Figure 10 is a schematic structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0096] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0097] It should be understood that artificial intelligence (AI) is a theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines to enable machines to have the functions of perception, reasoning, and decision-making.

[0098] Artificial intelligence technology is an interdisciplinary subject that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, autonomous driving, and intelligent transportation.

[0099] Among them, the solution provided in the embodiments of this application mainly relates to the machine learning (ML) technology of artificial intelligence. Machine learning is an interdisciplinary subject that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and rote learning.

[0100] Specifically, please refer to Figure 1 , Figure 1 which is a schematic structural diagram of a network architecture provided by the embodiments of this application. As Figure 1 shown, the network architecture may include a business server 2000 and a user terminal cluster. Among them, the user terminal cluster may specifically include one or more user terminals, and the number of user terminals in the user terminal cluster will not be limited here. As Figure 1As shown, multiple user terminals may specifically include user terminal 3000a, user terminal 3000b, user terminal 3000c, …, user terminal 3000n; user terminal 3000a, user terminal 3000b, user terminal 3000c, …, user terminal 3000n may be directly or indirectly network-connected to the service server 2000 through wired or wireless communication methods, so that each user terminal can perform data interaction with the service server 2000 through this network connection.

[0101] Among them, the service server 2000 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0102] Among them, each user terminal in the user terminal cluster may include: intelligent terminals with account level adjustment functions such as smart phones, tablet computers, laptop computers, desktop computers, smart homes, wearable devices, in-vehicle terminals, etc. It should be understood that, as Figure 1 shown, each user terminal in the user terminal cluster may be integrated with an application client (i.e., terminal program). When the application client runs on each user terminal, it can perform data interaction with the above-mentioned Figure 1 shown service server 2000 respectively. Among them, the application client may specifically include: in-vehicle client, smart home client, entertainment client (such as game client), multimedia client (such as video client), social client, and information client (such as news client), etc.

[0103] For ease of understanding, embodiments of the present application may select one user terminal as the target user terminal among the Figure 1 shown multiple user terminals. For example, embodiments of the present application may use the Figure 1 shown user terminal 3000a as the target user terminal, and an application client with account level adjustment function may be integrated in this target user terminal. At this time, this target user terminal can achieve data interaction with the service server 2000 through this application client.

[0104] For ease of understanding, embodiments of the present application may collectively refer to the information flow content (such as videos, pictures and texts) uploaded by a certain user (such as user Y1) through the application client as the target sample content. For ease of understanding, embodiments of the present application may collectively refer to the information flow content (such as videos, pictures and texts) recommended to a certain user (such as user Y2) as the target recommended content.

[0105] It should be understood that the application client in the embodiments of the present application can be integrated into a certain client (for example, a social client). For example, when the application client is integrated into the social client, the application client can be QQ Browser. Optionally, the application client in the embodiments of the present application can also be a client independent of the above social client (for example, a news client). For example, when the application client is an independent client, the application client can be KuaiBao. The embodiments of the present application do not limit the type of the application client.

[0106] It can be understood that in the embodiments of the present application, a user who logs in to the application client through a first user account (for example, account information 1) can be referred to as a first user (for example, the above user Y1). The user terminal corresponding to the first user can be referred to as a first terminal, and the application client integrated and installed on the first terminal can be a first client. In the embodiments of the present application, any user terminal in the above user terminal cluster can be selected as the first terminal. For example, in the embodiments of the present application, the user terminal 3000a in the above user terminal cluster can be used as the first terminal. It should be understood that the first user in the embodiments of the present application can be a user who uploads information flow content (for example, target sample content) through the first client, that is, a content uploader.

[0107] It can be understood that in the embodiments of the present application, a user who logs in to the application client through a second user account (for example, account information 2) can be referred to as a second user (for example, the above user Y2). The user terminal corresponding to the second user can be referred to as a second terminal, and the application client integrated and installed on the second terminal can be a second client. In the embodiments of the present application, any user terminal in the above user terminal cluster can be selected as the second terminal. For example, in the embodiments of the present application, the user terminal 3000b in the above user terminal cluster can be used as the second terminal. It should be understood that the second user in the embodiments of the present application can be a user who receives information flow content (for example, target recommended content) through the second client, that is, a content receiver.

[0108] It should be understood that the first user in the embodiments of the present application can be both the above content uploader and the above content receiver. For example, the first user can become a content uploader through the first client in the first terminal, and the first user can also become a content receiver through the first client in the first terminal. Similarly, the second user in the embodiments of the present application can be both the above content receiver and the above content uploader. For example, the second user can become a content receiver through the second client in the second terminal, and the second user can also become a content uploader through the second client in the second terminal.

[0109] Among them, it can be understood that the business scenarios applicable to the above network framework may specifically include: video distribution scenarios, video search scenarios, etc. Here, specific business scenarios will not be listed one by one.

[0110] For example, in a video distribution scenario, a computer device (e.g., the above business server 2000) can obtain target recommended content for recommending to user Y2 from a content recommendation pool based on the user profile of user Y2. Among them, the business server 2000 can obtain the content uploaded by high-level accounts from the content recommendation pool as the target recommended content. For another example, in a video search scenario, when user Y2 conducts content search, the business server 2000 can obtain target recommended content for recommending to user Y2 from the content recommendation pool based on the search content entered by user Y2. Among them, the business server 2000 can obtain the content uploaded by high-level accounts from the content recommendation pool as the target recommended content corresponding to the search results.

[0111] For ease of understanding, further, please refer to Figure 2 , Figure 2 which is a schematic diagram of a scenario for data interaction provided by an embodiment of the present application. As Figure 2 shown, the server 20a can be the business server 2000 in the corresponding embodiment above. As Figure 1 shown, the user terminal 20b (i.e., the first terminal) can be any user terminal in the user terminal cluster in the corresponding embodiment above. For ease of understanding, in the embodiment of the present application, the user terminal 3000a shown above is taken as the user terminal 20b as an example to elaborate Figure 2 the specific process of data interaction between the server 20a and the user terminal 20b shown. Among them, an application client (i.e., the first client) is installed on the user terminal 20b, and the application client can upload target sample content through the first user account. Among them, the user corresponding to the user terminal 20b can be user 20c (i.e., the first user). As Figure 1 shown, the database 200c can include a content database 200a and a statistical database 200b. Figure 1 Figure 2 Figure 2 Figure 2 Figure 2 shown, the content database 200a can include multiple databases, and the multiple databases can specifically include

[0112] Among them, as Figure 2 shown, the content database 200a can include multiple databases, and the multiple databases can specifically include Figure 2The databases 21a, …, 21n shown. This means that the content database 200a can be used to store data corresponding to accounts of different account levels (e.g., target sample content and sample metadata information). For example, the database 21a can be used to store the information flow content corresponding to an account with an account level of D1 and the sample metadata information of this information flow content, …, the database 21n can be used to store the information flow content corresponding to an account with an account level of D2 and the sample metadata information of this information flow content.

[0113] Among them, as Figure 2 shown, the statistical database 200b can include multiple databases, and the multiple databases can specifically include Figure 2 the databases 22a, …, 22n shown. This means that the statistical database 200b can be used to store data corresponding to accounts of different account levels (e.g., post flow information). For example, the database 22a can be used to store the post flow information corresponding to an account with an account level of D1, …, the database 22n can be used to store the post flow information corresponding to an account with an account level of D2.

[0114] As Figure 2 shown, the user 20c can upload the target sample content to the server 20a through the application client in the user terminal 20b. Therefore, when the server 20a receives the target sample content, it can obtain the sample metadata information and post flow information of the target sample content, and then store the target sample content, sample metadata information, and post flow information in the database 200c. Among them, the server 20a can store the sample metadata information and the target sample content in the content database 200a, and store the post flow information in the statistical database 200b.

[0115] It can be understood that when the server 20a needs to determine the account level corresponding to the first user account, it can obtain the account level model 23c used for account grading of the first user account. Here, the account level model 23c can include a prior level model and a posterior level model. As Figure 2 shown, the server 20a can obtain the prior information used by the prior level model for prior level identification and the posterior information used by the posterior level model for posterior level identification from the database 200c. For example, the prior information here can be the prior information 23a, and the posterior information here can be the posterior information 23b.

[0116] Among them, the prior information 23a may include account statistical information and content metadata information. The account statistical information may be obtained from the statistical database 200b, and the content metadata information may be obtained from the content database 200a. For example, the server 20a may obtain the content metadata information from the database 21a and the account statistical information from the database 22a. Among them, the posterior information 23b may include account contribution information and content metadata information. The account contribution information may be obtained from the statistical database 200b, and the content metadata information may be obtained from the content database 200a. For example, the server 20a may obtain the content metadata information from the database 21a and the account contribution information from the database 22a. It can be understood that the information used by the prior ranking model and the posterior ranking model for ranking identification may belong to the prior information 23a and the posterior information 23b respectively. The content metadata information in the prior information 23a and the content metadata information in the posterior information 23b may contain different content or the same content.

[0117] As Figure 2 shown, after obtaining the prior information 23a from the database 200c, the server 20a may input the prior information 23a into the prior ranking model, and the prior ranking model may identify the account level of the first user account to obtain the prior level information of the first user account. Further, the server 20a may obtain the posterior information 23b from the database 200c, input the posterior information 23b into the posterior ranking model, and the posterior ranking model may identify the account level of the first user account to obtain the posterior level information of the first user account. Among them, the prior ranking identification and the posterior ranking identification do not occur simultaneously here. The posterior ranking identification requires data accumulation, so the occurrence time of the posterior ranking identification is later than that of the prior ranking identification. Further, after determining the prior level information and the posterior level information of the first user account, the server 20a may correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0118] Among them, it can be understood that the server 20a may identify the account level of the first user account through the prior ranking model at time T1, and identify the account level of the first user account through the posterior ranking model at time T2. Similarly, the account level of the first user account is identified through the posterior ranking model at time T3. Among them, time T2 may be the next moment of time T1, and time T3 may be the next moment of time T2.

[0119] It can be seen that the embodiments of the present application can intelligently identify and adjust the account level of the first user account through the account level model. Here, the account level model can include a prior level model and a posterior level model. The prior level model can determine the prior level information of the first user account based on the prior information associated with the first user account, and the posterior level model can determine the posterior level information of the first user account based on the posterior information associated with the first user account. It can be understood that when the prior level information and the posterior level information of the first user account are identified, the server can dynamically adjust the account level of the first user account based on the prior level information and the posterior level information, thereby realizing the dynamic adjustment of the account level and improving the efficiency of level adjustment.

[0120] Among them, for the specific implementation method of data interaction between the application client in the server 20a and the user terminal 20b, reference can be made to the following Figures 3 - 8 description of data interaction between the server and the first client in the corresponding embodiment.

[0121] Further, please refer to Figure 3 , Figure 3 which is a schematic flowchart of an account level adjustment method provided by the embodiments of the present application. This method can be executed by the server, or by the first client, or jointly executed by the server and the first client. The server can be the server 20a in the above Figure 2 corresponding embodiment, and the first client can be the application client in the above Figure 2 corresponding embodiment. For ease of understanding, this embodiment is described by taking the method being executed by the server as an example. Among them, the account level adjustment method can include the following steps S101 - step S105:

[0122] Step S101, when receiving the target sample content uploaded by the first client through the first user account, obtain the sample metadata information and the posting flow information of the target sample content, add the sample metadata information and the target sample content to the content database to be distributed, and add the posting flow information to the statistical database;

[0123] Specifically, the server can obtain the sample metadata information and the posting flow information of the target sample content when receiving the target sample content uploaded by the first client through the first user account through the uplink and downlink content interface service. Among them, the posting flow information can include the posting time and content type of the target sample content. Further, the server can add the sample metadata information and the target sample content to the content database to be distributed, and send the posting flow information to the statistical reporting service, and add the posting flow information to the statistical database through the statistical reporting service.

[0124] Among them, the content types of the information flow content may include text and pictures (i.e., articles), picture sets, vertical short videos, and horizontal short videos. The text and pictures may include videos or pictures. The vertical short videos and horizontal short videos may be collectively referred to as videos. It can be understood that a picture set can be regarded as a special form of text and pictures. Therefore, in the embodiments of this application, text and pictures and picture sets may be collectively referred to as text and pictures.

[0125] It can be understood that the sample metadata information may include first auxiliary information, second auxiliary information, and third auxiliary information. The first auxiliary information is the information entered by the first user when uploading the target sample content. The second auxiliary information is obtained after the content classification model classifies the target sample content. The third auxiliary information is obtained after transcoding the target sample content. Optionally, when the server adds the sample metadata information and the target sample content to the content database to be distributed, the current timestamp also needs to be added to the content database as the storage time. It should be understood that the embodiments of this application do not limit the model type of the content classification model.

[0126] Among them, the first auxiliary information may include the first type of label information, theme information (i.e., title), time information (i.e., release time), abstract information, author information (i.e., account author), and source channel of the target sample content. The second auxiliary information may include the second type of label information and classification information of the target sample content obtained through classification by the content classification model. The third auxiliary information may include the size information of the target sample content (i.e., file size, for example, 3.2MB), cover image link, bitrate information (for example, 4818kbps), dimension information (i.e., specification, for example, 1280*720 pixels), and format information (for example, avi). It should be understood that the embodiments of this application do not limit the information types included in the first auxiliary information, second auxiliary information, and third auxiliary information.

[0127] Among them, the first type of label information may be the label information entered by the first user in the first client. The second type of label information may be the label information obtained after the content classification model identifies the target sample content. Optionally, the content database may also include the third type of label information of the target sample content. Here, the third type of label information may be the high-level abstract label information identified manually. Among them, the first type of label information, the second type of label information, and the third type of label information may be collectively referred to as label information.

[0128] Among them, the classification information stored in the content database may include primary classification, secondary classification, and tertiary classification, and the number of tag information may be one or more. For example, for a piece of information flow content about a mobile phone (e.g., mobile phone J, where mobile phone J can be a mobile phone brand or model), the primary classification can be technology, the secondary classification can be mobile phones, the tertiary classification can be domestic mobile phones, and the tag information can be mobile phone J, Snapdragon 855, etc.

[0129] It should be understood that the first user account can be the account information corresponding to a self-media or content production agency (e.g., a multi-channel network MCN, Multi-Channel Network). The first user corresponding to the first user account can upload target sample content through the content upload interface, and then the server can obtain the target sample content uploaded by the first user through the content upload interface through the up and down content interface service.

[0130] Among them, the target video data uploaded by the first user through the content upload interface can be PGC (Professional Generated Content, professional production content, expert production content, or Professionally Produced Content, abbreviated as PPC) content, UGC (User Generated Content, user-generated content, user original content), or PUGC (Professional User Generated Content) content of a self-media or content production agency. Among them, PGC content can generally refer to content personalization, diverse perspectives, democratic dissemination, and virtual social relations. Among them, UGC content emerged along with the Web2.0 concept characterized by advocating personalization. It is not a specific business but a new way for users to use the Internet, that is, from mainly downloading to both downloading and uploading equally. Among them, PUGC content is used to represent professional audio content in the form of UGC and the output of PGC. Among them, MCN is a product form of a multi-channel network that can combine PGC content and, with the strong support of capital, ensure the continuous output of content, and ultimately achieve stable commercial monetization.

[0131] Optionally, the content database and the statistical database can also be the same database. In this way, the server can add the received sample metadata information, target sample content, and post flow information to the same database.

[0132] Step S102, obtain an account level model for grading the first user account;

[0133] Among them, the account level model includes a prior level model and a posterior level model. Among them, the prior level model may include a first prior sub-model, a second prior sub-model, a third prior sub-model, and a fourth prior sub-model; the posterior account level model may include a first posterior sub-model, a second posterior sub-model, and a third posterior sub-model.

[0134] It can be understood that the first prior sub-model is used to analyze the influence of the first user account to obtain the account influence degree of the first user account; the second prior sub-model is used to analyze the stability of the first user account to obtain the content stability degree of the first user account; the third prior sub-model is used to analyze the verticality of the first user account to obtain the content verticality degree of the first user account; the fourth prior sub-model is used to analyze the matching degree of the first user account to obtain the content matching degree of the first user account.

[0135] It can be understood that the first posterior sub-model is used to analyze the contribution degree of the first user account to obtain the platform contribution degree of the first user account; the second posterior sub-model is used to analyze the interaction degree of the first user account to obtain the account interaction degree of the first user account; the third posterior sub-model is used to analyze the difference of the first user account to obtain the content difference degree of the first user account.

[0136] Step S103, obtain the account statistical information associated with the first user account from the statistical database, and obtain the content metadata information associated with the first user account from the content database, input the account statistical information and the content metadata information into the prior level model, and let the prior level model identify the account level of the first user account, and use the identified account level of the first user account as the prior level information of the first user account;

[0137] Specifically, the server can obtain the account statistical information associated with the first user account from the statistical database and obtain the content metadata information associated with the first user account from the content database. Among them, the content metadata information includes sample metadata information; the account statistical information includes post flow information. Further, the server can obtain the account influence parameter associated with the first prior sub-model, the content stability parameter associated with the second prior sub-model, the content verticality parameter associated with the third prior sub-model, and the content matching parameter associated with the fourth prior sub-model from the account statistical information and the content metadata information. Further, the server can input the account influence parameter into the first prior sub-model, and the first prior sub-model can perform influence analysis on the first user account based on the account influence parameter to obtain the account influence degree of the first user account. Further, the server can input the content stability parameter into the second prior sub-model, and the second prior sub-model can perform stability analysis on the first user account based on the content stability parameter to obtain the content stability degree of the first user account. Further, the server can input the content verticality parameter into the third prior sub-model, and the third prior sub-model can perform verticality analysis on the first user account based on the content verticality parameter to obtain the content verticality degree of the first user account. Further, the server can input the content matching parameter into the fourth prior sub-model, and the fourth prior sub-model can perform matching degree analysis on the first user account based on the content matching parameter to obtain the content matching degree of the first user account. Further, the server can perform prior data fusion on the account influence degree, the content stability degree, the content verticality degree, and the content matching degree to obtain the prior data fusion result of the first user account, and based on the prior data fusion result, use the identified account level of the first user account as the prior level information of the first user account.

[0138] Among them, it can be understood that the server can perform prior data fusion on the account influence degree, the content stability degree, the content verticality degree, and the content matching degree to obtain the prior data fusion result of the first user account (for example, the quantization score corresponding to the prior data fusion result). Further, the server can determine the prior account level of the first user account based on the prior data fusion result, use the prior account level as the account level of the identified first user account, and then use the identified account level of the first user account as the prior level information of the first user account.

[0139] Among them, the account influence parameters include a first account set and a second account set corresponding to the first user account; among them, the first account set is the account set that follows the first user account, and the second account set is the account set that unfollows the first user account. It should be understood that the specific process of the server performing influence analysis on the first user account can be described as follows: The server can obtain the number of followed accounts and the number of followed-by accounts corresponding to the first accounts in the first account set, and determine a first weight associated with the first accounts based on the number of followed accounts and the number of followed-by accounts. Further, the server can determine a first average weight corresponding to each first account based on the first weight associated with each first account in the first account set and the first timestamp when each first account follows the first user account. Further, the server can obtain the number of unfollowed accounts and the number of unfollowed-by accounts corresponding to the second accounts in the second account set, and determine a second weight associated with the second accounts based on the number of unfollowed accounts and the number of unfollowed-by accounts. Further, the server can determine a second average weight corresponding to each second account based on the second weight associated with each second account in the second account set and the second timestamp when each second account unfollows the first user account. Further, the server can determine the number of accounts in the first account as the auxiliary account number, and the first prior sub-model can obtain the account influence degree of the first user account based on the auxiliary account number, the first average weight corresponding to each first account, and the second average weight corresponding to each second account.

[0140] It can be understood that the number of accounts of the third account in the followed account set is the number of followed accounts, the number of accounts of the fourth account in the followed-by account set is the number of followed-by accounts, the number of accounts of the fifth account in the unfollowed account set is the number of unfollowed accounts, and the number of accounts of the sixth account in the unfollowed-by account set is the number of unfollowed-by accounts. Among them, the first account, the second account, the third account, the fourth account, the fifth account, and the sixth account are all accounts that meet the account screening conditions. Here, the account screening conditions can be accounts that have been viewed twice or more within a month, non-zombie accounts (i.e., active accounts), and non-reposting accounts (i.e., true fans). In other words, for example, the server can identify zombie accounts and reposting accounts among all the fan accounts that follow the first user account, remove the zombie accounts and reposting accounts from all the fan accounts, and use the accounts obtained after the removal as the first accounts.

[0141] Among them, the way for the server to perform influence analysis on the first user account and obtain the account influence degree of the first user account can be seen in the following formula (1):

[0142]

[0143] Among them, A iRepresents the total number of the latest fans (i.e., the number of auxiliary accounts) of the $i$-th account (e.g., the first user account); Represents the set of users who follow the $i$-th account (i.e., the first account set), Represents the set of users who unfollow the $i$-th account (i.e., the second account set); Represents the behavior date when user $u$ follows the $i$-th account (i.e., the first timestamp), Represents the behavior date when user $u$ unfollows the $i$-th account (i.e., the second timestamp), and $t_0$ represents the current date; Represents the weight when user $u$ follows the $i$-th account (i.e., the first weight), Represents the weight when user $u$ unfollows the $i$-th account (i.e., the second weight); $\delta$ and $\eta$ are parameters for smoothing effect. For example, $\delta = 1.0$ and $\eta = 10.0$; $\alpha$ is a control parameter. For example, $\alpha = 2.0$. Among them, Can represent the number of days between $t_0$ and When it is the same day between $t_0$ and it is equal to 0. Can represent the number of days between $t_0$ and When is the day before $t_0$, it is equal to 1.

[0144] Among them, the calculation process of the first weight can be seen in the following formula (2)

[0145]

[0146] Among them, is the number of accounts in the set of accounts followed by user $u$ (i.e., the number of followed accounts), $|P$ + $|$ is the number of accounts in the set of accounts that user $u$ is followed by (i.e., the number of accounts being followed).

[0147] Among them, the calculation process of the second weight can be seen in the following formula (3):

[0148]

[0149] Among them, is the number of accounts in the set of accounts unfollowed by user $u$ (i.e., the number of unfollowed accounts), $|P$ - $|$ is the number of accounts in the set of accounts that user $u$ is unfollowed by (i.e., the number of accounts being unfollowed).

[0150] Among them, it can be understood that the account influence degree can regard the accounts with more fans, more recent follows, fewer unfollows, more high-quality follows, and fewer low-quality follows as higher-level accounts. ​

[0151] Among them, the content stability parameter includes the number of samples of the sample content of the first user account within the first period. It should be understood that the specific process of the server performing stability analysis on the first user account can be described as follows: The server can perform mean processing on the number of samples to obtain the sample mean of the first user account within the first period. Further, the server can perform variance processing on the number of samples based on the sample mean to obtain the sample variance of the first user account within the first period. Further, the server can obtain the content stability degree of the first user account by the second prior sub-model based on the sample mean and the sample variance.

[0152] Optionally, the server can also divide the first period into multiple sub-periods, perform mean processing on the number of sub-samples in the multiple sub-periods to obtain the sub-sample mean of the first user account in the multiple sub-periods, and then perform an addition operation on the sub-sample means corresponding to the multiple sub-periods respectively to obtain the sample mean of the first user account within the first period. Further, the server can perform variance processing on the number of sub-samples based on the sub-sample mean to obtain the sub-sample variance of the first user account in the multiple sub-periods, and then perform an addition operation on the sub-sample variances corresponding to the multiple sub-periods respectively to obtain the sample variance of the first user account within the first period. For example, the first period can be in units of one month, and the sub-period can be in units of one week. Among them, the sub-periods can be not completely the same. For example, when the sub-period is in units of weeks, the sub-period can be 7 days, 8 days or 9 days.

[0153] Among them, the way for the server to perform stability analysis on the first user account and obtain the content stability degree of the first user account can be seen in the following formula (4):

[0154]

[0155] Among them, σ1 represents the mean of the number of posts per month (i.e., the sample mean), and α2 represents the variance of the number of posts per month (i.e., the sample variance); η represents the smoothing coefficient. For example, η = 10.0; α represents the control parameter of the mean, and β represents the control parameter of the variance. For example, α = 1, β = 2.

[0156] Among them, it can be understood that the content stability degree can regard the accounts that adhere to posting and have good recent posting performance as higher-level accounts, and this content stability degree can be used to characterize the stability of content posting and the recent activity.

[0157] Among them, the content vertical parameter includes the classification information of the sample content of the first user account in the first period. It should be understood that the specific process of the server analyzing the verticality of the first user account can be described as follows: The server can determine the number of posting vertical categories of the classification information, and based on the number of each vertical category corresponding to the vertical category information in the number of posting vertical categories and the number of posting vertical categories, determine the vertical category feature probability corresponding to each vertical category information in the first period. Further, the server can obtain the content verticality of the first user account by the third prior sub-model based on the vertical category feature probability corresponding to each vertical category information.

[0158] Among them, the way for the server to analyze the verticality of the first user account and obtain the content verticality of the first user account can be seen in the following formula (5):

[0159]

[0160] Among them, i represents the i-th posting vertical category (i.e., the vertical category information, the first-level classification of the posting content); n represents the total number of posting vertical categories (i.e., the number of posting vertical categories); P i represents the proportion of the i-th vertical category posting (i.e., the vertical category feature probability). For example, the number of sample contents can be 3, and the 3 sample contents respectively correspond to 3 first-level classifications. Among them, the first-level classifications of 2 sample contents can be life, and the first-level classification of 1 sample content can be technology. Then P i can be equal to 1 / 3 and 2 / 3.

[0161] Among them, it can be understood that the content verticality is a statistical feature. It can regard the account with more concentrated posting vertical categories in the account as a higher-level account. In other words, the content verticality can regard the account with more similar first-level classifications in the account as a higher-level account.

[0162] Among them, the content matching parameters include the tag information of the sample content of the first user account within the first period, the publishing timestamp of the sample content, the account profile of the first user account, and the account name of the first user account. It should be understood that the specific process of the server performing a matching degree analysis on the first user account can be described as follows: The server can perform word segmentation on the account profile and the account name to obtain the text word segmentation of the first user account, and perform string matching between the tag information and the text word segmentation to obtain a string matching result. Further, if the string matching result indicates that there is a text word segmentation in the text word segmentation that matches the tag information, the server can determine the number of text word segmentations that match the tag information as the matching quantity corresponding to the sample content. Further, if the string matching result indicates that there is no text word segmentation in the text word segmentation that matches the tag information, the server can determine the matching quantity corresponding to the sample content based on the absence of a text word segmentation that matches the tag information. Further, the server can determine the first weight information corresponding to each sample content according to the publishing timestamp. Further, the server can determine the number of tags of the tag information of each sample content, and the fourth prior sub-model can obtain the content matching degree of the first user account based on the number of tags corresponding to each sample content, the first weight information corresponding to each sample content, and the matching quantity corresponding to each sample content.

[0163] Among them, the way for the server to perform a matching degree analysis on the first user account and obtain the content matching degree of the first user account can be seen in the following formula (6):

[0164]

[0165] Among them, M represents the matching degree between the post and the account (i.e., the content matching degree), the larger the score, the less matching; n represents the number of posts of the account in one month (i.e., the number of samples of the sample content); i represents the i-th picture / text / video sorted in the order of publishing time; HitTags i represents the number of tags (i.e., the matching quantity) of the i-th picture / text / video that hits the account name (i.e., the account name) and the profile (i.e., the account profile), CntTags i represents the number of tags of all the tag information of the i-th picture / text / video; w i represents the time weight (i.e., the first weight information) of the i-th picture / text / video post, and the post closer to the current time has a greater weight.

[0166] Among them, it can be understood that the content matching degree can describe the degree to which the tag information of all posts of an account hits the account profile and the account name, and is used to describe the matching degree between the posted sample content (e.g., pictures / texts, videos) and the account.

[0167] For easy understanding, please refer to Figure 4, Figure 4 is a schematic flowchart of a prior level model provided by an embodiment of the present application. As Figure 4 shown, the account influence parameter 40a, the content smoothness parameter 40b, the content verticality parameter 40c, and the content matching parameter 40d can be parameters of an account (for example, the first user account) obtained from a content database and a statistical database. As Figure 4 shown, the first prior sub-model 41a, the second prior sub-model 41b, the third prior sub-model 41c, and the fourth prior sub-model 41d can be collectively referred to as a prior level model.

[0168] As Figure 4 shown, the account influence parameter 40a is input into the first prior sub-model 41a, and the first prior sub-model 41a performs an influence analysis on the first user account based on the account influence parameter 40a, and an account influence degree 42a can be obtained. The content smoothness parameter 40b is input into the second prior sub-model 41b, and the second prior sub-model 41b performs a smoothness analysis on the first user account based on the content smoothness parameter 40b, and a content smoothness degree 42b can be obtained. The content verticality parameter 40c is input into the third prior sub-model 41c, and the third prior sub-model 41c performs a verticality analysis on the first user account based on the content verticality parameter 40c, and a content verticality degree 42c can be obtained. The content matching parameter 40d is input into the fourth prior sub-model 41d, and the fourth prior sub-model 41d performs a matching degree analysis on the first user account based on the content matching parameter 40d, and a content matching degree 42d can be obtained.

[0169] As Figure 4 shown, by performing prior data fusion on the account influence degree 42a, the content smoothness degree 42b, the content verticality degree 42c, and the content matching degree 42d, a prior data fusion result 43a can be obtained. Based on this prior data fusion result 43a, prior level information 43b of the first user account can be obtained.

[0170] Among them, the process of prior data fusion can refer to the following formula (7):

[0171] S i =(1 + I i ) α *(C i + μ) β *(H i + π) χ *(1 / M + σ) γ (7)

[0172] Among them, S iIt represents the final level score (i.e., the prior data fusion result) of the account (i.e., the self-media account) i; α represents the control parameter of the influence of the self-media account, β represents the control parameter of the stability of the content published by the self-media account, χ represents the control parameter of the verticality of the content published by the self-media account, and γ represents the control parameter of the matching degree between the graphics, texts and videos published by the self-media account and the account. These four control parameters here can be understood as the relative value weights of these four important dimensions. For example, α = 2, β = 1, χ = 1, γ = 0.5. μ represents the smoothing coefficient of the content stability, π represents the smoothing coefficient of the content verticality, and σ represents the smoothing coefficient of the content matching degree. For example, μ = 0.04, π = 0.0001, σ = 0.0001.

[0173] It can be understood that by setting different thresholds for different levels, different accounts can be classified into different levels. At this time, the server can sort the prior data fusion results of all accounts, and then classify all accounts into different levels based on the prior data fusion results, and set different recommendation priorities for accounts of different levels. For example, when the levels can be divided into three categories, the three levels can be the first level (e.g., high quality), the second level (e.g., ordinary), and the third level (e.g., low quality). The recommendation priority of the first level is higher than that of the second level, and the recommendation priority of the second level is higher than that of the third level. For example, the server can determine the accounts with prior data fusion results in the top 1 - 300 as the accounts of the first level, the accounts with prior data fusion results in the range of 301 - 1000 as the accounts of the second level, and the accounts with prior data fusion results after 1000 as the accounts of the second level.

[0174] Step S104, add the target sample content to the content recommendation pool associated with the content database based on the prior level information; Step S105, before the target sample content is sent through the content recommendation pool, obtain the account contribution information associated with the first user account from the statistical database, input the account contribution information and the content metadata information into the posterior level model, and the posterior level model outputs the posterior level information of the first user account, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0175] Specifically, before the server distributes the target sample content through the content recommendation pool, it can obtain the account contribution information associated with the first user account from the statistical database. The account contribution information includes posting transaction information. Further, the server can obtain the platform contribution parameter associated with the first posterior sub-model, the account interaction parameter associated with the second posterior sub-model, and the content difference parameter associated with the third posterior sub-model from the account contribution information and the content metadata information. Further, the server can input the platform contribution parameter into the first posterior sub-model, and the first posterior sub-model can perform a contribution degree analysis on the first user account based on the platform contribution parameter to obtain the platform contribution degree of the first user account. Further, the server can input the account interaction parameter into the second posterior sub-model, and the second posterior sub-model can perform an interaction degree analysis on the first user account based on the account interaction parameter to obtain the account interaction degree of the first user account. Further, the server can input the content difference parameter into the third posterior sub-model, and the third posterior sub-model can perform a difference analysis on the first user account based on the content difference parameter to obtain the content difference degree of the first user account. Further, the server can perform posterior data fusion on the platform contribution degree, the account interaction degree, and the content difference degree to obtain the posterior data fusion result of the first user account. Based on the posterior data fusion result, the server can obtain the posterior level information of the first user account, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0176] Among them, it can be understood that the server can perform posterior data fusion on the platform contribution degree, the account interaction degree, and the content difference degree to obtain the posterior data fusion result of the first user account (for example, the quantization score corresponding to the posterior data fusion result). Further, the server can determine the posterior account level of the first user account based on the posterior data fusion result, use the posterior account level as the new account level of the identified first user account, and then use the new account level of the identified first user account as the posterior level information of the first user account. Further, the server can correct the account level indicated by the prior level information through the account level indicated by the posterior level information (i.e., the new account level).

[0177] Among them, the platform contribution parameters include the post content of the first user account and the viewing behavior corresponding to the post content; the viewing behavior includes the viewing duration and viewing time of the viewing user viewing the post content. It should be understood that the specific process of the server analyzing the contribution degree of the first user account can be described as follows: The server can obtain the post content within the validity period from the post content and determine the obtained post content as the effective post content. Further, the server can perform validity analysis on the viewing behavior corresponding to the effective post content to obtain the effective viewing behavior of the viewing user viewing the effective post content within the second period, and determine the number of viewing users corresponding to the effective viewing behavior as the effective viewing quantity. Further, the server can obtain the effective viewing duration of the effective post content within the second period from the viewing duration, and determine the second weight information corresponding to the effective viewing duration according to the effective viewing time corresponding to the effective viewing duration; the effective viewing time is obtained from the viewing time. Further, the server can obtain the consumption duration corresponding to the effective post content based on the effective viewing duration and the second weight information, and obtain the target consumption duration that meets the duration acquisition condition from the consumption duration. Further, the server can obtain the platform contribution degree of the first user account based on the effective viewing quantity and the target consumption duration by the first posterior sub-model.

[0178] Among them, the post content of different content types has different validity periods. For example, the validity period of text and pictures can be 7 days, and the validity period of videos can be 3 months. It should be understood that the embodiments of the present application do not limit the specific time of the validity period of the post content.

[0179] It can be understood that the viewing behavior corresponding to the effective post content can include effective viewing behavior and invalid viewing behavior. Here, the viewing behavior can include the viewing times, viewing speed, and viewing completion degree of the viewing user viewing the effective post content. Among them, the viewing times can represent the number of clicks of the first user on the effective post content. The viewing speed is determined by the viewing duration and the number of words of the viewing user viewing the effective post content. The viewing completion degree is the viewing completion situation of the viewing user for the effective post content within the viewing duration.

[0180] Among them, it can be understood that the server can obtain the threshold value associated with the viewing times, the speed threshold value associated with the viewing speed, and the completion threshold value associated with the viewing completion degree, and then based on the threshold value of times, the speed threshold value, and the completion threshold value, divide the viewing behavior of the effective post content into effective viewing behavior and invalid viewing behavior. For example, when the viewing times corresponding to the effective post content are greater than the threshold value of times, the viewing speed corresponding to the effective post content is less than the speed threshold value, and the viewing completion degree corresponding to the effective post content is greater than the completion threshold value, the viewing behavior of the effective post content is determined as the effective viewing behavior.

[0181] It can be understood that when obtaining the target consumption duration that meets the duration acquisition condition from the consumption duration, the server can remove a maximum value and a minimum value from the consumption duration, or remove multiple (e.g., 3) values that are relatively large from the mean of the consumption duration. Further, the server can obtain the median of the consumption duration from the consumption duration and use the obtained median as the target consumption duration that meets the duration acquisition condition. Optionally, the server can also perform a mean processing on the consumption duration to use the consumption duration obtained after the mean processing as the target consumption duration that meets the duration acquisition condition.

[0182] Among them, the way for the server to perform a contribution degree analysis on the first user account to obtain the platform contribution degree of the first user account can be referred to the following formula (8):

[0183] Q i =||S i,j ||3 (8)

[0184] Among them, the norm is a basic concept in mathematics, and it is often used to measure the length or size of each vector in a certain space (or matrix). The specific definition is as follows: x = [x1, x2, …, x n T , then Here, the 3-norm can be used, that is, P is equal to 3. Among them, S i,j =(S i,1 , S i,2 , …, S i,n ), and the specific calculation process of S i,j can be referred to the following formula (9):

[0185] S i,j =(η + distinct uv) α *(η + median(read duration)) β (9)

[0186] Among them, distinct uv represents the count of unique consumption content users (i.e., the effective viewing count) for the jth enabled content (i.e., the effective published content) of account i within one month; n represents the number of effective published contents; median(read duration) represents the median of the per-piece content consumption duration within one month for the jth enabled content (i.e., the effective published content) of account i (i.e., the target consumption duration); α and β respectively represent different weights corresponding to the effective viewing count and the target consumption duration. For example, α = 1 and β = 2; η represents a smoothing coefficient. For example, η = 10.0.

[0187] ​Among them, by setting the weight corresponding to the effective viewing quantity to be less than the weight of the target consumption duration (i.e., setting α less than β), it is possible to avoid having only popular accounts ranked at the front while some accounts with good quality are ranked at the back. Among them, it can be understood that the platform contribution degree can regard the accounts corresponding to the content with high popularity (i.e., a large number of effective readers) and high user immersion degree (i.e., a long average reading time) as higher-level accounts. In addition, the closer the time when the user views is to the current time, the greater the weight of the duration (i.e., the second weight information).

[0188] Among them, the second period includes the T i th day; the account interaction parameter includes the viewing information of the effective posted content on the T i th day, the first interaction information on the T i th day, and the second interaction information on the T i th day. It should be understood that the specific process of the server analyzing the interaction degree of the first user account can be described as follows: The server can perform weighted summation on the first interaction information and the second interaction information to obtain the interaction quantity corresponding to the T i th day, and based on the interaction quantity and the viewing information, obtain the average interaction quantity of the first user account on the T i th day. Further, the server can perform an accumulation process on the interaction quantities of each day of the first user account within the second period to obtain the total interaction quantity of the first user account within the second period. Further, the server can obtain the account interaction degree of the first user account based on the total interaction quantity and the average interaction quantity by the second posterior sub-model.

[0189] Among them, the way for the server to analyze the interaction degree of the first user account and obtain the account interaction degree of the first user account can be referred to the following formula (10):

[0190]

[0191] Among them, pv k represents the click-through rate of the graphic and text content on the kth day (i.e., the viewing information associated with the graphic and text content), vv k represents the playback volume of the video content on the kth day (i.e., the viewing information associated with the video content); zanCnt k represents the number of likes on the kth day (i.e., the first interaction quantity), shareCnt k represents the number of shares on the kth day (i.e., the second interaction quantity), λ represents the weight occupied by the number of shares. For example, λ = 3; cnt represents the total interaction quantity of the account's posts and enabled content in a month, where cnt = zanCnt k +shareCnt k *λ; δ and η represent smoothing coefficients. For example, δ = 1.0 and η = 10.0.

[0192] Among them, it can be understood that the account interaction degree can rank the accounts with more high - like / high - sharing - rate and high - confidence good content in the account at the forefront. Assuming that under the same exposure conditions, the like and share performance of good content will be better, that is, the like rate / share rate of good content is higher. Therefore, the quality score of aggregating content to the account will show stronger noise resistance.

[0193] Among them, the content difference parameter includes the tag information of the sample content of the first user account in the second period; the sample content includes sample content S i and sample content S i+1 ; sample content S i+1 is the next sample content of sample content S i . It should be understood that the specific process of the server performing differential analysis on the first user account can be described as: the server can determine the tag information of sample content S i as the first tag information, and determine the tag information of sample content S i+1 as the second tag information. Further, the server can perform an intersection process on the first tag information and the second tag information to obtain intersection tag information, and determine the number of tags corresponding to the intersection tag information as the first tag number. Further, the server can perform a union process on the first tag information and the second tag information to obtain union tag information, and determine the number of tags corresponding to the union tag information as the second tag number. Further, the server can determine the content similarity between sample content S i and sample content S i+1 based on the first tag number and the second tag number, and the third posterior sub - model can obtain the content difference degree of the first user account based on the content similarity.

[0194] Among them, the way for the server to perform differential analysis on the first user account to obtain the content difference degree of the first user account can be seen in the following formula (11):

[0195]

[0196] Among them, T represents the similarity score (i.e., content difference degree) of the post tag information (i.e., tag). The larger the score, the greater the difference between the posts; n represents the number of posts of the self - media account in a month, i represents the i - th picture - text / video sorted in the order of posting time; IntersectionSize i,i+1 represents the size of the tag intersection (i.e., the first tag number corresponding to the intersection tag information) between the i - th picture - text / video and the (i + 1) - th picture - text / video, and UnionSize i,i+1 represents the size of the tag union (i.e., the second tag number corresponding to the union tag information) between the i - th picture - text / video and the (i + 1) - th picture - text / video.

[0197] It can be understood that the content difference degree can sort the tag information of the posts of an account according to the posting time, and is used to depict the similarity between the tag information of two adjacent picture texts / videos, so as to describe the difference between the picture texts / videos.

[0198] For ease of understanding, please refer to Figure 5 , Figure 5 which is a schematic flowchart of a tag matching provided by an embodiment of the present application. As Figure 5 shown, the tag information 50a, tag information 50b,..., tag information 50n can be the tag information of the sample content. The tag information 50a can be the tag information of the sample content 1, the tag information 50b can be the tag information of the sample content 2,..., and the tag information 50n can be the tag information of the sample content N.

[0199] Among them, the tag information 50a can include tags B1, B2, and B3, the tag information 50b can include tags B5, B6, and B3, and the tag information 50n can include tags B3, B5, and B7. It should be understood that the embodiments of the present application do not limit the number of tags in the tag information 50a, tag information 50b,..., tag information 50n. Here, an example is given where the number of tags in the tag information 50a, tag information 50b, and tag information 50n is 3.

[0200] As Figure 5 shown, the server can obtain the tag information 50a and tag information 50b from the tag information 50a, tag information 50b,..., tag information 50n. At this time, if the sample content 1 can be S i , and the sample content 2 can be S i+1 , the tag information 50a can be the first tag information, and the tag information 50b can be the second tag information. Further, the server can perform an intersection process on the tag information 50a and tag information 50b, that is, perform an intersection process on the tags in the tag information 50a and tag information 50b to obtain the intersection tag information 500a; the server can perform a union process on the tag information 50a and tag information 50b, that is, perform a union process on the tags in the tag information 50a and tag information 50b to obtain the union tag information 500b.

[0201] Among them, the intersection tag information 500a can include the tag B3, and the union tag information 500b can include the tags B1, B2, B3, B5, and B6. Therefore, the first tag quantity can be equal to 1, and the second tag quantity can be equal to 5.

[0202] For ease of understanding, please refer to Figure 6 ,Figure 6 is a schematic flowchart of a posteriori ranking model provided by an embodiment of the present application. As Figure 6 shown, the platform contribution parameter 60a, the account interaction parameter 60b, and the content difference parameter 60c can be parameters of an account (e.g., the first user account) obtained from a content database and a statistical database. As Figure 4 shown, the first posterior sub-model 61a, the second posterior sub-model 61b, and the third posterior sub-model 61c can be collectively referred to as the posteriori ranking model.

[0203] As Figure 6 shown, input the platform contribution parameter 60a into the first posterior sub-model 61a. The first posterior sub-model 61a performs a contribution degree analysis on the first user account based on the platform contribution parameter 60a, and the platform contribution degree 62a can be obtained. Input the account interaction parameter 60b into the second posterior sub-model 61b. The second posterior sub-model 61b performs an interaction degree analysis on the first user account based on the account interaction parameter 60b, and the account interaction degree 62b can be obtained. Input the content difference parameter 60c into the third posterior sub-model 61c. The third posterior sub-model 61c performs a difference analysis on the first user account based on the content difference parameter 60c, and the content difference degree 62c can be obtained.

[0204] As Figure 6 shown, by performing posteriori data fusion on the platform contribution degree 62a, the account interaction degree 62b, and the content difference degree 62c, the posteriori data fusion result 63a can be obtained. Based on the posteriori data fusion result 63a, the posteriori ranking information 63b of the first user account can be obtained.

[0205] Among them, the process of posteriori data fusion can refer to the following formula (12):

[0206] S = (μ + Q i ) α *(π + P i ) β *(σ + 1 / T) γ (12)

[0207] Among them, S represents the final posteriori ranking score (i.e., the posteriori data fusion result) of an account (i.e., a self-media account); α represents the weight parameter of the platform contribution degree, β represents the weight parameter of the account interaction degree, γ represents the weight parameter of the content difference degree. For example, α = 0.5, β = 5, γ = 0.5; μ represents the smoothing coefficient of the platform contribution degree, π represents the smoothing coefficient of the account interaction degree, σ represents the smoothing coefficient of the content difference degree. For example, μ = 0.04, π = 0.08, σ = 0.0001.

[0208] It can be seen that the embodiments of the present application can quickly and accurately identify the prior level information and posterior level information of the first user account through the prior level model and the posterior level model with level recognition functions. The prior level information can be used to describe the prior account level of the first user account (i.e., the account level indicated by the prior level information), and the posterior level information can be used to describe the posterior account level of the first user account (i.e., the account level indicated by the posterior level information). The prior account level can be used to evaluate the account level of the first user account, and the posterior account level can be used to dynamically update the account level of the first user account. In other words, the posterior account level can be used to correct the prior account level. Based on this, the embodiments of the present application can improve the efficiency of level recognition and realize the dynamic adjustment of the account level based on the prior level model and the posterior level model, thereby improving the efficiency of level adjustment.

[0209] Further, please refer to Figure 7 , Figure 7 which is a schematic flowchart of an account level adjustment method provided by the embodiments of the present application. This method can be executed by the server, or by the first client, or jointly by the server and the first client. The server can be the server 20a in the corresponding implementation of the above Figure 2 , and the first client can be the application client in the corresponding embodiment of the above Figure 2 . For ease of understanding, this embodiment is described by taking the method being executed by the server as an example. Among them, the account level adjustment method may include the following steps S201 - step S210:

[0210] Step S201, when receiving the target sample content uploaded by the first client through the first user account, obtain the sample metadata information and the post flow information of the target sample content, add the sample metadata information and the target sample content to the content database to be distributed, and add the post flow information to the statistical database associated with the second client;

[0211] It can be understood that when the uplink and downlink content interface service receives the target sample content uploaded by the first user account, it can send the target sample content to the scheduling center service, and the target sample content is sent to the duplicate checking service through the scheduling center server, so that the duplicate checking service configures the target fingerprint feature for the target sample content. Further, the duplicate checking service can obtain the historical fingerprint features of the recommended content in the content database and determine the vector distance (e.g., Euclidean distance) between the target fingerprint feature and the historical fingerprint feature. Further, if the vector distance is greater than the vector threshold associated with the historical fingerprint feature, the duplicate checking service can determine that the target fingerprint feature meets the duplicate checking processing condition. Further, the duplicate checking service can determine that there is no recommended content in the content database that matches the target sample content, and then store the target sample content uploaded by the first user account in the content database.

[0212] Optionally, if the vector distance is less than the vector threshold, the duplicate checking service can determine that the target fingerprint feature does not meet the duplicate checking processing condition. Further, the duplicate checking service can determine that there is recommended content in the content database that matches the target sample content, and then delete the target sample content uploaded by the first user account. In this way, a copy of duplicate or similar data can be retained and enabled in the content database (i.e., the content pool), and then subsequent processing can be performed on this copy of data to reduce duplicate or similar data in the subsequent processing process, so as to effectively construct the data that needs to be processed in the system audit process and improve the efficiency of system audit.

[0213] It can be understood that when the target sample content is text and image content, the duplicate checking service can perform duplicate checking processing on the text data in the text and image content; when the target sample content is video content, the duplicate checking service can perform duplicate checking processing on the video data (i.e., extract video fingerprints to construct vectors) and audio data (i.e., extract audio fingerprints to construct vectors) in the video content. Optionally, the duplicate checking service can also perform duplicate checking processing on the cover image, title, etc. of the target sample content.

[0214] It can be understood that the duplicate checking service can use a model to configure the target fingerprint feature. For example, when configuring the fingerprint feature for the text data, the model here can be the simmhash and BERT (Bidirectional Encoder Representations from Transformers) network model. The simmhash can quickly calculate the digital fingerprint of the text content, and the BERT network model can quickly represent the digital fingerprint as the target fingerprint feature.

[0215] It can be understood that the dispatching center service can call the audit system to read the target sample content and the sample metadata information of the target sample content from the content database through the audit system. Among them, the sample metadata information may include first auxiliary information, second auxiliary information, and third auxiliary information, and the content included in the first auxiliary information, second auxiliary information, and third auxiliary information may refer to the description of step S101 in the corresponding embodiment above. Further, the audit system can perform a first content audit on the target sample content based on the sample metadata information to obtain an initial audit result. Further, if the initial audit result indicates that the target sample content is legal, the audit system can perform a second content audit on the target sample content to obtain a target audit result, and use the target audit result as the content audit result of the target sample content. Among them, the content of the audit can come from the acquisition by the web crawler from the public network and the active release of the self-media. Figure 3 The first content audit can be carried out manually to perform a round of preliminary filtering on whether the target sample content involves porn, gambling, drugs (i.e., porn, gambling, drugs) and politically sensitive characteristics, that is, to audit the legality of the target sample content.

[0216] It should be understood that since machine learning is not yet fully mature, it is not completely accurate to determine the label information and classification information of the target sample content completely through machine learning. It is necessary to manually audit the second auxiliary information on the basis of the content classification model obtaining the second auxiliary information, and through human-machine cooperation, improve the accuracy and efficiency of the annotation of the target sample content. It can be understood that the second content audit can be carried out manually to confirm the label information and classification information of the target sample content, that is, to audit the correctness of the meta-information (for example, the second auxiliary information) of the target sample content. In addition, the second audit can also be carried out manually to audit the content quality (such as lack of nutrition, incomplete) and safety and other quality problems that are difficult for machines to identify of the target sample content, that is, to audit the integrity of the target sample content.

[0217] Among them, it can be understood that the content audit result will be transmitted back to the content database for storage, and the content audit result is also an important basis for measuring the filtering effect of the content quality algorithm in the future. The content quality algorithm here can identify and mark the quality of the target sample content through machines, for example, advertising content, clickbait, and old news, etc.

[0218] Step S202, obtain an account level model for grading the first user account;

[0219] Among them, the account level model includes a prior level model (i.e., an account comprehensive quality model) and a posterior level model.

[0220]

[0221] ​Step S203, call the prior level model through the account grading service;

[0222] Among them, the account grading service can be called through the scheduling center service, and the prior level model for grading the first user account can be called through the account grading service.

[0223] Step S204, obtain the account statistical information associated with the first user account from the statistical database, and obtain the content metadata information associated with the first user account from the content database. Input the account statistical information and the content metadata information into the prior level model. The prior level model identifies the account level of the first user account, and uses the identified account level of the first user account as the prior level information of the first user account;

[0224] Among them, the content metadata information includes sample metadata information; the account statistical information includes post flow information.

[0225] Among them, the prior level model considers four key dimensions: (1) Account influence (i.e., account impact); (2) Stability of the content published by the account (i.e., content stability); (3) Verticality of the content published by the account (i.e., content verticality); (4) Degree of matching between the articles / videos published and the account (i.e., content matching). For the specific process of the server identifying the account level of the first user account through the prior level model, reference can be made to the description in step S103 above, which will not be elaborated here. It should be understood that the dimensions for the prior level model to determine the prior level information of the first user account include but are not limited to the above four key dimensions.

[0226] Step S205, call the audit system through the account grading service, conduct a system review of the prior level information through the audit system, and write the prior level information that passes the system review into the level database;

[0227] Among them, the system review can check whether there are abnormalities or deficiencies in the prior level information through configured rules (for example, the ranking of the head account level with clear institutional endorsement is very high) and accounts reported by manual feedback (for example, the whitelist provided by the operation). Through the combination of manual and machine, the rating and dynamic update of the first user account can be better and more objectively carried out to ensure the quality of the prior level information.

[0228] It can be understood that the embodiments of the present application can be used to grade user accounts on the information flow platform corresponding to the first client, and the user account here can be the first user account corresponding to the first client. Optionally, when the first user account is a newly acquired account and the new account does not have posting data, the server can directly call the audit system through the account grading service without identifying the account level of the first user account through the prior level model, and determine the account level of the new account through the audit system.

[0229] Step S206: Add the target sample content to the content recommendation pool associated with the content database based on the prior level information;

[0230] It can be understood that the server can determine the account level indicated by the prior level information. When the account level indicated by the prior level information is a higher level (for example, the first level), the target sample content uploaded by the first user account is preferentially added to the content recommendation pool based on the prior level information.

[0231] Step S207: Call the posterior level model through the account grading service;

[0232] Among them, the account grading service can be called through the scheduling center service, and the posterior level model for grading the first user account can be called through the account grading service.

[0233] Step S208: Before the target sample content is sent down through the content recommendation pool, obtain the account contribution information associated with the first user account from the statistical database, input the account contribution information and the content metadata information into the posterior level model, output the posterior level information of the first user account by the posterior level model, correct the account level indicated by the prior level information through the account level indicated by the posterior level information, and write the posterior level information into the level database;

[0234] It can be understood that after the server determines the posterior level information of the first user account, it can send down the target sample content through the content recommendation pool based on the posterior level information (that is, send the target sample content to the application client). For example, the server can send down the target sample content to the second client through the content recommendation pool. Here, the application client is taken as the second client for illustration. In this way, when the second client receives the target recommended content sent down by the server (the target recommended content here can include the target sample content uploaded by the above first user account), it can execute the following Step S209 and Step S210. Optionally, when the server determines the prior level information of the first user account, it can also directly send down the target sample content through the content recommendation pool based on the prior level information.

[0235] Among them, the second client can present at least one target recommended content (i.e., information flow content) to the second user (i.e., content receiver) in the form of an information flow (i.e., Feeds stream). Here, the Feeds stream is a data format, and the server can recommend the information flow content that the second user is interested in or the latest to the second client, so that the second client can display this information flow content in the form of a Feeds stream and quickly refresh this information flow content. Among them, the Feeds stream is usually sorted in the form of a timeline (i.e., Timeline), that is, the timeline is the most intuitive, original, and basic display form of the Feeds stream.

[0236] It should be understood that the aggregator can aggregate the above Feeds streams together. Here, the aggregator refers to software used to aggregate Feeds streams. For example, the aggregator can be software specifically used to subscribe to websites (different websites correspond to different servers), and the aggregator can also be called an RSS (Really Simple Syndication) reader, a feed reader, a news reader, etc.

[0237] It can be understood that in the application client, various different self-media accounts (e.g., the first user account) can create their own content, and users (e.g., the second user corresponding to the second user account) can subscribe to this content. Then, when the content is updated, the corresponding content is recommended to the second user in the way of B2C (Business to Consumer) downlink to be presented in the Feeds stream. Of course, the second user can also actively refresh the Feeds stream to obtain the latest content.

[0238] Optionally, the server can determine the account level indicated by the posterior level information, and then add the target sample content to the content recommendation pool associated with the content database based on the posterior level information. It can be understood that when the account level indicated by the posterior level information is a higher level (e.g., the first level), the server can preferentially add the target sample content uploaded by the first user account to the content recommendation pool based on the posterior level information.

[0239] Among them, it can be understood that the server can determine the recommendation priority of the target sample content in the content recommendation pool based on the account level indicated by the prior level information; the server can determine the recommendation priority of the target sample content in the content recommendation pool based on the account level indicated by the posterior level information. Among them, the self-media account level is a dynamically changing process, and the level of the account will change during its life cycle (especially for accounts other than non-top and non-authoritative large accounts without clear institutional endorsement). The prior level information and the posterior level information can be the level information of the first user account at different times.

[0240] It can be understood that when the account level indicated by the prior-level information is the same as the account level indicated by the posterior-level information, the server may not need to adjust the recommendation priority of the target sample content in the content recommendation pool. Optionally, when the account level indicated by the posterior-level information is higher than the account level indicated by the prior-level information, the server may increase the recommendation priority of the target sample content in the content recommendation pool. Optionally, when the account level indicated by the posterior-level information is lower than the account level indicated by the prior-level information, the server may decrease the recommendation priority of the target sample content in the content recommendation pool. In this way, it is possible to prevent the first user account, as a high-quality leading account, from wasting cold-start traffic.

[0241] Among them, the recommendation priority in the content recommendation pool can be used to adopt different recommendation strategies when distributing the target sample content. For example, low-quality content may not be recommended, and high-quality content may be displayed in front of ordinary content. Another example is to ensure that high-quality content meets a higher exposure volume and determine that ordinary content meets a lower exposure volume. It should be understood that the embodiments of the present application do not limit the specific recommendation strategies used in the recommendation strategy based on the account level.

[0242] It should be understood that the embodiments of the present application can enable the content of high-quality self-media accounts to be preferentially enabled and enter the content recommendation pool for distribution. By means of statistical modeling methods (i.e., the prior-level model and the posterior-level model), manual intervention can be reduced, and abnormal accounts can be detected in a timely manner and downgraded, improving the effectiveness of level adjustment. In addition, the embodiments of the present application can also concentrate traffic on real high-quality content creators, reduce the waste of high-quality traffic, and enable these high-quality content creators to receive the greatest incentives.

[0243] It can be understood that the server can store the level information of the user account in the level database to save a detailed record of the reasons for level changes for easy traceability. Therefore, the level database can be used to store the level information of user accounts on the information flow platform. For example, the level database can store the transformation process of the account level of the first user account, that is, the prior-level information and the posterior-level information of the first user account can be stored in the level database at the same time.

[0244] Among them, the posterior rank model considers three key dimensions: (1) the posterior performance of the account works and the contribution to the platform (i.e., the platform contribution degree); (2) the posterior performance of the account content and the interaction results, such as data on comments, likes, shares, etc. (i.e., the account interaction degree); (3) whether the differences between the published articles / videos are large, which is used to measure whether the account has mutated and blindly chased hot topics (i.e., the content difference degree). For the specific process of the server identifying the account level of the first user account through the posterior rank model, reference can be made to the description of step S105 above, which will not be elaborated here. It should be understood that the dimensions for the posterior rank model to determine the posterior rank information of the first user account include but are not limited to the above three key dimensions.

[0245] It can be understood that the prior rank model and the posterior rank model can be automatically run regularly to determine the rank information of the first user account. For example, the prior rank model runs once a month, and the posterior rank model can be automatically run every day.

[0246] Step S209, obtain the network environment, exposure data, and interaction behaviors of the second user account corresponding to the second client for the target sample content through the statistical reporting service;

[0247] It can be understood that the second client can communicate with the content distribution outlet service and the uplink and downlink content interface service, receive the target sample content sent by the server through the content recommendation pool through the index information of the target sample content (for example, the URL (Uniform Resource Locator) address), and then record the network environment, exposure data, and interaction behaviors corresponding to the trigger operation in response to the trigger operation performed by the second user corresponding to the second client for the target sample content.

[0248] Among them, the network environment is determined by the caching time of the second client caching the target sample content; the exposure data includes the target viewing time and target viewing duration of the second user account for the target sample content; the interaction behaviors include the first interaction operation of the second user account for the target sample content and the second interaction operation for the first user account.

[0249] It can be understood that the first interaction operation can include comment operations, forwarding operations, sharing operations, favorite operations, like operations, etc. performed by the second user through the second user account for the target sample content; the second interaction operation can include follow operations and unfollow (i.e., cancel follow) operations performed by the second user through the second user account for the first user account.

[0250] Step S210, determine the network environment, exposure data, and interaction behaviors as the end statistical information associated with the second client, and add the end statistical information to the statistical database.

[0251] It can be understood that the end statistical information corresponding to the second client can be used to determine the account level of the first user account, and some of the information in the end statistical information can belong to the above-mentioned account contribution information. It should be understood that the account information can include some of the information in the end statistical information and can also include some of the information in the end statistical information corresponding to other clients.

[0252] For ease of understanding, please refer to Figure 8 , Figure 8 which is a system flowchart for dynamically determining the level of a self-media account provided by an embodiment of the present application. As Figure 8 shown, steps S11 - S20 can be an execution path, steps S31 - S34 can be an execution path, and steps S41 - S49 can be an execution path. These 3 execution paths can be executed synchronously and crosswise.

[0253] As Figure 8 shown, in the execution path of steps S11 - S20, the content production end (for example, the first client corresponding to the first user) can execute step S11, and upload the target sample content to the uplink and downlink content interface service through the content upload interface. When the uplink and downlink content interface service obtains the target sample content uploaded by the first user through the first user account, it can execute step S12, obtain the sample metadata information (i.e., meta information) and the post flow information (i.e., account posts) of the target sample content (i.e., content), and write the target sample content and the sample metadata information into the content database.

[0254] Furthermore, the uplink and downlink content interface service can execute step S13, directly upload the target sample content to the scheduling center service for subsequent content processing and circulation. At this time, the target sample content can trigger the start of the scheduling process, and the target sample content enters the scheduling system. Therefore, the scheduling center service can execute step S14, call the duplicate checking service to perform duplicate checking on the target sample content, and write the result of the duplicate checking process into the content database. Among them, when the duplicate checking process passes, the scheduling center service can dispatch the manual review system to review this target sample content through step S15, and update the meta information in the content database based on the result of the duplicate checking process through step S17; when the duplicate checking process fails, the scheduling center service can delete this target sample content. Therefore, the target sample content that fails the duplicate checking process will not be reviewed by the manual review system.

[0255] Further, when the result indication of duplicate elimination processing indicates that the elimination processing has passed, the manual review system may execute step S16 to read the target sample content (i.e., the original content) and the sample metadata information from the content database, and perform the first content review (i.e., preliminary review) and the second content review (i.e., review) on the read information. It can be understood that when the first content review of the target sample content passes, the second content review can be performed on the target sample content. When the second content review of the target sample content passes, the target sample content can be used as the distributable content. When the first content review of the target sample content passes but the second content review fails, steps S15 and S17 can be executed to update the meta-information of the target sample content, and the target sample content that meets the integrity requirements can be used as the distributable content. Among them, the manual review system is usually a system developed based on a web database with complex operations.

[0256] Further, the server may execute step S18 to start content distribution. Based on the account level of the first user account, the target sample content is added to the content recommendation pool (i.e., the recommendation pool) associated with the content database through the recommendation distribution system. Further, the server may execute step S19 to enable content distribution through the content distribution export service, obtain at least one distributable content from the recommendation pool of the recommendation distribution system, and distribute at least one distributable content (i.e., the target recommended content) to the content consumption end (e.g., the second client corresponding to the second user) through step S20. Among them, the content distribution export service is usually a group of access services deployed nearby the users.

[0257] As Figure 8 shown, in the execution path of steps S31 - S34, when the upstream and downstream content interface service obtains the document flow information of the target sample content in the above step S11, it may report the account document (i.e., the document flow information) to the statistical reporting interface service by executing step S31, and the statistical reporting interface service executes step S34 to write the account document into the statistical database. At the same time, the content consumption end may execute step S32. Based on the index address of at least one distributable content (i.e., the access address entry and index information of the content) issued by the content distribution export service, the actual at least one distributable content is obtained from the content database through the upstream and downstream content interface service based on the index address. Further, when the content consumption end performs operations on at least one distributable content, the server may execute step S33 to upload the network environment, exposure data, and interaction behavior for at least one distributable content as the end statistical information to the statistical reporting interface service, and then the statistical reporting interface server executes step S34 to write the end statistical information into the statistical database.

[0258] As Figure 8As shown, in the execution path of steps S41 - S49, the account grading service can call the prior account level model (i.e., the prior level model), and through the prior account level model, execute steps S41 and S42 to obtain content metadata (i.e., content metadata information) and account statistical data (i.e., account statistical information) respectively, to obtain prior level information, and then return the prior level information to the account grading service through step S43. Further, the account grading service can execute step S44, send the prior level information to the manual review system, and conduct system review on the prior level information through the manual review system to execute step S45 and write the prior level information after system review into the level database (i.e., the self-media account level database).

[0259] Further, the account definition service can call the posterior account level model (i.e., the posterior level model), and through the posterior account level model, execute steps S46 and S47 to obtain content metadata (i.e., content metadata information) and account contribution data (i.e., account contribution information) respectively, to obtain posterior level information, and then return the posterior level information to the account grading service through step S48. Further, the account grading service can execute step S45 and write the posterior level information into the self-media account level database.

[0260] As Figure 8 shown, the scheduling center service can schedule the account grading service through step S49 to obtain the level information of the first user account. It should be understood that the latest level information of the first user account can be stored in the self-media account level database, and the latest level information here can be prior level information or posterior level information. Therefore, the level information obtained by the scheduling center service can be the prior level information or the posterior level information of the first user account.

[0261] It should be understood that after the target sample content is uploaded to the uplink and downlink content interface service at the content production end, the target sample content can enter the server through the uplink and downlink content interface service. Figure 8 The content distribution outlet service, uplink and downlink content interface service, statistical reporting interface service, scheduling center service, duplicate checking service, and account grading service shown can be server programs deployed on the server and specifically providing remote network services for the application client.

[0262] It can be seen that the embodiments of the present application can intelligently evaluate and correct the account level of the first user account through the account level model. The account level model here can include a prior level model and a posterior level model. The prior level model can be used to determine the prior level information of the first user account, and the posterior level model can be used to determine the posterior level information of the first user account. Among them, the prior level model can include a first prior sub-model, a second prior sub-model, a third prior sub-model, and a fourth prior sub-model. The first prior sub-model here can be used to determine the account influence degree of the first user account, the second prior sub-model here can be used to determine the account stability of the first user account, the third prior sub-model here can be used to determine the content verticality of the first user account, and the first prior sub-model here can be used to determine the content matching degree of the first user account. Among them, the posterior level model can include a first posterior sub-model, a second posterior sub-model, and a third posterior sub-model. The first posterior sub-model here can be used to determine the platform contribution degree of the first user account, the second posterior sub-model here can be used to determine the account interaction degree of the first user account, and the third posterior sub-model here can be used to determine the content difference degree of the first user account. Based on this, when the prior level information is obtained through evaluation in the embodiments of the present application, the account level indicated by the prior level information can be corrected by the account level indicated by the posterior level information, and the dynamic adjustment of the account level of the first user account can be realized, thereby improving the efficiency of level adjustment.

[0263] Further, please refer to Figure 9 , Figure 9 FIG. is a schematic structural diagram of an account level adjustment device provided by an embodiment of the present application. The account level adjustment device 1 may include: a content receiving module 10, a model obtaining module 20, a first recognition module 30, a content adding module 40, and a second recognition module 50; further, the account level adjustment device 1 may further include: a first calling module 60, a system review module 70, a second calling module 80, a data obtaining module 90, and a data adding module 100;

[0264] The content receiving module 10 is configured to, when receiving the target sample content uploaded by the first client through the first user account, obtain the sample metadata information and the post flow information of the target sample content, add the sample metadata information and the target sample content to the content database to be distributed with data, and add the post flow information to the statistical database;

[0265] The model obtaining module 20 is configured to obtain an account level model for grading the first user account; the account level model includes a prior level model and a posterior level model;

[0266] The first identification module 30 is used to obtain the account statistical information associated with the first user account from the statistical database, and obtain the content metadata information associated with the first user account from the content database, input the account statistical information and the content metadata information into the prior level model, and the prior level model identifies the account level of the first user account, and takes the identified account level of the first user account as the prior level information of the first user account; the content metadata information includes sample metadata information; the account statistical information includes the post flow information;

[0267] Among them, the prior level model includes a first prior sub-model, a second prior sub-model, a third prior sub-model and a fourth prior sub-model;

[0268] The first identification module 30 includes: a first acquisition unit 301, a first analysis unit 302, a second analysis unit 303, a third analysis unit 304, a fourth analysis unit 305, and a first fusion unit 306;

[0269] The first acquisition unit 301 is used to obtain the account influence parameter associated with the first prior sub-model, the content stability parameter associated with the second prior sub-model, the content vertical parameter associated with the third prior sub-model, and the content matching parameter associated with the fourth prior sub-model from the account statistical information and the content metadata information;

[0270] The first analysis unit 302 is used to input the account influence parameter into the first prior sub-model, and the first prior sub-model performs influence analysis on the first user account based on the account influence parameter to obtain the account influence degree of the first user account;

[0271] Among them, the account influence parameter includes a first account set and a second account set corresponding to the first user account; the first account set is the account set that follows the first user account; the second account set is the account set that unfollows the first user account;

[0272] The first analysis unit 302 includes: a first determination subunit 3021, a first averaging subunit 3022, a second determination subunit 3023, a second averaging subunit 3024, and a first output subunit 3025;

[0273] The first determination subunit 3021 is used to obtain the number of followed accounts and the number of followed-by accounts corresponding to the first account in the first account set, and determine the first weight associated with the first account based on the number of followed accounts and the number of followed-by accounts;

[0274] The first averaging subunit 3022 is used to determine the first average weight corresponding to each first account based on the first weight associated with each first account in the first account set and the first timestamp when each first account follows the first user account;

[0275] A second determination subunit 3023, configured to obtain the number of unfollowed accounts and the number of accounts being unfollowed corresponding to the second account in the second account set, and determine a second weight associated with the second account based on the number of unfollowed accounts and the number of accounts being unfollowed;

[0276] A second averaging subunit 3024, configured to determine a second average weight corresponding to each second account based on the second weight associated with each second account in the second account set and the second timestamp when each second account unfollows the first user account;

[0277] A first output subunit 3025, configured to determine the number of accounts of the first account as the number of auxiliary accounts, and obtain the account influence degree of the first user account by the first prior sub-model based on the number of auxiliary accounts, the first average weight corresponding to each first account, and the second average weight corresponding to each second account.

[0278] Wherein, for the specific implementation manners of the first determination subunit 3021, the first averaging subunit 3022, the second determination subunit 3023, the second averaging subunit 3024, and the first output subunit 3025, reference may be made to the description of step S103 in the corresponding embodiment above, which will not be elaborated here. Figure 3 The description of step S103 in the corresponding embodiment above will not be elaborated here.

[0279] A second analysis unit 303, configured to input the content smoothness parameter into the second prior sub-model, and perform smoothness analysis on the first user account by the second prior sub-model based on the content smoothness parameter to obtain the content smoothness degree of the first user account;

[0280] Wherein, the content smoothness parameter includes the number of samples of the sample content of the first user account within the first period;

[0281] The second analysis unit 303 includes: a mean processing subunit 3031, a variance processing subunit 3032, and a second output subunit 3033;

[0282] The mean processing subunit 3031 is configured to perform mean processing on the number of samples to obtain the sample mean of the first user account within the first period;

[0283] The variance processing subunit 3032 is configured to perform variance processing on the number of samples according to the sample mean to obtain the sample variance of the first user account within the first period;

[0284] The second output subunit 3033 is configured to obtain the content smoothness degree of the first user account by the second prior sub-model based on the sample mean and the sample variance.

[0285] Among them, for the specific implementation manners of the mean processing subunit 3031, the variance processing subunit 3032, and the second output subunit 3033, reference may be made to the description of step S103 in the corresponding embodiment above, which will not be elaborated here. Figure 3 The description of step S103 in the corresponding embodiment will not be elaborated here.

[0286] The third analysis unit 304 is configured to input the content vertical parameter into the third prior sub-model, and the third prior sub-model performs verticality analysis on the first user account based on the content vertical parameter to obtain the content verticality of the first user account.

[0287] Among them, the content vertical parameter includes the classification information of the sample content of the first user account within the first period.

[0288] The third analysis unit 304 includes: a probability determination subunit 3041 and a third output subunit 3042.

[0289] The probability determination subunit 3041 is configured to determine the number of posting vertical categories of the classification information, and based on the number of each vertical category corresponding to the vertical category information in the number of posting vertical categories and the number of posting vertical categories, determine the vertical category feature probability corresponding to each vertical category information within the first period.

[0290] The third output subunit 3042 is configured to obtain the content verticality of the first user account by the third prior sub-model based on the vertical category feature probability corresponding to each vertical category information.

[0291] Among them, for the specific implementation manners of the probability determination subunit 3041 and the third output subunit 3042, reference may be made to the description of step S103 in the corresponding embodiment above, which will not be elaborated here. Figure 3 The description of step S103 in the corresponding embodiment will not be elaborated here.

[0292] The fourth analysis unit 305 is configured to input the content matching parameter into the fourth prior sub-model, and the fourth prior sub-model performs matching degree analysis on the first user account based on the content matching parameter to obtain the content matching degree of the first user account.

[0293] Among them, the content matching parameter includes the tag information of the sample content of the first user account within the first period, the posting timestamp of the sample content, the account profile of the first user account, and the account name of the first user account.

[0294] The fourth analysis unit 305 includes: a word segmentation processing subunit 3051, a first matching subunit 3052, a second matching subunit 3053, a weight determination subunit 3054, and a fourth output subunit 3055.

[0295] The word segmentation processing subunit 3051 is used to perform word segmentation on the account profile and account name to obtain the text word segmentation of the first user account, and perform string matching between the tag information and the text word segmentation to obtain a string matching result;

[0296] The first matching subunit 3052 is used to, if the string matching result indicates that there is a text word segmentation in the text word segmentation that matches the tag information, determine the number of text word segmentations that match the tag information as the matching quantity corresponding to the sample content;

[0297] The second matching subunit 3053 is used to, if the string matching result indicates that there is no text word segmentation in the text word segmentation that matches the tag information, determine the matching quantity corresponding to the sample content based on the absence of text word segmentations that match the tag information;

[0298] The weight determination subunit 3054 is used to determine the first weight information corresponding to each sample content in the sample content according to the post timestamp;

[0299] The fourth output subunit 3055 is used to determine the number of tags of the tag information of each sample content, and the first user account's content matching degree is obtained by the fourth prior sub-model based on the number of tags corresponding to each sample content, the first weight information corresponding to each sample content, and the matching quantity corresponding to each sample content.

[0300] Among them, for the specific implementation manners of the word segmentation processing subunit 3051, the first matching subunit 3052, the second matching subunit 3053, the weight determination subunit 3054, and the fourth output subunit 3055, reference can be made to the description of step S103 in the corresponding embodiments above, and details will not be elaborated here. Figure 3 The description of step S103 in the corresponding embodiments above will not be repeated here.

[0301] The first fusion unit 306 is used to perform prior data fusion on the account influence degree, content smoothness, content verticality, and content matching degree to obtain a prior data fusion result of the first user account, and based on the prior data fusion result, use the identified account level of the first user account as the prior level information of the first user account.

[0302] Among them, for the specific implementation manners of the first acquisition unit 301, the first analysis unit 302, the second analysis unit 303, the third analysis unit 304, the fourth analysis unit 305, and the first fusion unit 306, reference can be made to the description of step S103 in the corresponding embodiments above, and details will not be elaborated here. Figure 3 The description of step S103 in the corresponding embodiments above will not be repeated here.

[0303] The content addition module 40 is used to add the target sample content to the content recommendation pool associated with the content database based on the prior level information;

[0304] The second recognition module 50 is used to obtain, before the target sample content is sent through the content recommendation pool, the account contribution information associated with the first user account from the statistical database, input the account contribution information and the content metadata information into the posterior level model, output the posterior level information of the first user account by the posterior level model, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0305] Among them, the posterior account level model includes a first posterior sub-model, a second posterior sub-model, and a third posterior sub-model;

[0306] The second recognition module 50 includes: a second acquisition unit 501, a fifth analysis unit 502, a sixth analysis unit 503, a seventh analysis unit 504, and a second fusion unit 505;

[0307] The second acquisition unit 501 is used to obtain, from the account contribution information and the content metadata information, the platform contribution parameter associated with the first posterior sub-model, the account interaction parameter associated with the second posterior sub-model, and the content difference parameter associated with the third posterior sub-model;

[0308] The fifth analysis unit 502 is used to input the platform contribution parameter into the first posterior sub-model, and the first posterior sub-model performs contribution degree analysis on the first user account based on the platform contribution parameter to obtain the platform contribution degree of the first user account;

[0309] Among them, the platform contribution parameter includes the published content of the first user account and the viewing behavior corresponding to the published content; the viewing behavior includes the viewing duration and viewing time of the viewing user viewing the published content;

[0310] The fifth analysis unit 502 includes: a content acquisition sub-unit 5021, a quantity determination sub-unit 5022, a data acquisition sub-unit 5023, a duration acquisition sub-unit 5024, and a fifth output sub-unit 5025;

[0311] The content acquisition sub-unit 5021 is used to obtain the published content within the validity period from the published content, and determine the obtained published content as the effective published content;

[0312] The quantity determination sub-unit 5022 is used to perform validity analysis on the viewing behavior corresponding to the effective published content, obtain the effective viewing behavior of the viewing user viewing the effective published content within the second period, and determine the number of viewing users corresponding to the effective viewing behavior as the effective viewing quantity;

[0313] A data acquisition subunit 5023, configured to obtain the effective viewing duration of the effective published content within the second period from the viewing duration, and determine the second weight information corresponding to the effective viewing duration according to the effective viewing time corresponding to the effective viewing duration; the effective viewing time is obtained from the viewing time.

[0314] A duration acquisition subunit 5024, configured to obtain the consumption duration corresponding to the effective published content based on the effective viewing duration and the second weight information, and obtain the target consumption duration that meets the duration acquisition condition from the consumption duration.

[0315] A fifth output subunit 5025, configured to obtain the platform contribution degree of the first user account by the first posterior sub-model based on the effective viewing quantity and the target consumption duration.

[0316] Among them, for the content acquisition subunit 5021, the quantity determination subunit 5022, the data acquisition subunit 5023, the duration acquisition subunit 5024, and the specific implementation manners of the fifth output subunit 5025, reference may be made to the description of step S105 in the corresponding embodiments above, which will not be elaborated here. Figure 3 The description corresponding to the embodiments will not be repeated here.

[0317] A sixth analysis unit 503, configured to input the account interaction parameter into the second posterior sub-model, and perform an interaction degree analysis on the first user account by the second posterior sub-model based on the account interaction parameter to obtain the account interaction degree of the first user account.

[0318] Among them, the second period includes the T i th day; the account interaction parameter includes the viewing information of the effective published content on the T i th day, the first interaction information on the T i th day, and the second interaction information on the T i th day.

[0319] The sixth analysis unit 503 includes: a weighted summation subunit 5031, an accumulation processing subunit 5032, and a sixth output subunit 5033.

[0320] The weighted summation subunit 5031 is configured to perform a weighted summation on the first interaction information and the second interaction information to obtain the interaction quantity corresponding to the T i th day, and obtain the average interaction quantity of the first user account on the T i th day based on the interaction quantity and the viewing information.

[0321] The accumulation processing subunit 5032 is configured to perform an accumulation process on the interaction quantities of each day of the first user account within the second period to obtain the total interaction quantity of the first user account within the second period.

[0322] The sixth output subunit 5033 is configured to obtain the account interaction degree of the first user account based on the total interaction volume and the average interaction quantity by the second posterior sub-model.

[0323] Among them, for the specific implementation manners of the weighted summation subunit 5031, the accumulation processing subunit 5032, and the sixth output subunit 5033, reference may be made to the description of step S105 in the corresponding embodiment above, which will not be elaborated here. Figure 3 The description of step S105 in the corresponding embodiment above will not be elaborated here.

[0324] The seventh analysis unit 504 is configured to input the content difference parameter into the third posterior sub-model, and perform a difference analysis on the first user account by the third posterior sub-model based on the content difference parameter to obtain the content difference degree of the first user account;

[0325] Among them, the content difference parameter includes the label information of the sample content of the first user account in the second period; the sample content includes sample content S i and sample content S i+1 ; sample content S i+1 is the next sample content of sample content S i ;

[0326] The seventh analysis unit 504 includes: a label determination subunit 5041, an intersection processing subunit 5042, a union processing subunit 5043, and a seventh output subunit 5044;

[0327] The label determination subunit 5041 is configured to determine the label information of sample content S i as the first label information, and determine the label information of sample content S i+1 as the second label information;

[0328] The intersection processing subunit 5042 is configured to perform an intersection process on the first label information and the second label information to obtain intersection label information, and determine the number of labels corresponding to the intersection label information as the first label quantity;

[0329] The union processing subunit 5043 is configured to perform a union process on the first label information and the second label information to obtain union label information, and determine the number of labels corresponding to the union label information as the second label quantity;

[0330] The seventh output subunit 5044 is configured to determine the content similarity between sample content S i and sample content S i+1 based on the first label quantity and the second label quantity, and obtain the content difference degree of the first user account by the third posterior sub-model based on the content similarity.

[0331] Among them, for the specific implementation manners of the label determination subunit 5041, the intersection processing subunit 5042, the union processing subunit 5043, and the seventh output subunit 5044, reference may be made to the description of step S105 in the corresponding embodiment above, which will not be elaborated here. Figure 3 The description of step S105 in the corresponding embodiment will not be elaborated here.

[0332] The second fusion unit 505 is configured to perform posterior data fusion on the platform contribution degree, the account interaction degree, and the content difference degree to obtain a posterior data fusion result of the first user account, and based on the posterior data fusion result, obtain posterior level information of the first user account, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0333] Among them, for the specific implementation manners of the second acquisition unit 501, the fifth analysis unit 502, the sixth analysis unit 503, the seventh analysis unit 504, and the second fusion unit 505, reference may be made to the description of step S105 in the corresponding embodiment above, which will not be elaborated here. Figure 3 The description of step S105 in the corresponding embodiment will not be elaborated here.

[0334] Optionally, the first invocation module 60 is configured to invoke a prior level model through an account rating service to obtain prior level information indicated by the prior level model;

[0335] The system review module 70 is configured to invoke an audit system through an account rating service, perform system review on the prior level information through the audit system, and write the prior level information that passes the system review into a level database;

[0336] The second invocation module 80 is configured to invoke a posterior level model through an account rating service to obtain posterior level information indicated by the posterior level model, and write the posterior level information into the level database.

[0337] Optionally, the data acquisition module 90 is configured to obtain the network environment, exposure data, and interaction behaviors of the second user account corresponding to the second client for the target sample content through a statistical reporting service; the network environment is determined by the caching time of the second client caching the target sample content; the exposure data includes the target viewing time and target viewing duration of the second user account for the target sample content; the interaction behaviors include the first interaction operation of the second user account for the target sample content and the second interaction operation for the first user account;

[0338] The data addition module 100 is configured to determine the network environment, exposure data, and interaction behaviors as end statistical information associated with the second client, and add the end statistical information to a statistical database.

[0339] Among them, for the specific implementation manners of the content receiving module 10, the model obtaining module 20, the first recognition module 30, the content adding module 40, and the second recognition module 50, reference may be made to the descriptions of steps S101 - S105 in the corresponding embodiments above, and details will not be elaborated here. Optionally, for the specific implementation manners of the first calling module 60, the system review module 70, the second calling module 80, the data obtaining module 90, and the data adding module 100, reference may be made to the descriptions of steps S201 - S210 in the corresponding embodiments above, and details will not be elaborated here. In addition, the description of the beneficial effects of adopting the same method will not be elaborated either. Figure 3 For the specific implementation manners of the content receiving module 10, the model obtaining module 20, the first recognition module 30, the content adding module 40, and the second recognition module 50, reference may be made to the descriptions of steps S101 - S105 in the corresponding embodiments above, and details will not be elaborated here. Optionally, for the specific implementation manners of the first calling module 60, the system review module 70, the second calling module 80, the data obtaining module 90, and the data adding module 100, reference may be made to the descriptions of steps S201 - S210 in the corresponding embodiments above, and details will not be elaborated here. Figure 7 For the specific implementation manners of the content receiving module 10, the model obtaining module 20, the first recognition module 30, the content adding module 40, and the second recognition module 50, reference may be made to the descriptions of steps S101 - S105 in the corresponding embodiments above, and details will not be elaborated here. Optionally, for the specific implementation manners of the first calling module 60, the system review module 70, the second calling module 80, the data obtaining module 90, and the data adding module 100, reference may be made to the descriptions of steps S201 - S210 in the corresponding embodiments above, and details will not be elaborated here. In addition, the description of the beneficial effects of adopting the same method will not be elaborated either.

[0340] Further, please refer to Figure 10 , Figure 10 which is a schematic structural diagram of a computer device provided by an embodiment of the present application. As Figure 10 shown, the computer device 1000 may include: a processor 1001, a network interface 1004, and a memory 1005. In addition, the computer device 1000 may further include: a user interface 1003 and at least one communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. Among them, the user interface 1003 may include a display screen (Display) and a keyboard (Keyboard). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. Optionally, the memory 1005 may further be at least one storage device located far from the aforementioned processor 1001. As Figure 10 shown, the memory 1005, as a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application program.

[0341] In the computer device 1000 as Figure 10 shown, the network interface 1004 can provide network communication functions; the user interface 1003 is mainly used to provide an input interface for users; and the processor 1001 can be used to call the device control application program stored in the memory 1005 to implement:

[0342] When receiving the target sample content uploaded by the first client through the first user account, obtain the sample metadata information and the document publishing flow information of the target sample content, add the sample metadata information and the target sample content to the content database to be distributed, and add the document publishing flow information to the statistical database;

[0343] Obtain an account level model for grading the first user account; the account level model includes a prior level model and a posterior level model;

[0344] Obtain the account statistical information associated with the first user account from the statistical database, and obtain the content metadata information associated with the first user account from the content database. Input the account statistical information and the content metadata information into the prior level model, and let the prior level model identify the account level of the first user account. Take the identified account level of the first user account as the prior level information of the first user account; the content metadata information includes the sample metadata information; the account statistical information includes the document publishing flow information;

[0345] Based on the prior level information, add the target sample content to the content recommendation pool associated with the content database;

[0346] Before distributing the target sample content through the content recommendation pool, obtain the account contribution information associated with the first user account from the statistical database. Input the account contribution information and the content metadata information into the posterior level model, and let the posterior level model output the posterior level information of the first user account. Correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

[0347] It should be understood that the computer device 1000 described in the embodiments of the present application can execute the description of the account level adjustment method in the corresponding embodiments mentioned above Figure 3 or Figure 7 The description of the corresponding embodiments, and can also execute the description of the account level adjustment device 1 in the corresponding embodiments mentioned above Figure 9 which will not be elaborated here. In addition, the description of the beneficial effects of adopting the same method will not be elaborated either.

[0348] In addition, it should be pointed out here that: the embodiments of the present application also provide a computer-readable storage medium, and the computer-readable storage medium stores the computer program executed by the account level adjustment device 1 mentioned above, and the computer program includes program instructions. When the processor executes the program instructions, it can execute the above Figure 3 or Figure 7The description of the account level adjustment method in the corresponding embodiments will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated. For the technical details not disclosed in the embodiments of the computer-readable storage medium involved in this application, please refer to the description of the method embodiments of this application.

[0349] In addition, it should be noted that: The embodiments of this application also provide a computer program product or a computer program. The computer program product or the computer program may include computer instructions, and the computer instructions may be stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor may execute the computer instructions, so that the computer device executes the Figure 3 or Figure 7 description of the account level adjustment method in the corresponding embodiments. Therefore, it will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated. For the technical details not disclosed in the embodiments of the computer program product or the computer program involved in this application, please refer to the description of the method embodiments of this application.

[0350] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above various methods. Among them, the storage medium can be a magnetic disk, an optical disc, a read-only memory (ROM), or a random access memory (RAM), etc.

[0351] The foregoing disclosure is only the preferred embodiments of this application. Of course, the scope of the rights of this application cannot be limited thereby. Therefore, equivalent changes made according to the claims of this application still fall within the scope covered by this application.

Claims

1. An account level adjustment method, characterized in that, Including: When receiving target sample content uploaded by a first client through a first user account, obtaining sample metadata information and document publishing flow information of the target sample content, adding the sample metadata information and the target sample content to a content database to be distributed, and adding the document publishing flow information to a statistical database; Obtaining an account grading model for grading the first user account; the account grading model includes a prior grading model and a posterior grading model; Obtaining account statistical information associated with the first user account from the statistical database, and obtaining content metadata information associated with the first user account from the content database, inputting the account statistical information and the content metadata information into the prior grading model, and having the prior grading model identify the account grade of the first user account, and taking the identified account grade of the first user account as the prior grade information of the first user account; The content metadata information includes the sample metadata information; The account statistical information includes the document publishing flow information; Based on the prior grade information, adding the target sample content to a content recommendation pool associated with the content database; Before distributing the target sample content through the content recommendation pool, obtaining account contribution information associated with the first user account from the statistical database, inputting the account contribution information and the content metadata information into the posterior grading model, having the posterior grading model output the posterior grade information of the first user account, and correcting the account grade indicated by the prior grade information through the account grade indicated by the posterior grade information.

2. The method according to claim 1, wherein The prior grading model includes a first prior sub-model, a second prior sub-model, a third prior sub-model, and a fourth prior sub-model; The step of inputting the account statistical information and the content metadata information into the prior grading model, having the prior grading model identify the account grade of the first user account, and taking the identified account grade of the first user account as the prior grade information of the first user account includes: Obtaining an account influence parameter associated with the first prior sub-model, a content stability parameter associated with the second prior sub-model, a content vertical parameter associated with the third prior sub-model, and a content matching parameter associated with the fourth prior sub-model from the account statistical information and the content metadata information; Inputting the account influence parameter into the first prior sub-model, and having the first prior sub-model perform an influence analysis on the first user account based on the account influence parameter to obtain the account influence degree of the first user account; Inputting the content stability parameter into the second prior sub-model, and having the second prior sub-model perform a stability analysis on the first user account based on the content stability parameter to obtain the content stability degree of the first user account; Input the content vertical parameter into the third prior sub-model, and the third prior sub-model performs a verticality analysis on the first user account based on the content vertical parameter to obtain the content verticality of the first user account; Input the content matching parameter into the fourth prior sub-model, and the fourth prior sub-model performs a matching degree analysis on the first user account based on the content matching parameter to obtain the content matching degree of the first user account; Perform prior data fusion on the account influence degree, the content stability, the content verticality, and the content matching degree to obtain the prior data fusion result of the first user account. Based on the prior data fusion result, use the identified account level of the first user account as the prior level information of the first user account.

3. The method according to claim 2, characterized in that, The account influence parameter includes a first account set and a second account set corresponding to the first user account; the first account set is the set of accounts that follow the first user account; the second account set is the set of accounts that unfollow the first user account; The step of inputting the account influence parameter into the first prior sub-model, and the first prior sub-model performs an influence analysis on the first user account based on the account influence parameter to obtain the account influence degree of the first user account, includes: Obtain the number of followed accounts and the number of followed-by accounts corresponding to the first account in the first account set, and determine the first weight associated with the first account based on the number of followed accounts and the number of followed-by accounts; Based on the first weight associated with each first account in the first account set and the first timestamp when each first account follows the first user account, determine the first average weight corresponding to each first account; Obtain the number of unfollowed accounts and the number of unfollowed-by accounts corresponding to the second account in the second account set, and determine the second weight associated with the second account based on the number of unfollowed accounts and the number of unfollowed-by accounts; Based on the second weight associated with each second account in the second account set and the second timestamp when each second account unfollows the first user account, determine the second average weight corresponding to each second account; Determine the number of accounts in the first account as the auxiliary account number, and the first prior sub-model obtains the account influence degree of the first user account based on the auxiliary account number, the first average weight corresponding to each first account, and the second average weight corresponding to each second account.

4. The method according to claim 2, wherein The content stability parameter includes the number of sample contents of the first user account within the first period; The step of inputting the content stability parameter into the second prior sub-model, and the second prior sub-model performs a stability analysis on the first user account based on the content stability parameter to obtain the content stability of the first user account, includes: Perform a mean processing on the number of samples to obtain the sample mean of the first user account within the first period; Perform variance processing on the sample quantity according to the sample mean to obtain the sample variance of the first user account within the first period; Based on the sample mean and the sample variance, the second prior sub-model obtains the content smoothness of the first user account.

5. The method according to claim 2, wherein The content vertical parameter includes the classification information of the sample content of the first user account within the first period; Inputting the content vertical parameter into the third prior sub-model, and the third prior sub-model performs verticality analysis on the first user account based on the content vertical parameter to obtain the content verticality of the first user account, including: Determine the number of posting vertical categories of the classification information, and based on the number of each vertical category corresponding to the vertical category information in the number of posting vertical categories and the number of posting vertical categories, determine the vertical category feature probability corresponding to each vertical category information within the first period; Based on the vertical category feature probability corresponding to each vertical category information, the third prior sub-model obtains the content verticality of the first user account.

6. The method according to claim 2, characterized in that The content matching parameter includes the label information of the sample content of the first user account within the first period, the posting timestamp of the sample content, the account profile of the first user account, and the account name of the first user account; Inputting the content matching parameter into the fourth prior sub-model, and the fourth prior sub-model performs matching degree analysis on the first user account based on the content matching parameter to obtain the content matching degree of the first user account, including: Perform word segmentation on the account profile and the account name to obtain the text word segmentation of the first user account, and perform string matching between the label information and the text word segmentation to obtain a string matching result; If the string matching result indicates that there is a text word segmentation in the text word segmentation that matches the label information, then determine the number of text word segmentations that match the label information as the matching quantity corresponding to the sample content; If the string matching result indicates that there is no text word segmentation in the text word segmentation that matches the label information, then determine the matching quantity corresponding to the sample content based on the absence of text word segmentations that match the label information; According to the posting timestamp, determine the first weight information corresponding to each sample content in the sample content; Determine the number of labels of the label information of each sample content, and the fourth prior sub-model obtains the content matching degree of the first user account based on the number of labels corresponding to each sample content, the first weight information corresponding to each sample content, and the matching quantity corresponding to each sample content.

7. The method according to claim 1, characterized in that, The posterior level model includes a first posterior sub-model, a second posterior sub-model, and a third posterior sub-model; Inputting the account contribution information and the content metadata information into the posterior level model, and the posterior level model outputs the posterior level information of the first user account, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information, including: Obtain a platform contribution parameter associated with the first posterior sub-model, an account interaction parameter associated with the second posterior sub-model, and a content difference parameter associated with the third posterior sub-model from the account contribution information and the content metadata information; Input the platform contribution parameter into the first posterior sub-model, and have the first posterior sub-model perform a contribution degree analysis on the first user account based on the platform contribution parameter to obtain the platform contribution degree of the first user account; Input the account interaction parameter into the second posterior sub-model, and have the second posterior sub-model perform an interaction degree analysis on the first user account based on the account interaction parameter to obtain the account interaction degree of the first user account; Input the content difference parameter into the third posterior sub-model, and have the third posterior sub-model perform a difference analysis on the first user account based on the content difference parameter to obtain the content difference degree of the first user account; Perform posterior data fusion on the platform contribution degree, the account interaction degree, and the content difference degree to obtain a posterior data fusion result of the first user account. Based on the posterior data fusion result, obtain the posterior level information of the first user account, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

8. The method according to claim 7, wherein The platform contribution parameter includes the posted content of the first user account and the viewing behavior corresponding to the posted content; the viewing behavior includes the viewing duration and viewing time of the viewing user viewing the posted content; The step of inputting the platform contribution parameter into the first posterior sub-model, and having the first posterior sub-model perform a contribution degree analysis on the first user account based on the platform contribution parameter to obtain the platform contribution degree of the first user account includes: Obtain the posted content within the validity period from the posted content, and determine the obtained posted content as the valid posted content; Perform validity analysis on the viewing behavior corresponding to the valid posted content to obtain the valid viewing behavior of the viewing user viewing the valid posted content within the second period, and determine the number of viewing users corresponding to the valid viewing behavior as the valid viewing quantity; Obtain the valid viewing duration of the valid posted content within the second period from the viewing duration, and determine the second weight information corresponding to the valid viewing duration according to the valid viewing time corresponding to the valid viewing duration; the valid viewing time is obtained from the viewing time; Based on the valid viewing duration and the second weight information, obtain the consumption duration corresponding to the valid posted content, and obtain the target consumption duration that meets the duration acquisition condition from the consumption duration; Have the first posterior sub-model obtain the platform contribution degree of the first user account based on the valid viewing quantity and the target consumption duration.

9. The method according to claim 8, wherein The second period includes the T i th day; the account interaction parameter includes the viewing information of the effective posted content on the T i th day, the first interaction information on the T i th day, and the second interaction information on the T i th day; The step of inputting the account interaction parameter into the second posterior sub-model, and having the second posterior sub-model perform an interaction degree analysis on the first user account based on the account interaction parameter to obtain the account interaction degree of the first user account includes: Perform a weighted sum of the first interaction information and the second interaction information to obtain the interaction quantity corresponding to the T i th day, and based on the interaction quantity and the viewing information, obtain the average interaction quantity of the first user account on the T i th day; Accumulate the interaction quantities of the first user account for each day within the second period to obtain the total interaction quantity of the first user account within the second period; Based on the total interaction quantity and the average interaction quantity, the second posterior sub-model obtains the account interaction degree of the first user account.

10. The method according to claim 7, characterized in that, The content difference parameter includes the label information of the sample content of the first user account in the second period; the sample content includes sample content S i and sample content S i+1 ; the sample content S i+1 is the next sample content of the sample content S i ; The step of inputting the content difference parameter into the third posterior sub-model, and the third posterior sub-model performs a difference analysis on the first user account based on the content difference parameter to obtain the content difference degree of the first user account includes: Determine the tag information of the sample content S i as the first tag information, and determine the tag information of the sample content S i+1 as the second tag information; Perform an intersection process on the first tag information and the second tag information to obtain intersection tag information, and determine the number of tags corresponding to the intersection tag information as the first tag number; Perform a union process on the first tag information and the second tag information to obtain union tag information, and determine the number of tags corresponding to the union tag information as the second tag number; Determine the sample content S based on the first tag quantity and the second tag quantity i and the sample content S i+1 The content similarity degree. The third posterior sub-model obtains the content difference degree of the first user account based on the content similarity degree 11. The method according to claim 1, wherein The method further includes: Invoke the prior level model through the account rating service to obtain the prior level information indicated by the prior level model; Invoke the audit system through the account rating service, and the audit system performs a system review on the prior level information, and write the prior level information that passes the system review into the level database; Invoke the posterior level model through the account rating service to obtain the posterior level information indicated by the posterior level model, and write the posterior level information into the level database.

12. The method according to claim 1, characterized in that, The method further includes: Obtain the network environment, exposure data, and interaction behaviors of the second user account corresponding to the second client for the target sample content through the statistical reporting service; the network environment is determined by the caching time of the second client caching the target sample content; the exposure data includes the target viewing time and target viewing duration of the second user account for the target sample content; the interaction behaviors include the first interaction operation of the second user account for the target sample content and the second interaction operation for the first user account; Determine the network environment, the exposure data, and the interaction behaviors as the terminal statistical information associated with the second client, and add the terminal statistical information to the statistical database.

13. An account level adjustment device, characterized in that, Includes: A content receiving module, configured to, when receiving the target sample content uploaded by the first client through the first user account, obtain the sample metadata information and the posting flow information of the target sample content, add the sample metadata information and the target sample content to the content database to be distributed, and add the posting flow information to the statistical database; A model acquisition module, configured to acquire an account level model for rating the first user account; the account level model includes a prior level model and a posterior level model; A first recognition module, configured to obtain account statistical information associated with the first user account from the statistical database, and obtain content metadata information associated with the first user account from the content database, input the account statistical information and the content metadata information into the prior level model, and have the prior level model recognize the account level of the first user account, and use the recognized account level of the first user account as the prior level information of the first user account; The content metadata information includes the sample metadata information; The account statistical information includes the post publishing stream information; A content addition module, configured to add the target sample content to a content recommendation pool associated with the content database based on the prior level information; A second recognition module, configured to, before the target sample content is sent down through the content recommendation pool, obtain account contribution information associated with the first user account from the statistical database, input the account contribution information and the content metadata information into the posterior level model, have the posterior level model output the posterior level information of the first user account, and correct the account level indicated by the prior level information through the account level indicated by the posterior level information.

14. A computer device, characterized in that, Comprising: A processor and a memory; The processor is connected to the memory, wherein the memory is used to store a computer program, and the processor is used to call the computer program to cause the computer device to execute the method according to any one of claims 1-12.

15. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, and the computer program is suitable for being loaded and executed by a processor to cause a computer device having the processor to execute the method according to any one of claims 1-12.

Images