Service video recommendation method and device, service video display method and device and computer equipment

Abstract

The invention discloses a service video recommendation method and device, a service video display method and device and computer equipment, and belongs to the technical field of computers. Under the condition that the video recommendation request of the service account is received, on one hand, the service behavior characteristics of the service account are extracted, on the other hand, the respective video attribute characteristics of the plurality of service videos are extracted, so that the video recommendation request is obtained by utilizing the service behavior characteristics and each video attribute characteristic; according to the embodiment of the invention, personalized service video recommendation can be performed for the current service account, the target video of which the matching degree with the service account meets the recommendation condition is determined from the plurality of service videos, and the target video is recommended to the service account, so that the target video and the service account are ensured to be relatively adaptive, and the user experience is improved. Therefore, the target video accords with the watching intention of the service account as much as possible, and the recommendation accuracy of the service video is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a method and device for recommending and displaying business videos, as well as a computer device. Background Art

[0002] With the development of computer technology, users can use terminals to start game sessions anytime and anywhere. Among them, Multiplayer Online Battle Arena (MOBA) games are a relatively popular type of game. Currently, users can enter the game forum through the forum control in the business client of MOBA games to view pre-configured business videos, including character teaching videos, equipment strategy videos, gameplay explanation videos, etc.

[0003] Since there are a large number of virtual characters in MOBA games, users are more likely to be uninterested in the pre-configured business videos. Therefore, the matching degree between the pre-configured business videos and the users' viewing intentions is poor, and the recommendation accuracy of business videos is low. Summary of the Invention

[0004] Embodiments of this application provide a method and device for recommending and displaying business videos, as well as a computer device, which can improve the recommendation accuracy of business videos. The technical solution is as follows:

[0005] On the one hand, a method for recommending business videos is provided. The method includes:

[0006] In response to a video recommendation request of a business account, obtain the business behavior information of the business account, where the business behavior information is the information of the business behavior executed by the business account in the business client;

[0007] Based on the business behavior information, extract the business behavior characteristics of the business account, where the business behavior characteristics represent the characteristics of the business behavior executed by the business account in the business client;

[0008] Based on the video attribute information of multiple business videos, extract the video attribute characteristics of each of the multiple business videos, where the video attribute characteristics represent the characteristics of the video attributes of the business videos;

[0009] Based on the business behavior characteristics and the video attribute characteristics, recommend target videos for the business account from the multiple business videos, where the matching degree between the target videos and the business account meets the recommendation conditions.

[0010] On the other hand, a method for displaying business videos is provided. The method includes:

[0011] When a business video is included in the interaction interface of a business client, a video recommendation request for a business account is sent. The video recommendation request is used to obtain a target video recommended for the business account. The business account is the account logged in to the business client, and the matching degree between the target video and the business account meets the recommendation conditions;

[0012] Receive video recommendation information returned based on the video recommendation request. The video recommendation information indicates the target video;

[0013] Pull the target video based on the video recommendation information;

[0014] Display the target video in the interaction interface.

[0015] On the one hand, a recommendation device for business videos is provided. The device includes:

[0016] An acquisition module, configured to obtain business behavior information of the business account in response to a video recommendation request of the business account. The business behavior information is information about business behaviors performed by the business account in the business client;

[0017] A first extraction module, configured to extract business behavior characteristics of the business account based on the business behavior information. The business behavior characteristics represent the characteristics of business behaviors performed by the business account in the business client;

[0018] A second extraction module, configured to extract video attribute characteristics of each of the multiple business videos based on video attribute information of the multiple business videos. The video attribute characteristics represent the characteristics of video attributes of the business videos;

[0019] A recommendation module, configured to recommend a target video for the business account from the multiple business videos based on the business behavior characteristics and the video attribute characteristics. The matching degree between the target video and the business account meets the recommendation conditions.

[0020] In some embodiments, the recommendation module includes:

[0021] An input unit, configured to input the business behavior characteristics into a video recommendation model. The video recommendation model is used to predict a video to be recommended for the business account based on business behavior characteristics of the business account;

[0022] A determination unit, configured to determine the target video to be recommended from the multiple business videos through the video recommendation model based on the video attribute characteristics of each of the multiple business videos;

[0023] A recommendation unit, configured to recommend the target video to the business account.

[0024] In some embodiments, the determining unit is configured to:

[0025] Based on the service behavior characteristics and the video attribute characteristics of each service video, determine the matching probability between the service account and each service video, where the matching probability indicates the degree of matching between the service account and the service video;

[0026] Based on the matching probability between the service account and each service video, determine the target video from the multiple service videos.

[0027] In some embodiments, the recommending unit is configured to:

[0028] Send video recommendation information of the target video to the terminal logged in by the service account, so that the terminal pulls the target video based on the video recommendation information; or, send the target video to the terminal.

[0029] In some embodiments, the first extraction module is configured to:

[0030] Input the service behavior information into a behavior feature extraction model, and perform feature extraction on the service behavior information through the behavior feature extraction model to obtain the service behavior characteristics, where the behavior feature extraction model is used to extract the service behavior characteristics of a service account.

[0031] In some embodiments, the apparatus further includes:

[0032] A query module, configured to query a candidate video list associated with the virtual character based on the character identifier in the case where the character identifier of the virtual character is carried in the video recommendation request;

[0033] A determination module, configured to determine the multiple service videos from the multiple candidate videos included in the candidate video list.

[0034] In some embodiments, the determination module is configured to:

[0035] Create a video pulling timing task for the virtual character, where the video pulling timing task is used to pull the candidate video list associated with the virtual character at regular intervals;

[0036] Based on the video pulling timing task, pull the candidate video list at intervals of a target duration, where the candidate video list contains multiple candidate videos associated with the virtual character;

[0037] Determine the multiple candidate videos that have passed the review from the candidate video list as the multiple service videos.

[0038] In some embodiments, the second extraction module is configured to:

[0039] Input the video attribute information of each business video into a video feature extraction model. Through the video feature extraction model, perform feature extraction on the video attribute information to obtain the video attribute features of the business video. The video feature extraction model is used to extract the video attribute features of business videos.

[0040] In some embodiments, the business behavior information includes at least one of the operation behavior information of the business account or the historical video records. The operation behavior information indicates the accumulated game behavior information of the business account in the virtual game. The historical video records indicate the interaction behavior information of the business account for the recommended historical videos.

[0041] In some embodiments, the video attribute information includes at least one of the publisher account information, video popularity information, video details information, or associated video information of the business video.

[0042] On the one hand, a display device for business videos is provided. The device includes:

[0043] A sending module, configured to send a video recommendation request of the business account when the interactive interface of the business client includes a business video. The video recommendation request is used to obtain a target video recommended for the business account. The business account is the account logged in to the business client, and the matching degree between the target video and the business account meets the recommendation conditions.

[0044] A receiving module, configured to receive video recommendation information returned based on the video recommendation request. The video recommendation information indicates the target video.

[0045] A pulling module, configured to pull the target video based on the video recommendation information.

[0046] A display module, configured to display the target video in the interactive interface.

[0047] In some embodiments, the interactive interface includes at least one of the following:

[0048] The function interface of the virtual character, which is used to introduce the virtual character or configure the equipment of the virtual character.

[0049] The preparation interface before the start of the practice mode or the virtual environment after the start.

[0050] The notification interface of the client version, which is used to prompt the updated content of the client version.

[0051] In some embodiments, when the interactive interface is a function interface of the virtual character, the role identifier of the virtual character is carried in the video recommendation request; or, when the interactive interface is a preparation interface before the start of the practice mode or a virtual environment after the start, the role identifier of the virtual character controlled by the service account is carried in the video recommendation request.

[0052] On the one hand, a computer device is provided. The computer device includes one or more processors and one or more memories. At least one computer program is stored in the one or more memories. The at least one computer program is loaded and executed by the one or more processors to implement the recommendation method or display method of business videos in any of the above possible implementation manners.

[0053] On the one hand, a storage medium is provided. At least one computer program is stored in the storage medium. The at least one computer program is loaded and executed by a processor to implement the recommendation method or display method of business videos in any of the above possible implementation manners.

[0054] On the one hand, a computer program product or a computer program is provided. The computer program product or the computer program includes one or more program codes. The one or more program codes are stored in a computer-readable storage medium. One or more processors of a computer device can read the one or more program codes from the computer-readable storage medium. The one or more processors execute the one or more program codes, so that the computer device can execute the recommendation method or display method of business videos in any of the above possible implementation manners.

[0055] The beneficial effects brought by the technical solutions provided in the embodiments of the present application at least include:

[0056] When receiving a video recommendation request of a service account, on the one hand, the business behavior characteristics of the service account are extracted, and on the other hand, the video attribute characteristics of each of multiple business videos are extracted. Therefore, by using the business behavior characteristics and each video attribute characteristic, personalized business video recommendations can be made for the current service account, a target video that meets the recommendation conditions in terms of the matching degree with the service account is determined from multiple business videos, and the target video is recommended to the service account, thereby ensuring that the target video is more suitable for the service account, making the target video as much as possible meet the viewing intention of the service account, and improving the recommendation accuracy of business videos. Description of the Drawings

[0057] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. 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.

[0058] Figure 1 It is a schematic diagram of the implementation environment of a business video recommendation method provided by an embodiment of the present application;

[0059] Figure 2 It is a logical architecture diagram of a video recommendation system provided by an embodiment of the present application;

[0060] Figure 3 It is a flowchart of a business video recommendation method provided by an embodiment of the present application;

[0061] Figure 4 It is a schematic diagram of the principle of a game video recommendation provided by an embodiment of the present application;

[0062] Figure 5 It is a flowchart of a business video recommendation method provided by an embodiment of the present application;

[0063] Figure 6 It is a flowchart of video management and review provided by an embodiment of the present application;

[0064] Figure 7 It is a schematic diagram of the principle of a business video screening method provided by an embodiment of the present application;

[0065] Figure 8 It is a development flowchart of a video recommendation model provided by an embodiment of the present application;

[0066] Figure 9 It is an interaction flowchart of a business video recommendation method provided by an embodiment of the present application;

[0067] Figure 10 It is a schematic diagram of the functional interface of a virtual character provided by an embodiment of the present application;

[0068] Figure 11 It is a schematic diagram of the virtual environment after starting the practice mode provided by an embodiment of the present application;

[0069] Figure 12 It is a schematic diagram of the full-screen display of a game video provided by an embodiment of the present application;

[0070] Figure 13 It is a schematic diagram of the notification interface of the client version provided by an embodiment of the present application;

[0071] Figure 14It is a schematic structural diagram of a recommended device for business videos provided by an embodiment of the present application;

[0072] Figure 15 It is a schematic structural diagram of a display device for business videos provided by an embodiment of the present application;

[0073] Figure 16 It is a schematic structural diagram of a computer device provided by an embodiment of the present application;

[0074] Figure 17 It is a schematic structural diagram of another computer device provided by an embodiment of the present application. Detailed implementation manners

[0075] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0076] In the present application, terms such as "first" and "second" are used to distinguish identical or similar items with basically the same functions and effects. It should be understood that there is no logical or chronological dependence between "first", "second", and "nth", nor are the quantity and execution order limited.

[0077] In the present application, the term "at least one" means one or more, and the meaning of "multiple" is two or more. For example, multiple business videos mean two or more business videos.

[0078] The term "including at least one of A or B" in the present application covers the following situations: only including A, only including B, and including both A and B.

[0079] The user-related information (including but not limited to device information, personal information, behavior information, etc.), data (including but not limited to data for analysis, stored data, displayed data, etc.), and signals involved in the present application, when applied to specific products or technologies using the methods of the embodiments of the present application, are all with the permission, consent, authorization of the user or fully authorized by all parties, and the collection, use, and processing of the relevant information, data, and signals need to comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the business behavior information and video attribute information involved in the present application are obtained under full authorization.

[0080] First, introduce the nouns involved in the embodiments of the present application:

[0081] Virtual environment: It is a virtual environment displayed (or provided) when an application runs on a terminal. This virtual environment can be a three-dimensional virtual environment or a two-dimensional virtual environment. This three-dimensional virtual environment can be a simulation environment of the real world, a semi-simulation and semi-fictional environment, or a purely fictional environment.

[0082] Virtual character: It refers to an active object in the virtual environment provided by the service client. This active object can be a virtual person, a virtual animal, an anime character, etc. Before the start of a game round, the user can select the virtual character they want to control in this round. Usually, since different virtual characters usually play different lane positions, the user can select a certain virtual character belonging to that lane position according to the lane position assigned in this round. Among them, taking the 5V5 battle in a MOBA game as an example, the lane positions include at least one of the solo lane, the mid lane, the jungle, the marksman, or the support. Optionally, when the virtual environment is a three-dimensional virtual environment, the virtual character is a three-dimensional solid model created based on animation skeleton technology. Each virtual character has its own shape and volume in the three-dimensional virtual environment and occupies a part of the space in the three-dimensional virtual environment. Optionally, when the virtual environment is a two-dimensional virtual environment, the virtual character is a two-dimensional plane model created based on animation technology. Each virtual character has its own shape and area in the two-dimensional virtual environment and occupies a part of the area in the two-dimensional virtual environment.

[0083] MOBA game (Multiplayer Online Battle Arena): It is a game that provides several strongholds in a virtual environment. Users from different teams control virtual characters to fight in the virtual environment, aiming to capture strongholds or destroy the strongholds of the opposing team. For example, a MOBA game can divide users into two opposing teams, scatter the virtual characters controlled by the users in the virtual environment to compete with each other, and take destroying or capturing the innermost stronghold (such as the crystal) of the enemy as the victory condition. A MOBA game is played in rounds, and the duration of one round of a MOBA game is from the start of the game to the moment when the victory condition is achieved.

[0084] Service client: It refers to the application installed on the terminal that can be used to access the service. According to the different service types of the service, the service client also has different types. For example, when the service type is a game, the service client refers to the game client; when the service type is a virtual simulation activity, the service client refers to the simulation client.

[0085] Service account: It refers to the account that logs in to the service client and can access the service. For example, when the service client is a game client, the service account refers to the game account. The game account can be an account registered by the user in the service client or an account of other platforms authorized by the user to log in. The embodiments of this application do not make specific limitations on this.

[0086] Business video: It refers to a video related to the business services provided by a business client. For example, when the business client is a game client, the business video refers to a game video, which can be a character teaching video, an equipment guide video, a gameplay explanation video, a version update video, etc.

[0087] KOL (Key Opinion Leader): It refers to a person who has more and more accurate product information, is accepted or trusted by the relevant group, and has a great influence on the conversion behavior of this group. In the embodiments of this application, a KOL account refers to an account that publishes or creates business videos. Usually, the number of fans of a KOL account is greater than a preset fan quantity threshold. Taking the MOBA game as an example, a KOL account may be some game streamer accounts, strategy author accounts, game news accounts, or high-end player accounts, etc. Of course, ordinary players can also upload the business videos they create, and the embodiments of this application do not make specific limitations on this.

[0088] AI (Artificial Intelligence): It is to use a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, sense the environment, acquire knowledge, and use knowledge to obtain the best results of theory, methods, technologies, and application systems. In other words, artificial intelligence is a comprehensive technology in computer science. It 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, so that the machine has the functions of perception, reasoning, and decision-making.

[0089] Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, including both hardware-level technologies and software-level technologies. Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technologies 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.

[0090] ML (Machine Learning): It is an interdisciplinary subject involving multiple fields such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how a computer simulates or realizes human learning behavior to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve its own performance. Machine learning is the core of artificial intelligence and the fundamental way to make a computer intelligent. 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 teaching learning.

[0091] Machine learning training: A method of training existing data using a machine learning model to obtain a corresponding computational model. In the embodiments of this application, the video recommendation model, the behavior feature extraction model, and the video feature extraction model all belong to machine learning models.

[0092] CDN (Content Delivery Network): A distributed network established and covering the bearer network, consisting of a group of edge node servers distributed in different regions. Usually, the video platform to which the service video belongs adopts the CDN architecture, allowing terminals in different regions to pull the service video from the edge node server closest to themselves. This way of pulling nearby makes the transmission process of the service video have a lower delay and a higher playback smoothness.

[0093] AB Test: A version test method for business clients, which refers to testing the reaction differences of users to two versions (only this element is different) of the same element A and B of a certain product, so as to make subsequent trade-off judgments.

[0094] With the research and progress of artificial intelligence technology, artificial intelligence technology has been studied and applied in multiple fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, driverless, autonomous driving, drones, robots, smart healthcare, smart customer service, vehicle networking, autonomous driving, intelligent transportation, etc. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role. The solution provided in the embodiments of this application involves technologies such as machine learning of artificial intelligence, which will be specifically described through the following several embodiments.

[0095] Hereinafter, the implementation environment of the embodiments of this application will be described.

[0096] Optionally, the server is 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 (Content Delivery Network), and big data and artificial intelligence platforms. Optionally, the terminal is a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, and this application does not make any restrictions here.

[0097] Figure 1It is a schematic diagram of the implementation environment of a method for recommending business videos provided by an embodiment of the present application. Refer to Figure 1 , in this implementation environment, it may include: a terminal 110 and a server 120, and the terminal 110 is connected to the server 120 through a wireless network or a wired network.

[0098] The terminal 110 installs and runs a business client that supports a virtual environment, and this business client can be a game client, a simulation client, a live broadcast client, a social client, etc. When the terminal runs the business client, an interactive interface of the business client is displayed on the screen of the terminal 110.

[0099] In the embodiments of the present application, taking the business client as a game client as an example for illustration, the game client includes but is not limited to any one of MOBA games, multiplayer online battle programs, virtual simulation programs, battle royale shooting games, and simulation strategy games (Simulation Game, SLG). In one example, the game client is a MOBA game. At this time, the user can use the terminal 110 to control a virtual character located in the virtual environment to perform activities, and the virtual character can be called the user's main controlled virtual character. The activities of the virtual character include but are not limited to at least one of adjusting the body posture, crawling, walking, running, cycling, flying, jumping, shooting, attacking, and releasing skills. Schematically, the virtual character is a virtual person, such as a simulated person or an anime character. The user can control the virtual character to wear different skins to present different appearance dressing effects.

[0100] The terminal 110 can generally refer to one of multiple terminals, and only the terminal 110 is used as an example in this embodiment. The device types of the terminal 110 include at least one of smart phones, tablet computers, multimedia playback devices, PCs (Personal Computers), intelligent robots, vehicle-mounted terminals, wearable devices, AR (Augmented Reality) devices, VR (Virtual Reality) devices, MR (Mixed Reality) devices, e-book readers, digital players, laptop computers, and desktop computers.

[0101] Figure 1Only two terminals are shown, but in different embodiments, there are multiple other terminals that can access the server 120. Optionally, there is also one or more terminals corresponding to the developer, on which a development and editing platform for the business client supporting the virtual environment is installed. The developer can edit and update the business client on this terminal, and transmit the installation package of the updated business client to the server 120 through a wired or wireless network. The terminal 110 can download the installation package of the business client from the server 120, thereby realizing the version update of the business client.

[0102] The server 120 includes at least one of a single server, multiple servers, a cloud computing platform, or a virtualization center. The server 120 is used to provide background services for the business client supporting the virtual environment. Optionally, the server 120 undertakes the main computing work, and the terminal 110 undertakes the secondary computing work; or, the server 120 undertakes the secondary computing work, and the terminal 110 undertakes the main computing work; or, a distributed computing architecture is adopted between the server 120 and the terminal 110 for collaborative computing.

[0103] In some embodiments, the server 120 is 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, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0104] Those skilled in the art can know that the number of the above terminals is more or less. For example, the above terminal is only one, or the above terminals are dozens or hundreds, or more. The embodiments of the present application do not limit the number and device type of the terminals.

[0105] Hereinafter, taking the business type as a game as an example, different functional modules of the server 120 in the above implementation environment will be introduced in detail.

[0106] Figure 2 is a logical architecture diagram of a video recommendation system provided by an embodiment of the present application. As Figure 2 shown, the business client running on the terminal refers to a game client, and the server providing background services for the game client is called a game server 200. In the game server 200, there are a game background 201, a log experience analysis system 202, a big data platform 203, a video recommendation system 204, a video feature management module 205, a video platform 206, and a video CDN module 207.

[0107] The game backend 201 is used to interact with the game client on the terminal and provide the main game services for the game client. For each game session, it provides a battle room for each user participating in the session to access the battle, such as 1V1 battle, 3V3 battle, 5V5 battle, etc.

[0108] The log experience analysis system 202 is used to collect the logs reported by the game accounts that log in to the game client. Since the logs may be used to report business behaviors, may also be used to report anomalies or errors (BUGs), or may also be used to report usage feedback, etc. Since the logs reported by the game accounts may be diverse and complex, the log experience analysis system 202 needs to perform preprocessing tasks such as filtering and screening on the logs reported by the game accounts. Then, the log experience analysis system 202 uploads the screened business logs related to business behaviors to the big data platform 203. It should be noted that the reporting, collection, and analysis of logs all require the full authorization and consent of the users.

[0109] The big data platform 203 is used to perform log analysis on the business logs uploaded by the log experience analysis system 202 to obtain the business behavior information of the game account. Then, it extracts the characteristics of the business behavior from the analyzed business behavior information to obtain the business behavior characteristics of the game account. Optionally, the big data platform 202 performs machine learning modeling on the business behavior information to train a behavior characteristic extraction model, and uses this behavior characteristic extraction model to extract the business behavior characteristics of each game account.

[0110] The video recommendation system 204 is used to implement the accurate personalized business video recommendation task for each game account. When each game account needs to pull a business video, it can send a video recommendation request to the game backend 201 through the game client. The game backend 201 routes the video recommendation request to the video recommendation system 204. Then, the video recommendation system 204 obtains the business behavior characteristics of the game account from the big data platform 203, obtains the video attribute characteristics of each of the multiple business videos from the video feature management module 205, and determines the target video to be recommended from the multiple business videos for the game account according to the pulled business behavior characteristics and the multiple video attribute characteristics. Then, the video recommendation system 204 returns the video recommendation information of the target video to the game backend 201, and the game backend 201 returns the video recommendation information to the game client.

[0111] The video feature management module 205 is used to manage the video attribute information of each business video, and extract the video attribute features of each business video according to the video attribute information of each business video. Optionally, the video feature management module 205 performs machine learning modeling on the video attribute information to train a video feature extraction model, and uses the video feature extraction model to extract the video attribute features of each business video. Among them, the business video can be a game video, and the game video includes character teaching videos, equipment strategy videos, gameplay explanation videos, version update videos, etc. Since the publisher accounts of business videos (usually KOL accounts) may take down or delete the previously uploaded business videos, and the newly uploaded business videos need to be reviewed and approved before they can be seen by the audience, the status of business videos (under review, approved, not approved, taken down, etc.) is maintained and updated through the video feature management module 205.

[0112] The video platform 206 is used to store and manage all business videos. Ordinary players and KOL users can upload or publish their contributed / created / secondary-created / edited business videos on the video platform 206. Here, the identity of the creator of the business video and the video content are not specifically limited. For example, the business video can be a teaching video for a certain virtual character, or a strategy video for a certain lane position, or an introduction video for the latest version, or a compilation of highlights from the video recording / screen recording of a game match.

[0113] The video CDN module 207 is used to cache and distribute personalized recommended business videos to the client. After the game backend 201 returns the video recommendation information to the game client, the game client will find the nearest CDN server from the video CDN module 207, and pull the target video indicated by the video recommendation information from the CDN server according to the video recommendation information. Among them, the CDN server first queries the target video from the cache. If the target video is found, it directly returns the target video to the game client through the streaming transmission method. If the target video is not found, it then pulls the target video from the video platform 206 to the cache through the streaming transmission method and returns it to the game client. It should be noted that in the streaming transmission method of the target video, the game client can play the loaded segments of the target video while continuing to pull the unloaded segments of the target video.

[0114] Next, the basic process of the business video recommendation method in the embodiments of the present application will be introduced.

[0115] Figure 3 is a flowchart of a business video recommendation method provided by an embodiment of the present application. See Figure 3 This embodiment is executed by a computer device, and the computer device can beFigure 1 In the server 120 in Figure 2 when the service type is a game, the computer device can also be the game server 200 in

[0116] 301. In response to a video recommendation request of a service account, the server obtains service behavior information of the service account, where the service behavior information is information about service behaviors performed by the service account in the service client.

[0117] Among them, the server is used to provide background services for the service client. The service client refers to an application program used to provide service services to users. According to different service types of service services, the service client also has different types. For example, when the service type is a game, the service client refers to a game client; when the service type is a virtual simulation activity, the service client refers to a simulation client.

[0118] Among them, the service account refers to an account logged in to the service client of the terminal and used to access the service service. For example, when the service client is a game client, the service account refers to a game account. The game account can be an account registered by the user in the service client or an account of other platforms authorized by the user to log in. The embodiments of the present application do not specifically limit this.

[0119] Among them, the service behavior information refers to information about service behaviors performed by the service account in the service client. The service behavior information includes operation behavior information, historical video records, historical game recordings, account attribute information, etc. The operation behavior information indicates the game behavior information accumulated by the service account in the virtual game. The historical video record indicates the interaction behavior information of the service account for the recommended historical videos. The historical game recording refers to the screen recording video of the service account's most recent N (N≥1) games, or can also refer to the screen recording video of the service account's most recent N games using each virtual character. The account attribute information refers to the attribute information of the service account on the service service. The embodiments of the present application do not specifically limit the content of the service behavior information, and the service behavior refers to the behavior related to the service service performed in the service client. It should be noted that the collection and use of service behavior information require the full authorization and consent of the user.

[0120] Among them, the video recommendation request is used to request the server to perform personalized business video recommendations for the business account. The video recommendation request carries at least the account identification (ID) of the business account, so that the server can query the business behavior information of the business account according to the account identification of the business account. Optionally, the video recommendation request may also carry the role identification of the virtual character. This facilitates the server to recommend only the business videos associated with the virtual character to the business account according to the role identification of the virtual character, making the business videos not only adapt to the business account but also to a specific virtual character. For example, teaching videos, strategy videos, etc. for a certain virtual character are recommended to the business account. The embodiments of the present application do not specifically limit whether the video recommendation request carries the role identification.

[0121] In some embodiments, the user logs in to the business account in the business client of the terminal and opens some interactive interfaces through the business account. If the currently opened interactive interface contains business videos, a video recommendation request for the business account is triggered to be sent to the server. The video recommendation request carries at least the account identification of the business account. Optionally, if the interactive interface is also associated with a certain virtual character, the video recommendation request may also carry the role identification of the virtual character.

[0122] Further, the server receives the video recommendation request of the business account, parses the video recommendation request to obtain the account identification of the business account, and then, according to the account identification of the business account, obtains the business behavior information of the business account. For example, the business behavior information of each business account is associated and stored with the account identification of each business account in the server, so that the business behavior information of the business account hit by the index can be queried with the account identification of the business account as the index. Optionally, if the video recommendation request also carries the role identification of the virtual character, then the role identification is used to screen out multiple business videos in step 303 below, but it is not necessary for the video recommendation request to carry the role identification of the virtual character.

[0123] In an exemplary scenario, Figure 4 is a schematic diagram of the recommendation of a game video provided by the embodiments of the present application. Taking Figure 4 as an example, in the case where the business type is a game, the business client refers to the game client, the business account refers to the game account, the business video refers to the game video, and the server refers to the game server. In this case, the game server contains a game background ( Figure 4(not shown in the figure), a log experience analysis system 401, a big data platform 402, a video recommendation system 403, a video feature management module 404, a video CDN module 405, and a video database (DB) 406. The game client 400 reports the logs generated during the operation of the client to the log experience analysis system 401. The log experience analysis system 401 is responsible for screening various logs reported by the game account, obtaining business logs related to business behaviors, and sending the screened business logs to the big data platform 402. Then, the big data platform 402 performs log analysis on the business logs uploaded by the log experience analysis system 401 to obtain the business behavior information of the game account. Further, when the interactive interface accessed by the game client 400 contains game videos, the game client 400 sends a video recommendation request to the game background. The video recommendation request carries at least the account identifier of the game account. The game background receives the video recommendation request and sends the video recommendation request to the video recommendation system 403. The video recommendation system 403 parses the video recommendation request to obtain the account identifier of the game account. Then, according to the account identifier of the game account, it pulls the business behavior information of the game account from the big data platform 402. Alternatively, the big data platform 402 not only maintains and stores the business behavior information of each game account, but also further extracts the business behavior characteristics of each game account for the business behavior information of each game account. In this case, the video recommendation system 403 can directly pull the business behavior characteristics of the game account from the big data platform 402 based on the account identifier of the game account.

[0124] 302. The server extracts the business behavior characteristics of the business account based on the business behavior information. The business behavior characteristics represent the characteristics of the business behaviors executed by the business account in the business client.

[0125] In some embodiments, the server performs feature extraction based on the business behavior information obtained in step 301 to obtain the business behavior characteristics of the business account. The business behavior characteristics are equivalent to transferring the business behavior information to the feature space and can better express the feature information hidden in the business behavior information. Optionally, the business behavior characteristics can be represented as a business behavior vector. For example, the business behavior vector can be an embedding vector or a latent space vector. The embodiments of the present application do not specifically limit the form of the business behavior vector, and the form of the business behavior vector depends on the feature extraction method.

[0126] In some embodiments, the server may use a behavior feature extraction model obtained through machine learning training to extract business behavior features. The feature extraction method of the behavior feature extraction model will be described in detail in the next embodiment and will not be elaborated here. It should be noted that the behavior feature extraction model is only a possible feature extraction method. The server may also use other feature extraction methods, such as embedding processing, one-hot encoding, etc., or separately extract sub-features of the corresponding modality for different modalities of business behavior information, and then fuse the sub-features of different modalities into business behavior features. The embodiments of the present application do not specifically limit the feature extraction method.

[0127] Still taking Figure 4 as an example for further illustration, the extraction of business behavior features may be implemented by the big data platform 402 or the video recommendation system 403. For example, the video recommendation system 403 only pulls business behavior information from the big data platform 402 and then extracts features from the business behavior information to obtain business behavior features; or the video recommendation system 403 directly pulls business behavior features from the big data platform 402. After the big data platform 402 analyzes business logs to obtain business behavior information, it will further extract features from the business behavior information to obtain business behavior features, and associate and store the account identifier, business behavior information, and business behavior features.

[0128] 303. The server extracts video attribute features of each of the multiple business videos based on the video attribute information of the multiple business videos, and the video attribute features characterize the characteristics of the video attributes possessed by the business videos.

[0129] Among them, the business video refers to a video related to the business service provided by the business client. For example, when the business client is a game client, the business video refers to a game video, which may be a character teaching video, an equipment strategy video, a gameplay explanation video, a version update video, etc. Another example is that when the business client is a simulation client, the business video refers to a simulation video, which may be a strategy video for virtual simulation activities, a map exploration video, a dungeon clearance video, a competitive highlights video, etc.

[0130] Among them, the video attribute information is used to characterize the video attributes of the service video. Optionally, the video attribute information includes publisher account information, video popularity information, video details information, associated video information, etc. The publisher account information refers to the attribute information related to the publisher account of the service video. The video popularity information refers to the comprehensive statistical information of the video platform on the interaction behaviors of each service account for the service video. The video details information refers to the inherent attribute information of the service video itself. The associated video information refers to the attribute information of other service videos associated with the service video. The embodiments of the present application do not specifically limit the content of the video attribute information. It should be noted that the collection and use of the video attribute information need to obtain the full authorization and consent of the video publisher.

[0131] In some embodiments, the server first determines multiple service videos available for recommendation for the service account, then obtains the video attribute information of each service video, and then extracts features from the video attribute information of each service video to obtain the video attribute features of each service video. The video attribute features are equivalent to transferring the video attribute information into the feature space and can better express the feature information hidden in the video attribute information. Optionally, the video attribute features can be represented as a video attribute vector. For example, the video attribute vector can be an Embedding vector, a latent space vector, or a One-Hot (one-hot encoding) vector, etc. The embodiments of the present application do not specifically limit the form of the video attribute vector, and the form of the video attribute vector depends on the feature extraction method.

[0132] In some embodiments, using all the service videos that have passed the review in the video platform as the multiple service videos can expand the range of service videos available for recommendation and improve the generalization ability of service video recommendation.

[0133] In some embodiments, when the role identifier of the virtual character is carried in the video recommendation request, only using all the service videos associated with the virtual character and passing the review as the multiple service videos can perform specific personalized vertical recommendation for a certain virtual character, so that the service video recommendation is not only for the service account but also for the specified virtual character, further improving the recommendation accuracy of the service video.

[0134] In some embodiments, the server may use a video feature extraction model obtained through machine learning training to extract video attribute features. The feature extraction method of the video feature extraction model will be described in detail in the next embodiment and will not be elaborated here. It should be noted that the video feature extraction model is only a possible feature extraction method. The server may also use other feature extraction methods, such as embedding processing, one-hot encoding, etc., or perform different data processing methods for video attribute information of different data types to obtain sub-features of one data type, and then fuse the sub-features of different data types into video attribute features. The embodiments of the present application do not specifically limit the feature extraction method.

[0135] Still taking Figure 4 as an example to continue the description, the extraction of video attribute features may be implemented by the video feature management module 404. After determining multiple business videos, the video recommendation system 403 may obtain the video attribute features of each business video from the video feature management module 404. Optionally, the video database 406 stores at least the video attribute information of each business video, and when the video attribute information changes, the video feature management module 404 is responsible for refreshing the video attribute information stored in the video database 406. For example, the number of video likes, the number of video collections, etc. may change, and the video feature management module 404 may regularly update the video attribute information of each business video.

[0136] Furthermore, after determining multiple business videos available for recommendation, the video recommendation system 403 requests the video attribute features of these multiple business videos from the video feature management module 404. The video feature management module 404 is responsible for pulling the video attribute information of each business video from the video database 406, performing feature extraction on the video attribute information of each business video to obtain the video attribute features of each business video, and returning them to the video recommendation system 403. Optionally, the video database 406 may also store the video attribute features of each business video. In this case, the video feature management module 404 may directly pull the video attribute features of each business video from the video database 406 without the need to calculate the video attribute features in real time, improving the acquisition efficiency of the video attribute features.

[0137] In an exemplary scenario, after receiving a request from the video recommendation system 403 for video attribute features, the video feature management module 404 first queries the video attribute features of each service video from the video database 406. If the video attribute features are queried, it indicates that the video attribute features have been calculated and cached previously, and then the queried video attribute features are directly read. If the video attribute features are not queried, only the video attribute information is read, and feature extraction is performed on the video attribute information to obtain the video attribute features, and then the calculated video attribute features are stored back in the video database 406. It should be noted that since the video attribute information may change, when the video feature management module 404 refreshes the video attribute information stored in the video database 406, it is also necessary to recalculate the video attribute features based on the latest video attribute information to achieve synchronous refreshing of the video attribute features, which can avoid accessing outdated video attribute features and improve the accuracy of the video attribute features. It should be noted that in the video database 406, the video identifier, video attribute information, and video attribute features are associated and stored. If the video attribute features are not stored, only the video identifier and video attribute information are associated and stored.

[0138] 304. The server recommends a target video for the service account from the multiple service videos based on the service behavior feature and the video attribute feature, and the matching degree between the target video and the service account meets the recommendation condition.

[0139] In some embodiments, the server performs personalized video recommendation for the service account from multiple service videos according to the service behavior feature of the service account extracted in step 302 and the video attribute features of each service video extracted in step 303, that is, determines the target video to be recommended from the multiple service videos and recommends the target video to the service account, where the matching degree between the target video and the service account meets the recommendation condition. The target video can be one or more, and the present application embodiment does not specifically limit the number of target videos.

[0140] Schematically, if the number of target videos is only one, then the matching probability between each service video and the service account can be calculated according to the service behavior feature and the video attribute feature of each service video, and the service video with the highest matching probability is determined as the target video, ensuring that the matching degree between the target video and the service account is the highest among the multiple service videos. For example, if the service behavior feature shows that the user's operation level of the virtual character belongs to a beginner, then the target video is usually a teaching video or an explanatory video for beginners, ensuring a high matching degree between the target video and the service account and improving the recommendation accuracy of the target video.

[0141] Schematically, if the number of target videos is greater than one, that is, the number of target videos is two or more. Suppose the number of target videos is K (K≥2). Similarly, according to the service behavior characteristics and the video attribute characteristics of each service video, calculate the matching probability between each service video and the service account, and sort the service videos in descending order of the matching probability. Determine the top K service videos in the sorting as the K target videos. This avoids recommending only one target video to the service account. If the service account is not interested in the only recommended target video, it can also switch to other recommended videos, further improving the recommendation accuracy of the target videos.

[0142] Schematically, if a probability threshold is preset, similarly, according to the service behavior characteristics and the video attribute characteristics of each service video, calculate the matching probability between each service video and the service account. If there is one or more service videos with a matching probability greater than the probability threshold, then determine all service videos with a matching probability greater than the probability threshold as target videos (ensuring that there is at least one target video). If there is no service video with a matching probability greater than the probability threshold, then determine the service video with the highest matching probability as the target video. This also improves the recommendation accuracy of the target videos.

[0143] In some embodiments, after determining the target videos to be recommended to the service account from multiple service videos, the server can return the video recommendation information of the target videos to the service client of the terminal, so that the service client can pull the target videos independently according to the video recommendation information, or directly return the target videos to the service client. The embodiments of the present application do not specifically limit the recommendation method of the target videos. Among them, the video recommendation information at least includes the video identifier of the target video, so that the video recommendation information can be used to uniquely indicate the target video.

[0144] Still taking Figure 4For illustration purposes, the video recommendation system 403 calculates the matching probability between each business video and the business account based on the business behavior characteristics obtained from the big data platform 402 and the video attribute characteristics of each of the multiple business videos obtained from the video feature management module 404. Then, based on the calculated matching probabilities, one or more target videos are determined from the multiple business videos. Next, the video recommendation information of the one or more target videos is returned to the game client 400. The game client 400 can pull the clicked target video from the video CDN module 405 according to the video recommendation information. The target video can be sent to the game client 400 from the video CDN module 405 in a streaming transmission manner, so that it can be played while loading, saving the communication overhead of business video transmission. Optionally, the video recommendation system 403 can also directly notify the video CDN module to return the target video to the game client 400. The embodiments of the present application do not specifically limit this.

[0145] All of the above optional technical solutions can be combined arbitrarily to form optional embodiments of the present disclosure, which will not be elaborated here one by one.

[0146] The method provided by the embodiments of the present application, in the case of receiving a video recommendation request of a business account, on the one hand extracts the business behavior characteristics of the business account, and on the other hand extracts the video attribute characteristics of each of the multiple business videos, so that by using the business behavior characteristics and each video attribute characteristic, personalized business video recommendation can be performed for the current business account, and target videos that meet the recommendation conditions in terms of the matching degree with the business account are determined from the multiple business videos, and the target videos are recommended to the business account, thus ensuring that the target videos and the business account are relatively well matched, so that the target videos can as much as possible meet the viewing intention of the business account, and improving the recommendation accuracy of the business videos.

[0147] In the previous embodiment, the basic process of the business video recommendation method on the server side was briefly introduced. In the embodiments of the present application, various possible implementation manners of each step of the business video recommendation method will be described in detail.

[0148] Figure 5 is a flowchart of a method for recommending a business video provided by the embodiments of the present application. Refer to Figure 5 , this embodiment is executed by a computer device, and the computer device can be Figure 1 the server 120 therein. When the business type is a game, the computer device can also be Figure 2 the game server 200 therein. Below, taking the computer device as the server as an example for illustration, this embodiment includes the following steps:

[0149] 501. In response to a video recommendation request of a service account, the server obtains the service behavior information of the service account, where the service behavior information is the information of the service behavior executed by the service account in the service client.

[0150] The above step 501 is the same as step 301 in the previous embodiment and will not be elaborated here.

[0151] In some embodiments, the service behavior information includes at least one of operation behavior information, historical video records, historical game records, or account attribute information. The embodiments of the present application do not specifically limit the content of the service behavior information. By considering various types of service behavior information, various types of service behavior information can fully represent the information of the service behavior executed by the service account in the service client, thereby enhancing the expression ability of the service behavior information, further improving the accuracy of the service behavior characteristics extracted from the service behavior information, and thus improving the accuracy of the service video recommendation based on the service behavior characteristics.

[0152] Among them, the operation behavior information indicates the accumulated game behavior information of the service account in the virtual game. Taking the service client as a MOBA game as an example, the operation behavior information of the service account, that is, the game account, in the MOBA game includes, but is not limited to: 1) The accumulated operation information for each virtual character, such as the character score of each virtual character (such as the single hero rating of hero A), the current character ranking, the historical highest character ranking, the proficiency level, the number of games played in battle, the character win rate, etc.; 2) The accumulated operation information for each lane position, such as the lane score of each lane position (such as the total rating of all heroes in the shooter lane), the current lane ranking, the historical highest lane ranking, etc.; 3) The rank, total number of games, number of winning games, total win rate, KDA (Kill-Death-Assist, battle loss rate), team participation rate, average damage value per game (the damage value caused by the user's own hero in each game on average), average damage received per game (the damage received by the user's own hero in each game on average), average tower damage value per game (the damage caused by the user's own hero to the defense tower in each game on average), the number of gold coins obtained per minute (the economic value obtained per minute on average in each game), the number of MVP (Most Valuable Player) times, the number of S evaluations, the number of A evaluations, the number of five consecutive kills, the number of four consecutive kills, the number of three consecutive kills, and other operation evaluation information. Through the above operation behavior information, the game level of the user can be comprehensively measured and evaluated. It should be noted that the collection and use of the operation behavior information require the full authorization and consent of the user.

[0153] Among them, the historical video record indicates the interaction behavior information of the business account for the recommended historical videos. For example, for each historical video that has been recommended, it will record whether the historical video has been played (clicked). If the historical video has been played, it will further record the playing duration of the historical video, whether it has been fully played, whether it has been fast-forwarded during the playing process, whether the video has been favorited, whether the publisher's account has been followed, whether it has been shared with other accounts, etc. That is, for each historical video, it will record the playing (i.e., clicking) behavior, favoriting behavior, following behavior, sharing behavior, etc. of the business account. Through the above historical video records, it is possible to comprehensively measure and evaluate whether the user is interested in the historical videos, which indirectly reflects the accuracy of the historical video recommendations. It should be noted that the collection and use of historical video records require the full authorization and consent of the user.

[0154] Among them, the historical game video is the recorded video of the business account's most recent N (N≥1) games, or it can also be the recorded video of the business account's most recent N games for each virtual character. The embodiments of the present application do not specifically limit this. It should be noted that the collection and use of historical game videos require the full authorization and consent of the user. Only after the business client has obtained the full authorization of the user will it upload the historical game videos to the server for use in the business video recommendation task.

[0155] Among them, the account attribute information refers to the attribute information of the business account on the business service, including but not limited to: daily activity, weekly activity, average daily online duration, average daily game duration, account level, achievement system, title system, etc. The embodiments of the present application do not specifically limit the content of the account attribute information. It should be noted that the collection and use of account attribute information require the full authorization and consent of the user.

[0156] In some embodiments, after receiving the video recommendation request of the business account, the server parses the video recommendation request to obtain at least the account identifier of the business account, and based on the account identifier of the business account, obtains the business behavior information of the business account. For example, the business behavior information of each business account is associated and stored with the account identifier of each business account in the server. In this way, the business behavior information of the business account hit by the index can be queried using the account identifier of the business account as the index. For example, the business behavior information of the current business account includes at least one of operation behavior information, historical video record, historical game video, or account attribute information. The server can configure more or fewer types of business behavior information according to the business requirements of the business account for business videos, and can also adjust which types of business behavior information to collect and adopt according to the different authorization situations of the business account. Different business accounts can collect and adopt different types of business behavior information. The embodiments of the present application do not specifically limit this.

[0157] In some embodiments, since the video recommendation request carries at least the account identifier of the service account and may also optionally carry the character identifier of the virtual character, among which the account attribute information in the service behavior information will not be affected by the change of the virtual character. Regardless of whether the video recommendation request carries the character identifier of the virtual character or which virtual character's identifier it carries, the same account attribute information will be retrieved for the same service account. In contrast, the operation behavior information, historical video records, and historical game records in the service behavior information can be specifically filtered according to different virtual characters. For example, if the video recommendation request does not carry the character identifier of the virtual character, then when retrieving the operation behavior information, there is no need to filter according to the virtual character, that is, retrieve the cumulative operation information for each virtual character, the cumulative operation information for each lane position, and various operation evaluation information of the service account in the current season. However, if the video recommendation request carries the character identifier of the virtual character, then when retrieving the operation behavior information, it is necessary to filter according to the virtual character, that is, only retrieve the cumulative operation information of the virtual character indicated by the carried character identifier, the cumulative operation information of the lane position to which the virtual character belongs, and various operation evaluation information of the service account in the games using this virtual character in the current season. This can filter out the operation behavior information that is unique to the virtual character in the service client for a specific virtual character, thereby improving the accuracy of the operation behavior information. Another example, if the video recommendation request does not carry the character identifier of the virtual character, then when retrieving the historical video records, there is no need to filter according to the virtual character, that is, retrieve the interaction behavior information of the service account for the recommended historical videos (historical videos of all virtual characters will be considered). However, if the video recommendation request carries the character identifier of the virtual character, then when retrieving the historical video records, it is necessary to filter according to the virtual character, that is, only retrieve the interaction behavior information of the service account for the recommended historical videos associated with this virtual character, without considering the interaction behavior information of historical videos unrelated to this virtual character. This can filter out the historical video records that are unique to the virtual character in the service client for a specific virtual character, thereby improving the accuracy of the historical video records.For another example, if the video recommendation request does not carry the character identifier of the virtual character, then there is no need to filter according to the virtual character when pulling the historical game video recordings, that is, the screencast videos of the last N games of the business account can be pulled (the historical game videos of all virtual characters controlled by the main account will be taken into consideration). However, if the video recommendation request carries the character identifier of the virtual character, then it is necessary to filter according to the virtual character when pulling the historical game video recordings, that is, only the screencast videos of the last N games of the business account using this virtual character are pulled. In this way, for a specific virtual character, the historical game video recordings that are unique to this virtual character in the business client of the business account can be filtered out, thereby improving the accuracy of the historical game video recordings. The above-mentioned secondary filtering of the complex business behavior information is achieved according to the character identifier of the virtual character. In the case where the character identifier of the virtual character is carried in the video recommendation request, the accuracy of the obtained business behavior information is improved. This is because even for the same user, there may be fluctuations in the operation level when controlling different virtual characters, thus giving rise to the viewing requirements for different business videos. In other words, not only can personalized and accurate business video recommendations for the same business account be realized, achieving the business video recommendation effect of "one size fits one person", but also when recommending for the same business account, different business videos with different adaptation degrees will be recommended according to the differences of different virtual characters, making the recommendation of business videos more granular, which helps to further improve the recommendation accuracy of business videos.

[0158] It should be noted that if the video recommendation request also carries the character identifier of the virtual character, then in addition to being able to filter out the irrelevant content in the business behavior information, the character identifier can also be used to screen out multiple business videos in steps A1 - A2 described in step 502 below. For detailed reference, see the description of steps A1 - A2 below, and no further elaboration will be provided.

[0159] It should also be noted that it is not necessary for the video recommendation request to carry the character identifier of the virtual character, but at least the account identifier of the business account needs to be carried. The embodiments of the present application do not specifically limit whether the character identifier of the virtual character is carried in the video recommendation request. If the character identifier of the virtual character is not carried in the video recommendation request, then there is no need to filter according to the virtual character when pulling the business behavior information, nor is it necessary to execute the optional steps A1 - A2.

[0160] 502. The server extracts the business behavior characteristics of the business account based on the business behavior information, and the business behavior characteristics represent the characteristics of the business behavior executed by the business account in the business client.

[0161] The above step 502 is the same as step 302 in the previous embodiment, and no further elaboration will be provided.

[0162] In some embodiments, the server side can use machine learning methods to train a behavior feature extraction model, and then extract business behavior features based on this behavior feature extraction model. That is, the server inputs the business behavior information obtained in step 501 into the behavior feature extraction model, and the behavior feature extraction model extracts features from the business behavior information to obtain the business behavior features. Here, the behavior feature extraction model is used to extract the business behavior features of business accounts. In this way, by training a behavior feature extraction model to automatically extract business behavior features, the extraction efficiency of business behavior features is greatly improved. Moreover, according to the user's interaction feedback on the recommended videos, the behavior feature extraction model can be continuously optimized or fine-tuned to achieve iterative updates of the behavior feature extraction model. By continuously improving the model performance of the behavior feature extraction model, the accuracy of the business behavior features extracted using the updated behavior feature extraction model is improved in turn.

[0163] In some embodiments, the server can periodically fine-tune the model parameters of the behavior feature extraction model according to the user's interaction feedback on the recommended videos within a period of time to achieve periodic updates of the behavior feature extraction model. Since the behavior feature extraction model may be updated periodically, the server side does not need to cache the business behavior features of each business account. Instead, each time it is needed, it uses the latest behavior feature extraction model to calculate the latest business behavior features in real time, ensuring the timeliness and accuracy of the business behavior features and saving the storage overhead of the server side.

[0164] In other embodiments, the server can periodically fine-tune the model parameters of the behavior feature extraction model according to the user's interaction feedback on the recommended videos within a period of time to achieve periodic updates of the behavior feature extraction model. However, even if the behavior feature extraction model may be updated periodically, the server side can also cache the business behavior features of each business account. Only after the behavior feature extraction model is updated each time, the updated behavior feature extraction model can be used to re-extract the business behavior features of each business account to achieve periodic updates of the business behavior features, ensuring that the latest, optimal-performance, and more accurate business behavior features can be used in each business video recommendation process. In this way, it is not necessary to calculate the business behavior features in real time each time a video recommendation request is received, which can greatly improve the memory access efficiency of the business behavior features, improve the real-time response speed to video recommendation requests, reduce the waiting latency of business accounts, and save the computing overhead of the server side.

[0165] It should be noted that the above method of training a behavior feature extraction model to extract business behavior features is only a possible feature extraction method. The server can also use other feature extraction methods, such as embedding processing, one-hot encoding, etc. Or, for business behavior information of different modalities, sub-features of the corresponding modality can be extracted respectively, and then the sub-features of different modalities can be fused into business behavior features. The fusion methods can be concatenation, element-wise addition, element-wise multiplication, bilinear pooling, etc. The embodiments of the present application do not specifically limit the feature extraction method of business behavior features.

[0166] 503. Based on the video attribute information of multiple business videos, the server extracts the video attribute features of each of the multiple business videos, and the video attribute features characterize the characteristics of the video attributes of the business videos.

[0167] The above step 503 is the same as step 303 in the previous embodiment and will not be elaborated here.

[0168] In some embodiments, the video attribute information includes at least one of the following: publisher account information, video popularity information, video details information, or associated video information. The embodiments of the present application do not specifically limit the content of the video attribute information. By considering various types of video attribute information, various types of video attribute information can fully represent the information of the video attributes of the business videos in different dimensions, thereby enhancing the expression ability of the video attribute information, further improving the accuracy of the video attribute features extracted according to the video attribute information, and thus improving the accuracy of business video recommendation based on the video attribute features.

[0169] Among them, the publisher account information refers to the attribute information related to the publisher account of the business video, including but not limited to: the account identifier of the publisher account, the account nickname of the publisher account, the number of fans of the publisher account, etc. It should be noted that the collection and use of the publisher account information require the full authorization and consent of the video publisher.

[0170] Among them, the video popularity information refers to the comprehensive statistical information of the interaction behaviors of each business account on the video platform for the business video, including but not limited to: the number of video collections, the number of video likes, the number of video plays, the number of video completions, the account identifiers of the commenters of excellent comments, etc. It should be noted that the collection and use of the video popularity information require the full authorization and consent of the video publisher.

[0171] Among them, the video details information refers to the inherent attribute information of the business video itself, including but not limited to: video source, adoption status, first-level classification label, second-level classification label, video duration, multi-cover image configuration, creation time, description information, keyword ID list (separated by English commas), cover image, editor nickname, label ID list (separated by English commas), video title, update time, etc. It should be noted that the collection and use of video details information require the full authorization and consent of the video publisher.

[0172] Among them, the associated video information refers to the attribute information of other business videos associated with this business video, including but not limited to: related recommended video list, publisher account information of each associated video in the related recommended video list, etc. It should be noted that the collection and use of associated video information require the full authorization and consent of the video publisher of each associated video.

[0173] In an exemplary scenario, the type name, data type, and business meaning of each type of information included in the video attribute information are shown in Table 1 below.

[0174] Table 1

[0175]

[0176]

[0177] It should be noted that the above Table 1 only provides an example of video attribute information. The server can collect more or fewer types of video attribute information according to business requirements, and different business videos can also be configured with different types of video attribute information according to different authorization situations. The embodiments of the present application do not specifically limit the types of video attribute information, nor do they specifically limit whether different business videos have the same types of video attribute information.

[0178] In some embodiments, if the role identifier of the virtual role is not carried in the video recommendation request, then all the business videos that have passed the review in the video platform are used as multiple business videos available for recommendation, which can expand the range of business videos available for recommendation and improve the generalization ability of business video recommendation. Further, for each business video, all the video attribute information with full authorization is retrieved, and then feature extraction is performed on the video attribute information of each business video to obtain the video attribute features of each business video.

[0179] In some other embodiments, if the video recommendation request carries the role identifier of a virtual character, then all business videos associated with the virtual character and passed the review can be used as the multiple business videos available for recommendation, enabling specific personalized vertical recommendation for a certain virtual character, so that the business video recommendation is not only for business accounts but also for the specified virtual character, further improving the recommendation accuracy and fine granularity of business videos. Further, for each business video, all fully authorized video attribute information is retrieved, and then feature extraction is performed on the video attribute information of each business video to obtain the video attribute features of each business video.

[0180] Next, through steps A1 to A2, a possible implementation manner for the server to screen out multiple business videos available for recommendation in the case where the video recommendation request carries the role identifier of a virtual character will be introduced:

[0181] A1. In the case where the video recommendation request carries the role identifier of a virtual character, the server queries the candidate video list associated with the virtual character based on the role identifier.

[0182] In some embodiments, in the case where the video recommendation request carries the role identifier of a virtual character in addition to the account identifier of the business account, the server parses the video recommendation request to obtain the account identifier of the business account and the role identifier of the virtual character, retrieves the business behavior information using the account identifier and extracts the business behavior features, which will not be elaborated here. Then, using the parsed role identifier as an index, the server queries the candidate video list associated with the virtual character. In the server, a candidate video list is maintained for each virtual character, and the video identifiers of all candidate videos associated with the virtual character are stored in the candidate video list. However, since the candidate videos need to pass the review before they can be publicly recommended, these candidate videos may be under review, and the review result may be passing the review or not passing the review. Even if they pass the review, they may be taken off the shelf, deleted, or the release may be revoked by the video publisher for some reasons (such as the need for secondary editing, correction of mistakes), and they are in an inaccessible state. The server cannot recommend business videos that have not passed the review or are in an inaccessible state to the business account. Through the following step A2, each candidate video stored in the candidate video list can be screened again to obtain the final multiple business videos available for recommendation.

[0183] Figure 6 is a video management and review flowchart provided by an embodiment of the present application, as Figure 6As shown, taking the publisher account of the video being a KOL account as an example, the video management and review process is introduced. After the KOL account finishes producing a video, the newly produced video is uploaded to the video platform. After that, it can be handed over to the operation staff for review, and also to the development staff and the planning staff for cross-review, which can ensure the accuracy of the video content review. Only when both reviews are passed, the newly reviewed video will be uploaded to the video platform and distributed by the video platform to each video CDN module. Optionally, the new video is not limited to being published only by the KOL account. For example, when the version is updated, the official account of the business client can also publish a video introducing the new version. Another example is that ordinary players can also publish their own recorded videos. The embodiments of the present application do not specifically limit this. Optionally, during the review of the video content, a machine learning model can also be used for review, which saves the labor cost of review. The review method of the embodiments of the present application is not specifically limited either.

[0184] In some embodiments, when maintaining a candidate video list for each virtual character, for any newly uploaded video, since the video publisher will add one or more classification tags (i.e., content tags) to their newly uploaded video when uploading the video, if any classification tag of the newly uploaded video matches the character name, character nickname, or character identifier of a certain virtual character, then the video identifier of this newly uploaded video is added to the candidate video list of the hit virtual character.

[0185] Optionally, since not every video is related to a virtual character, it is possible that a newly uploaded video will not be added to the candidate video list of any virtual character. For example, a novice teaching video may only introduce the gameplay ideas and scene maps and does not involve a specific virtual character. Optionally, in addition to being added by the video publisher himself, the above-mentioned classification tags can also be classification tags generated by the machine automatically performing video content or semantic recognition. The embodiments of the present application do not specifically limit the acquisition method of the classification tags. Optionally, the classification tags may also have different levels, such as first-level classification tags, second-level classification tags, etc. There is one or more second-level classification tags under the same first-level classification tag. The first-level classification tag realizes a relatively general large-category division, and the second-level classification tag realizes a fine sub-category division under a certain large category. The embodiments of the present application do not specifically limit the setting method of the classification tags. Optionally, when determining whether the classification tag matches the virtual character, keyword matching can be performed, or semantic recognition can be performed on the classification tag to determine whether the recognized semantic features match the character features of the virtual character. The embodiments of the present application do not specifically limit this.

[0186] A2. The server determines the multiple service videos from the multiple candidate videos included in the candidate video list.

[0187] In some embodiments, the server can query the review status of each candidate video from all the candidate videos included in the candidate video list queried in step A1, eliminate the candidate videos under review and those that fail the review, and only determine the candidate videos that pass the review as multiple business videos available for recommendation, ensuring that the finally selected business videos not only pass the review but also are associated with the virtual character.

[0188] In other embodiments, the server can traverse each candidate video in the candidate video list queried in step A1, query the review status of the candidate video. If the candidate video passes the review, the candidate video is determined as a business video, and continue to traverse the next candidate video. If the candidate video fails the review, then skip the candidate video, and the candidate video will not be determined as a business video, and continue to traverse the next candidate video until all candidate videos in the candidate video list are traversed, then multiple business videos available for recommendation can be screened out, ensuring that the finally selected business videos not only pass the review but also are associated with the virtual character.

[0189] Next, taking steps A21 to A23 as an example, a possible method for screening business videos will be introduced. By using a scheduled task to pull the candidate video list and screen business videos from the candidate video list, it is possible to save the access overhead of the server for repeatedly pulling the same candidate video list for video recommendation requests of different business accounts, and save the computing overhead of the server.

[0190] A21. The server creates a video pulling scheduled task for the virtual character, and this video pulling scheduled task is used to regularly pull the candidate video list associated with the virtual character.

[0191] For each virtual character, the server can create a video pulling scheduled task for the virtual character, and this video pulling scheduled task is used to regularly pull the candidate video list associated with the virtual character. Since the status of each video in the video platform will continuously change, such as review status change, new video upload, old video deletion, etc., therefore, the server can create a video pulling scheduled task to regularly pull the latest candidate video list associated with the virtual character, and the old videos that have been deleted by the video publisher or the platform will be eliminated from the candidate video list. The regular pulling of the candidate video list can be completed by the video feature management module in the server, and the video feature management module regularly pulls the latest candidate video list of each virtual character from the video platform.

[0192] Figure 7 is the schematic diagram of a method for screening business videos provided by the embodiments of the present application, as Figure 7As shown in the left part, the video platform assigns a unique character tag to each virtual character. The character tag can be a character identifier, a character name, or a character nickname. For each character tag, there is a candidate video list maintained. When a newly uploaded video carries a character tag, it will be added to the corresponding candidate video list. When an old video is deleted, the corresponding old video will also be removed from the candidate video list of the character tag to which it belongs. At the same time, any change in the review status of each candidate video will be recorded in real time in the video platform. For the video feature management module, after loading the character tags tag1~tagN of N virtual characters, for each virtual character, a video pulling scheduled task task1~taskN for this virtual character is created.

[0193] A22. Based on this video pulling scheduled task, the server pulls the candidate video list at intervals of a target duration. The candidate video list contains multiple candidate videos associated with this virtual character.

[0194] In some embodiments, for each video pulling scheduled task created by the server for a virtual character, it will pull the latest candidate video list of this virtual character at intervals of a target duration and record the pulled candidate video list in the cache. Among them, the candidate video list includes multiple candidate videos associated with this virtual character. In addition, the target duration refers to the cycle length of video pulling pre-configured for the video pulling scheduled task, such as every hour, every half hour, every ten minutes, etc. The embodiments of the present application do not specifically limit the target duration. And in the case where the character identifier of the virtual character is carried in the video recommendation request, it is only necessary to access the candidate video list pulled most recently by the video pulling scheduled task of this virtual character in the cache according to the character identifier of the virtual character.

[0195] Still taking Figure 7 as an example for illustration, in the video feature management module, each created video pulling scheduled task will regularly pull the latest candidate video list under the character tag to which it belongs from the video platform. Further referring to Figure 7 the right part of, according to the video pulling scheduled task of the virtual character, the latest candidate video list of the virtual character is pulled from the video platform, and then the following step A23 is executed.

[0196] A23. The server determines the multiple candidate videos that have passed the review from this candidate video list as the multiple business videos.

[0197] In some embodiments, among the candidate videos included in the latest candidate video list of the virtual character pulled by the server from step A22, the review status of each candidate video is queried, the candidate videos under review and those that fail the review are excluded, and the remaining multiple candidate videos that pass the review are determined as multiple business videos available for recommendation, ensuring that the finally selected business videos not only pass the review but also are associated with the virtual character.

[0198] Still taking Figure 7 the right part as an example for illustration, when the role identifier of the virtual character is carried in the video recommendation request, access the candidate video list pulled by the latest video pull scheduling task of the virtual character, traverse each candidate video in the candidate video list, and determine whether the candidate video passes the review. If so (i.e., it passes the review), determine the candidate video as a business video, pull the video attribute information of the business video, store the video attribute information in the Redis cache, and update the index of this business video in the Redis cache according to the role tags of the virtual character; if not (i.e., it fails the review or is still under review), then continue to traverse the next candidate video in the candidate video list, and repeat the above process until all candidate videos in the candidate video list are traversed. All business videos will be obtained, and the video attribute information of each business video will be loaded into the Redis cache. By loading the video attribute information of each business video into the Redis cache, it is not necessary to access the video platform again to pull the video attribute information of each business video after multiple business videos are selected, but only need to query the corresponding video attribute information in the Redis cache according to the allocated index, which can greatly improve the access efficiency of the video attribute information for business videos.

[0199] In the above steps A21 - A23, a possible business video screening method is provided. By using a scheduling task to periodically pull and cache the latest candidate video list, it can save the access overhead of the server for repeatedly pulling the same candidate video list for video recommendation requests of different business accounts, save the computing overhead of the server. Further, each candidate video in the candidate video list is screened one by one to determine whether it passes the review, and finally multiple business videos are obtained. When the screening of business videos is completed, the video attribute information of each business video can also be cached, which greatly improves the screening efficiency of business videos and also improves the access efficiency of video attribute information.

[0200] In some other embodiments, instead of creating a video pulling scheduled task for each virtual character, the server pulls the latest candidate video list from the video platform in real time, which can ensure that the pulled candidate video list has the highest timeliness. Since there may be an error delay of up to the target duration in the scheduled task, this can ensure that all candidate videos in the candidate video list have not been deleted, and can also ensure that the newly uploaded candidate videos are recorded in the candidate video list in a timely manner.

[0201] In the above steps A1 - A2, when the role identifier of the virtual character is carried in the video recommendation request, only all the business videos associated with the virtual character and passed the review are used as the multiple business videos available for recommendation. This can perform specific personalized vertical recommendation for a certain virtual character, making the business video recommendation not only target the business account but also the specified virtual character, further improving the recommendation accuracy of the business videos. In other words, not only can personalized and accurate business video recommendation for the same business account be achieved, reaching the effect of personalized business video recommendation for each user, but when recommending for the same business account, according to the differences of different virtual characters, business videos with different adaptation degrees will also be recommended, making the business video recommendation more granular.

[0202] On the basis of introducing the above determination method for the multiple business videos available for recommendation, for each determined business video, the business attribute information of the business video is loaded. The business attribute information can be separately loaded from the video platform by the video feature management module after determining the business video, or can be Figure 7 loaded into the Redis cache at the same time as the business video is determined, as shown in the right part.

[0203] Furthermore, the server side can use machine learning methods to train a video feature extraction model, and then extract video attribute features based on this video feature extraction model. That is, the server inputs the video attribute information of each business video into the video feature extraction model, and the video feature extraction model extracts features from this video attribute information to obtain the video attribute features of the business video. Here, the video feature extraction model is used to extract the video attribute features of business videos. In this way, by training a video feature extraction model to automatically extract video attribute features, the extraction efficiency of video attribute features is greatly improved. Moreover, according to the user's interaction feedback on the recommended videos, the video feature extraction model can be continuously optimized or fine-tuned to achieve iterative updates of the video feature extraction model. By continuously improving the model performance of the video feature extraction model, the accuracy of the video attribute features extracted using the updated video feature extraction model is improved in turn. Also, in the case where there are a large number of business videos available for recommendation, the video feature extraction model can quickly and efficiently extract the video attribute features of each business video, and there is a more significant improvement effect on the extraction efficiency of video attribute features.

[0204] In some embodiments, the server can regularly fine-tune the model parameters of the video feature extraction model according to the user's interaction feedback on the recommended videos within a period of time to achieve regular updates of the video feature extraction model. Since the video feature extraction model may be updated regularly, the server side does not need to cache the video attribute features of each business video. Instead, each time it is needed, it uses the latest video feature extraction model to calculate the latest video attribute features in real time, ensuring the timeliness and accuracy of the video attribute features and saving the storage overhead on the server side.

[0205] In some other embodiments, the server can regularly fine-tune the model parameters of the video feature extraction model according to the user's interaction feedback on the recommended videos within a period of time to achieve regular updates of the video feature extraction model. However, even if the video feature extraction model may be updated regularly, the server side can also cache the video attribute features of each business video. Only after the video feature extraction model is updated each time, the updated video feature extraction model can be used to re-extract the video attribute features of each business video to achieve regular updates of the video attribute features, ensuring that the latest, optimal-performance, and more accurate video attribute features can be used in each business video recommendation process. In this way, it is not necessary to calculate the video attribute features in real time each time a video recommendation request is received, which can greatly improve the memory access efficiency of the video attribute features, improve the real-time response speed to video recommendation requests, reduce the waiting delay of business videos, and save the computing overhead on the server side.

[0206] It should be noted that the above method of using a training video feature extraction model to extract video attribute features is only one possible feature extraction method. The server can also use other feature extraction methods, such as embedding processing, one-hot encoding, etc. Or, for video attribute information of different data types, different data processing methods can be used respectively to obtain sub-features of one data type, and then the sub-features of different data types are fused into video attribute features. The fusion method can be splicing, element-wise addition, element-wise multiplication, bilinear pooling, etc. The embodiments of the present application do not specifically limit the feature extraction method of video attribute features.

[0207] 504. The server inputs the service behavior feature into a video recommendation model, which is used to predict the video to be recommended for the service account based on the service behavior features of the service account.

[0208] In some embodiments, the server side can use machine learning methods to train a video recommendation model, and then based on this video recommendation model, predict which target video to recommend for a given service account from a given plurality of service videos. Optionally, the server includes a video recommendation system, and the model parameters of the video recommendation model are stored in the video recommendation system. The service behavior features extracted in step 502 are input into the video recommendation model.

[0209] In this way, by training a video recommendation model to automatically predict the target video from multiple service videos, the prediction efficiency of the target video can be improved. Moreover, according to the user's interaction feedback on the recommended video, the video recommendation model can be continuously optimized or fine-tuned to achieve iterative update of the video recommendation model. By continuously improving the model performance of the video recommendation model, the recommendation accuracy of the target video predicted by using the updated video recommendation model is improved in turn.

[0210] 505. The server determines the target video to be recommended from the plurality of service videos through the video recommendation model. The target video meets the recommendation conditions in terms of the matching degree with the service account.

[0211] In some embodiments, when the server calls the video recommendation model, it can use the service behavior features input in step 504 and the multiple video attribute features extracted in step 503. In the feature space, the matching degree between the service behavior features and the video attribute features can quantify the matching degree between each service video and the service account, so as to further screen out the target video to be recommended from the plurality of service videos. The number of target videos can be one or more. The embodiments of the present application do not specifically limit the number of target videos.

[0212] Next, taking steps B1 to B2 as an example, a possible method for screening target videos will be introduced. By using the matching probability to measure the matching degree between the business video and the business account, the matching degree between the business account and each business video in the feature space can be transformed into a normalized matching probability in the probability space, enabling the matching probability to intuitively measure the likelihood of mutual matching between the business video and the business account. The matching probability is positively correlated with the matching degree, which facilitates the use of the matching probability to quickly and efficiently screen out the target video from multiple business videos, improving the prediction efficiency and recommendation accuracy of the target video.

[0213] B1. The server determines the matching probability between the business account and each business video based on the business behavior characteristics and the video attribute characteristics of each business video, and this matching probability indicates the matching degree between the business account and the business video.

[0214] In some embodiments, for each business video among the multiple business videos available for recommendation, the video recommendation model can be invoked to encode based on the business behavior characteristics and the video attribute characteristics of the business video to obtain an encoded feature. The encoded feature integrates the business behavior characteristics and the video attribute characteristics, and then the encoded feature is subjected to exponential normalization (Softmax) to obtain the matching probability between the business account and the business video. Optionally, the video recommendation model includes one or more encoding layers and a Softmax layer. The last encoding layer will output the encoded feature, and the encoded feature is input into the Softmax layer to obtain a matching probability. Among them, when encoding, one or more encoding layers can be used to perform convolution operations, weighting operations, or fully connected operations, etc., and a residual mechanism or an attention mechanism can also be introduced in any encoding layer. The embodiments of the present application do not specifically limit the encoding method.

[0215] B2. The server determines the target video from the multiple business videos based on the matching probability between the business account and each business video.

[0216] In some embodiments, for each business video among the multiple business videos available for recommendation, the server can execute step B1 to obtain a matching probability. By traversing the multiple business videos, multiple matching probabilities can be obtained. Next, based on the multiple matching probabilities, the target video to be recommended is screened out from the multiple business videos. The target video can be one or more, and the embodiments of the present application do not specifically limit the number of target videos.

[0217] Schematically, if there is only one target video, the service video with the highest matching probability can be determined as the target video, ensuring that the target video has the highest matching degree with the service account among multiple service videos. For example, if the user's operation level of the virtual character shown by the service behavior characteristics belongs to a beginner, then the target video is usually a teaching video or an explanatory video for beginners, ensuring a high matching degree between the target video and the service account and improving the recommendation accuracy of the target video.

[0218] Schematically, if the number of target videos is greater than one, that is, the number of target videos is two or more. Assuming the number of target videos is K (K≥2), each service video can be sorted in descending order of the matching probability, and the K service videos ranked in the top K positions are all determined as K target videos. This avoids recommending only one target video to the service account. If the service account is not interested in the only recommended target video, it can also switch to other recommended videos, further improving the recommendation accuracy of the target video.

[0219] Schematically, in the case where a probability threshold is preset in advance, if there is one or more service videos with a matching probability greater than the probability threshold, then all service videos with a matching probability greater than the probability threshold are determined as target videos (ensuring that there is at least one target video). If there is no service video with a matching probability greater than the probability threshold, then the service video with the highest matching probability is determined as the target video. This also improves the recommendation accuracy of the target video.

[0220] In the above steps B1 to B2, by using the matching probability to measure the matching degree between the service video and the service account, the matching degree between the service account and each service video in the feature space can be transformed into a normalized matching probability mapped in the probability space, enabling the matching probability to intuitively measure the likelihood of mutual matching between the service video and the service account. The matching probability is positively correlated with the matching degree, facilitating the use of the matching probability to quickly and efficiently screen out the target video from multiple service videos and improving the prediction efficiency and recommendation accuracy of the target video.

[0221] In some other embodiments, the feature similarity between the service behavior feature and the video attribute feature can also be calculated for the video attribute features of each service video, so as to directly use the feature similarity in the feature space to measure the matching degree between the service video and the service account, thus also ensuring the recommendation accuracy of the target video. The embodiments of the present application do not specifically limit this.

[0222] 506. The server recommends the target video to the service account.

[0223] In some embodiments, after determining the target video to be recommended to the service account from multiple service videos, the server may send video recommendation information of the target video to the terminal logged in by the service account, so that the terminal pulls the target video based on the video recommendation information. The video recommendation information at least includes the video identifier of the target video, so that the video recommendation information can be used to uniquely indicate the target video. In the case where the number of target videos is more than one, the video recommendation information of each target video may be sent to the terminal; or, the server sends the target video to the terminal. The embodiments of the present application do not specifically limit the recommendation method of the target video.

[0224] When the server returns the video recommendation information of the target video to the terminal, the terminal can find the nearest CDN server from the video CDN module and pull the target video indicated by the video recommendation information from the CDN server according to the video recommendation information. The CDN server preferentially queries the target video from the cache. If the target video is found, it directly returns the target video to the terminal through the streaming transmission method. If the target video is not found, it then pulls the target video from the video platform to the cache through the streaming transmission method and returns it to the terminal. In the streaming transmission method of the target video, the game client can play the loaded segments in the target video while continuing to pull the unloaded segments in the target video, saving the communication overhead of the service video transmission. Moreover, since the target video is pulled from the CDN server nearby, the loading efficiency of the target video is also greatly improved, and the possible stuttering situation in the original playing process is improved.

[0225] When the server returns the target video to the terminal, the pulling process of the target video is simplified, and the communication rounds between the terminal and the server are saved. The embodiments of the present application do not specifically limit the pulling method of the target video.

[0226] In the above steps 504-506, a possible implementation manner of recommending the target video for the service account from the multiple service videos based on the service behavior characteristics and the video attribute characteristics is provided, where the matching degree between the target video and the service account meets the recommendation conditions. By using the video recommendation model to predict the target video for the service account, personalized service video recommendation can be realized for each service account, achieving the personalized service video recommendation effect of one person, one face.

[0227] Figure 8 is a development flowchart of a video recommendation model provided by the embodiments of the present application, as Figure 8As shown, when the server includes a big data platform, a video feature management module, and a video recommendation system, the video recommendation system is used to train a video recommendation model. During the training process, multiple sample accounts are determined, and multiple sample videos are constructed for each sample account. The sample videos include positive sample videos and negative sample videos. A positive sample video refers to a business video that is played after being recommended to the sample account, and a negative sample video refers to a business video that is not played after being recommended to the sample account. On this basis, in one iteration, for each sample account, the video recommendation system can obtain the business behavior characteristics of the sample account from the big data platform, obtain the video attribute characteristics of the multiple sample videos from the video feature management module, and then use the video recommendation model to predict the target video among the multiple sample videos. In this iteration, it can be in batch processing mode. For a batch of sample accounts, the target videos of each are predicted respectively. Then, according to whether the target videos predicted by the machine for this batch of sample accounts belong to positive sample videos or negative sample videos, the loss function value of this iteration can be calculated. The above steps are iteratively executed until the training is completed when the stop training condition is met. The stop training condition can be that the loss function value is less than the loss threshold, or the number of iteration steps reaches the set number of steps. The embodiments of the present application do not specifically limit this. After the video recommendation system finishes training the video recommendation model, it can be put into the test environment for index evaluation to obtain the performance test results of the video recommendation model. Among them, the test indicators such as accuracy, recall rate, F1 value, etc. The F1 value (F1 Score, also known as F1 score) is a weighted average of the accuracy and recall rate. Its maximum value is 1 and its minimum value is 0. The larger the F1 value, the better the model. The F1 value takes into account both the accuracy and recall rate of the video recommendation model. The embodiments of the present application do not specifically limit the test indicators. Further, after the test indicators meet the requirements, the video recommendation model is deployed to the production environment for AB testing. If the effect of the AB testing is good, then gray box testing is performed again. If the effect of the gray box testing is good, it is finally fully released in the game client, that is, the video recommendation system is launched in the game client, and the video recommendation model is used to implement personalized business video recommendations for business accounts.

[0228] All of the above optional technical solutions can be combined arbitrarily to form the optional embodiments of the present disclosure, which will not be elaborated one by one here.

[0229] The method provided by the embodiments of the present application, when receiving a video recommendation request of a service account, on the one hand extracts the service behavior characteristics of the service account, and on the other hand extracts the video attribute characteristics of each of multiple service videos. Thus, by using the service behavior characteristics and each video attribute characteristic, it is possible to perform personalized service video recommendations for the current service account, determine a target video that meets the recommendation conditions in terms of the matching degree with the service account from multiple service videos, and recommend the target video to the service account, thereby ensuring that the target video is relatively well-suited to the service account, making the target video as much as possible in line with the viewing intention of the service account, and improving the recommendation accuracy of service videos.

[0230] In the above embodiments, various details in the service video recommendation process are introduced in detail. In the embodiments of the present application, the interaction process of service video recommendation between the terminal and the server will be introduced in detail in combination with the terminal installed with the service client.

[0231] Figure 9 It is an interaction flowchart of a method for recommending service videos provided by the embodiments of the present application. Refer to Figure 9 , this embodiment is implemented through the interaction between the terminal and the server, and both the terminal and the server are exemplary descriptions of computer devices. Below, taking the example of the server returning video recommendation information of the target video to the terminal for illustration, this embodiment includes the following steps:

[0232] 901. When the interactive interface of the service client contains service videos, the terminal sends a video recommendation request of the service account to the server. This video recommendation request is used to obtain the target video recommended for this service account, and this service account is the account logged in to this service client.

[0233] In some embodiments, the user logs in to the service account in the service client of the terminal and opens some interactive interfaces through the service account. During the process of running the service client on the terminal, the user will switch or flow among different interactive interfaces through operations. When the currently displayed interactive interface contains service videos, the terminal sends a video recommendation request of the service account to the server. At least the account identifier of this service account is carried in this video recommendation request. Optionally, if the interactive interface is also associated with a certain virtual character, the role identifier of this virtual character can also be carried in this video recommendation request.

[0234] Hereinafter, taking the service client as a game client as an example, at this time the service account is a game account, and several possible interactive interfaces will be introduced separately.

[0235] (a) Function interface of the virtual character

[0236] Among them, this function interface is used to introduce this virtual character or configure the equipment of this virtual character.

[0237] Taking a MOBA game as an example, after a user logs in to the game account in the game client, the game lobby interface will be opened. In the game lobby interface, the user can start a matchmaking session or operate to open the function interface of any virtual character to view the basic background information introduction of the virtual character, or further configure the equipment plan of the virtual character. After completing the configuration of the equipment plan, if the user starts a matchmaking session and selects the virtual character as the main controlled virtual character, the configured equipment plan in the function interface can be switched with one key, so as to facilitate the purchase of equipment using the pre-configured equipment plan in the virtual game and improve the interaction efficiency of the user in the virtual game.

[0238] By providing business videos (i.e., game videos) in the function interface of the virtual character, when the user opens the function interface, the user can not only view the basic background information introduction of the virtual character or configure the equipment plan of the virtual character, but also support one-key playback of the game videos personalized for the game account, which increases the amount of information carried by the function interface of the virtual character, facilitates the user to obtain more skill information by watching game videos, and improves the information acquisition efficiency and human-computer interaction efficiency of the user.

[0239] In some embodiments, when the interaction interface is the function interface of the virtual character, due to the high correlation between the function interface and the virtual character, the account identifier of the business account and the character identifier of the virtual character can be carried in the video recommendation request sent to the server when the function interface of the virtual character is opened. Since the user opens the function interface of the virtual character, the viewing intention is usually more suitable for the virtual character. By adding the character identifier of the virtual character to the video recommendation request and controlling the server to only recommend the target video associated with the virtual character to the terminal, the recommendation accuracy of the target video can be further improved.

[0240] Figure 10 It is a schematic diagram of the function interface of a virtual character provided by an embodiment of the present application. As Figure 10 shown, in the function interface 1000 of the virtual character, there are provided the basic background information introduction of the virtual character X, skill icons, character positioning, and a character model. In addition, in the lower left corner of the function interface 1000, there is also provided a playback control 1001 for a game video. When the game client loads the function interface 1000, a video recommendation request needs to be sent to the server, and the account identifier of the game account and the character identifier of the virtual character X are carried in the video recommendation request. In addition, the user can also enter the practice mode through the practice control 1002 on the right side and automatically configure the virtual character X as the main controlled virtual character in the practice mode.

[0241] (b) Preparation interface before the start of the practice mode or the virtual environment after the start

[0242] In some embodiments, the user can directly open the practice mode in the game lobby interface, and a preparation interface before the start of the practice mode is displayed. The user can select the main controlled virtual character for the game in the preparation interface, or switch the already selected main controlled virtual character. After completing the selection of the main controlled virtual character, the user can further select the virtual skills, talents or runes to be carried, as well as the equipment scheme. After the user's confirmation operation, the virtual environment after the start will be opened. Among them, the above-mentioned virtual skills are different from the inherent character skills of the virtual character, but can be switched in different games, and are also called summoner skills. Alternatively, the user can also directly switch to the practice mode from the function interface of any virtual character, and a preparation interface before the start of the practice mode is displayed. In the preparation interface, the virtual character can be automatically preselected as the main controlled virtual character. Of course, the user can also switch to other main controlled virtual characters, further select the virtual skills, talents or runes to be carried, as well as the equipment scheme. After the user's confirmation operation, the virtual environment after the start will be opened.

[0243] On the one hand, by providing business videos (i.e., game videos) in the preparation interface before the start of the practice mode, after the user enters the practice mode, they can start watching the recommended game videos from the preparation interface, learn the relevant operations of the virtual character through the business videos, increase the amount of information carried by the preparation interface of the practice mode, facilitate the user to obtain more skill information by watching the game videos, and improve the user's information acquisition efficiency and human-computer interaction efficiency.

[0244] On the other hand, by providing business videos (i.e., game videos) in the virtual environment after the start of the practice mode, even after the user enters the practice mode, since the practice mode is usually a human-machine confrontation, the user can watch the game videos while operating the main controlled virtual character to carry out interactive practice, increase the amount of information carried by the virtual environment of the practice mode, facilitate the user to practice the operation skills of the virtual character while watching, and improve the user's information acquisition efficiency and human-computer interaction efficiency.

[0245] In some embodiments, when the interactive interface is the preparation interface before the start of the practice mode or the virtual environment after the start, due to the high correlation between the practice mode and the selected master virtual character, after opening the preparation interface or loading the virtual environment, the video recommendation request sent to the server may carry the account identifier of the service account and the character identifier of the virtual character mastered by the service account. Since the user usually opens the practice mode to practice the operation skills of the master virtual character, their viewing intention is usually more suitable for the master virtual character. By adding the character identifier of the master virtual character to the video recommendation request and controlling the server to only recommend target videos associated with the master virtual character to the terminal, the recommendation accuracy of the target videos can be further improved.

[0246] Figure 11 FIG. is a schematic diagram of a virtual environment after the start of a practice mode provided by an embodiment of the present application. As Figure 11 shown, it shows the virtual environment 1100 in the practice mode. In the virtual environment 1100, there is provided a character model 1101 of the master virtual character Y and a playback control 1102 for game videos. When the game client loads the virtual environment in the practice mode, it needs to send a video recommendation request to the server, and the video recommendation request carries the account identifier of the game account and the character identifier of the master virtual character Y. In addition, the user can also switch to full-screen display of the game video through the full-screen display control 1103 provided in the playback control 1102.

[0247] After the user performs a trigger operation on the full-screen display control 1103, as Figure 12 shown, Figure 12 FIG. is a schematic diagram of full-screen display of a game video provided by an embodiment of the present application. In the game client, the game video 1200 will be displayed in full screen. Optionally, the bullet screen information of the game video can be automatically hidden in the small window mode, and the user can view the bullet screen information in the game video through the full-screen mode.

[0248] (c) Notification interface of the client version

[0249] Among them, the notification interface is used to prompt the updated content of the client version.

[0250] In some embodiments, the user can directly open the notification interface of the client version in the game lobby interface, and the update content of the client version can be introduced in this notification interface. Since there may be many business videos related to version updates, for example, the official account will release version introduction videos, and different KOL accounts will also release their own version analysis videos. Thus, the business videos displayed in the notification interface can also be personalized recommended by the server for different business accounts, so as to facilitate the user to understand the changes in the new version, assist the user to adapt to the changes in the new version faster, improve the user's information acquisition efficiency and human-computer interaction efficiency, and also improve the user's gaming experience.

[0251] Figure 13 is a schematic diagram of a notification interface of a client version provided by an embodiment of the present application. As Figure 13 shown, in the notification interface 1300 of the client version, a playback control 1301 of a game video is displayed. When the game client loads the notification interface 1300 of the client version, it needs to send a video recommendation request to the server, and the account identifier of the game account is carried in the video recommendation request. The user can use the playback control 1301 to start playing the game video personalized recommended by the server.

[0252] 902. The server responds to the video recommendation request and returns video recommendation information of the target video to the terminal, and the matching degree of the target video with the business account meets the recommendation conditions.

[0253] The above step 902 is the same as steps 501-506 in the foregoing embodiments. Here, taking the server returning the video recommendation information of the target video to the terminal as an example for illustration, details are not described again.

[0254] 903. The terminal receives the video recommendation information returned by the server based on the video recommendation request, and the video recommendation information indicates the target video.

[0255] In some embodiments, the terminal receives the video recommendation information of the target video returned by the server. Since there may be more than one target video, the received video recommendation information may also be more than one, that is, the number of received video recommendation information depends on the number of target videos recommended by the server. The embodiments of the present application do not specifically limit this.

[0256] Still taking Figure 10 as an example for illustration, in the case where the server recommends multiple target videos to the game account, video switching controls 1003 and 1004 are also displayed in the playback control 1001. The user can use the video switching control 1003 to switch to the previous target video, or use the video switching control 1004 to switch to the next target video.

[0257] 904. The terminal pulls the target video from the server based on the video recommendation information.

[0258] In some embodiments, the terminal pulls the target video from the server based on the video recommendation information of the target video received in step 903. Hereinafter, the loading method of a single target video will be taken as an example for illustration. In the case of multiple target videos, the loading methods of each target video are the same and will not be elaborated.

[0259] In some embodiments, when the server includes a video platform and a video CDN module, the terminal can pull the cover image of the target video from the video CDN module according to the video recommendation information, and render the cover image into the interactive interface. After the user performs the play operation on the target video, the terminal further loads the video stream of the target video in a streaming manner from the video CDN module. Since there are multiple CDN servers distributed in various places in the video CDN module, the terminal can find the nearest CDN server through the video CDN module and pull the target video from the CDN server nearby, improving the transmission efficiency of the target video. It should be noted that it is possible that the target video is not cached in the CDN server, then the CDN server needs to pull the target video from the video platform first and then send it to the terminal.

[0260] In other embodiments, if the video platform does not adopt a CDN architecture, the terminal can also directly pull the cover image of the target video from the video platform, render the cover image into the interactive interface. After the user performs the play operation on the target video, the terminal further loads the video stream of the target video in a streaming manner from the video platform. It can also play while loading, improving the loading efficiency of the target video.

[0261] Still taking Figure 10 as an example for illustration, after the terminal receives the video recommendation information of the target video recommended for the game account personalized, and according to the video recommendation information, it can pull the cover image of the target video from the video CDN module, and render the cover image into the playback control 1001. If the user is interested in the target video, the user can perform a play operation through the playback control 1001, so that the terminal, in response to the play operation, continues to pull the target video from the video CDN module according to the video recommendation information. The target video can adopt a streaming transmission method, so that it can play while loading, improving the loading efficiency of the target video.

[0262] 905. The terminal displays the target video in the interactive interface.

[0263] In some embodiments, after the terminal pulls the cover image of the target video, it displays the playback control of the target video in the interactive interface and renders the cover image on the playback control. The user can perform a trigger operation on the playback control to enable the service client to play the target video. Since the target video is a video stream, it can be played while being loaded through streaming transmission, improving the loading efficiency of the target video.

[0264] Still taking Figure 10 as an example, after the user plays the target video, the user can also give a like to the target video through the like control 1005. Or, the user can also switch the target video to full-screen playback through the full-screen display control 1006. Or, the user can also switch the target video to floating-window playback through the floating-window display control 1007. In the floating-window playback mode, even if the user operates to close or exit the function interface 1000 of the virtual character, it does not affect the continuous playback of the target video in the floating-window playback mode, improving the playback smoothness of the target video. In the small-window playback mode, when the user exits the function interface 1000 of the virtual character, the target video will also be closed. However, when the user opens the function interface 1000 of the virtual character again, the target video can be played from the beginning, or the playback progress of the target video when it exited last time can be recorded and automatically resumed from the playback progress when it exited last time. The embodiments of the present application do not make specific limitations on this.

[0265] All the above optional technical solutions can be combined arbitrarily to form the optional embodiments of the present disclosure, which will not be elaborated one by one here.

[0266] The method provided by the embodiments of the present application displays the target video recommended for the service account personalized in the interactive interface of the service client. The user can view the target video with a high degree of matching without switching to other clients, improving the convenience for the user to view the target video. The user can learn operation skills while watching the target video, improving the user's information acquisition efficiency and human-computer interaction efficiency, and improving the user experience.

[0267] Figure 14 is a schematic structural diagram of a recommended device for service videos provided by the embodiments of the present application. As Figure 14 shown, the device includes:

[0268] An acquisition module 1401, configured to acquire the service behavior information of the service account in response to a video recommendation request of the service account, where the service behavior information is information on the service behavior performed by the service account in the service client;

[0269] The first extraction module 1402 is configured to extract the service behavior characteristics of the service account based on the service behavior information, where the service behavior characteristics characterize the characteristics of the service behavior executed by the service account in the service client;

[0270] The second extraction module 1403 is configured to extract the video attribute characteristics of each of the multiple service videos based on the video attribute information of the multiple service videos, where the video attribute characteristics characterize the characteristics of the video attributes of the service videos;

[0271] The recommendation module 1404 is configured to recommend a target video for the service account from the multiple service videos based on the service behavior characteristics and the video attribute characteristics, where the matching degree between the target video and the service account meets the recommendation conditions.

[0272] The device provided by the embodiment of the present application, when receiving a video recommendation request of a service account, on the one hand extracts the service behavior characteristics of the service account, and on the other hand extracts the video attribute characteristics of each of the multiple service videos, so that by using the service behavior characteristics and each video attribute characteristic, personalized service video recommendation can be performed for the current service account, a target video whose matching degree with the service account meets the recommendation conditions can be determined from the multiple service videos, and the target video is recommended to the service account, thereby ensuring that the target video and the service account are relatively well-matched, so that the target video as much as possible meets the viewing intention of the service account, and the recommendation accuracy of the service video is improved.

[0273] In some embodiments, based on Figure 14 the composition of the device, the recommendation module 1404 includes:

[0274] An input unit, configured to input the service behavior characteristics into a video recommendation model, where the video recommendation model is used to predict a video to be recommended for the service account based on the service behavior characteristics of the service account;

[0275] A determination unit, configured to determine the target video to be recommended from the multiple service videos through the video recommendation model based on the video attribute characteristics of each of the multiple service videos;

[0276] A recommendation unit, configured to recommend the target video to the service account.

[0277] In some embodiments, the determination unit is configured to:

[0278] Based on the service behavior characteristics and the video attribute characteristics of each service video, determine the matching probability between the service account and each service video, where the matching probability indicates the matching degree between the service account and the service video;

[0279] Determine the target video from the multiple business videos based on the matching probability between the business account and each business video.

[0280] In some embodiments, the recommendation unit is configured to:

[0281] Send video recommendation information of the target video to the terminal logged in by the business account, so that the terminal pulls the target video based on the video recommendation information; or, send the target video to the terminal.

[0282] In some embodiments, the first extraction module 1402 is configured to:

[0283] Input the business behavior information into a behavior feature extraction model, and extract features from the business behavior information through the behavior feature extraction model to obtain the business behavior features. The behavior feature extraction model is used to extract business behavior features of a business account.

[0284] In some embodiments, based on Figure 14 the device composition, the device further includes:

[0285] A query module, configured to query a candidate video list associated with the virtual character based on the character identifier in the case where the character identifier of the virtual character is carried in the video recommendation request;

[0286] A determination module, configured to determine the multiple business videos from the multiple candidate videos included in the candidate video list.

[0287] In some embodiments, the determination module is configured to:

[0288] Create a video pulling timing task for the virtual character, where the video pulling timing task is used to pull the candidate video list associated with the virtual character at regular intervals;

[0289] Based on the video pulling timing task, pull the candidate video list at intervals of a target duration. The candidate video list contains multiple candidate videos associated with the virtual character;

[0290] Determine the multiple candidate videos that have passed the review from the candidate video list as the multiple business videos.

[0291] In some embodiments, the second extraction module 1403 is configured to:

[0292] Input the video attribute information of each business video into a video feature extraction model, and extract features from the video attribute information through the video feature extraction model to obtain the video attribute features of the business video. The video feature extraction model is used to extract video attribute features of business videos.

[0293] In some embodiments, the service behavior information includes at least one of the operation behavior information of the service account or the historical video record. The operation behavior information indicates the accumulated game behavior information of the service account in the virtual game, and the historical video record indicates the interaction behavior information of the service account for the recommended historical videos.

[0294] In some embodiments, the video attribute information includes at least one of the publisher account information of the service video, the video popularity information, the video details information, or the associated video information.

[0295] All of the above optional technical solutions can be combined arbitrarily to form optional embodiments of the present disclosure, which will not be elaborated one by one here.

[0296] It should be noted that when the recommended device for the service video provided in the above embodiments recommends the target video, only the division of the above functional modules is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the computer device (such as a server) is divided into different functional modules to complete all or part of the functions described above. In addition, the recommended device for the service video provided in the above embodiments and the embodiments of the recommended method for the service video belong to the same concept. For the specific implementation process, please refer to the embodiments of the recommended method for the service video, which will not be elaborated here.

[0297] Figure 15 It is a schematic structural diagram of a display device for a service video provided by an embodiment of the present application, as Figure 15 shown. The device includes:

[0298] A sending module 1501, configured to send a video recommendation request of the service account when the interactive interface of the service client includes a service video. The video recommendation request is used to obtain a target video recommended for the service account. The service account is the account logged in to the service client, and the matching degree between the target video and the service account meets the recommendation conditions;

[0299] A receiving module 1502, configured to receive video recommendation information returned based on the video recommendation request. The video recommendation information indicates the target video;

[0300] A pulling module 1503, configured to pull the target video based on the video recommendation information;

[0301] A display module 1504, configured to display the target video in the interactive interface.

[0302] The device provided by the embodiment of the present application displays a target video recommended for personalization of a service account in the interaction interface of the service client. Without switching to other clients, users can watch the target video with a relatively high matching degree, which improves the convenience of users watching the target video. Users can learn operation skills while watching the target video, which improves the information acquisition efficiency and human-computer interaction efficiency of users and improves the user experience.

[0303] In some embodiments, the interaction interface includes at least one of the following:

[0304] The function interface of the virtual character, which is used to introduce the virtual character or configure the equipment of the virtual character;

[0305] The preparation interface before the start of the practice mode or the virtual environment after the start;

[0306] The notification interface of the client version, which is used to prompt the updated content of the client version.

[0307] In some embodiments, when the interaction interface is the function interface of the virtual character, the role identifier of the virtual character is carried in the video recommendation request; or, when the interaction interface is the preparation interface before the start of the practice mode or the virtual environment after the start, the role identifier of the virtual character controlled by the service account is carried in the video recommendation request.

[0308] All the above optional technical solutions can be combined arbitrarily to form optional embodiments of the present disclosure, which will not be elaborated one by one here.

[0309] It should be noted that when the service video display device provided in the above embodiment displays the target video, only the above-mentioned division of each functional module is used for illustration. In actual application, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the computer device (such as a terminal) is divided into different functional modules to complete all or part of the functions described above. In addition, the service video display device provided in the above embodiment and the service video display method embodiment belong to the same concept, and the specific implementation process can be seen in the service video display method embodiment, which will not be elaborated here.

[0310] Figure 16 It is a schematic structural diagram of a computer device provided by the embodiment of the present application, as Figure 16As shown, the computer device is taken as an example of the terminal 1600 for illustration. Optionally, the device types of the terminal 1600 include: smart phones, tablet computers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers or desktop computers. The terminal 1600 may also be referred to by other names such as user equipment, portable terminal, laptop terminal, desktop terminal, etc.

[0311] Generally, the terminal 1600 includes a processor 1601 and a memory 1602.

[0312] Optionally, the processor 1601 includes one or more processing cores, such as a 4-core processor, an 8-core processor, etc. Optionally, the processor 1601 is implemented in at least one of the following hardware forms: DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). In some embodiments, the processor 1601 includes a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 1601 integrates a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1601 further includes an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0313] In some embodiments, the memory 1602 includes one or more computer-readable storage media, optionally non-transitory. Optionally, the memory 1602 further includes high-speed random access memory, as well as non-volatile memory, such as one or more disk storage devices, flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 1602 is used to store at least one program code for being executed by the processor 1601 to implement the display method of the service video provided in various embodiments of the present application.

[0314] In some embodiments, the terminal 1600 may further optionally include: a peripheral device interface 1603 and at least one peripheral device. The processor 1601, the memory 1602, and the peripheral device interface 1603 can be connected through a bus or signal lines. Each peripheral device can be connected to the peripheral device interface 1603 through a bus, signal lines, or a circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 1604, a display screen 1605, a camera assembly 1606, an audio circuit 1607, and a power supply 1608.

[0315] The peripheral device interface 1603 can be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 1601 and the memory 1602. In some embodiments, the processor 1601, the memory 1602, and the peripheral device interface 1603 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 1601, the memory 1602, and the peripheral device interface 1603 are implemented on a separate chip or circuit board, and this embodiment does not limit this.

[0316] The radio frequency circuit 1604 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 1604 communicates with the communication network and other communication devices through electromagnetic signals. The radio frequency circuit 1604 converts electrical signals into electromagnetic signals for transmission, or converts the received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 1604 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and so on. Optionally, the radio frequency circuit 1604 communicates with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to: metropolitan area network, various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area network, and / or WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit 1604 further includes a circuit related to NFC (Near Field Communication), which is not limited in this application.

[0317] The display screen 1605 is used to display the UI (User Interface). Optionally, the UI includes graphics, text, icons, videos, and any combination thereof. When the display screen 1605 is a touch display screen, the display screen 1605 also has the ability to collect touch signals on or above the surface of the display screen 1605. The touch signals can be input as control signals to the processor 1601 for processing. Optionally, the display screen 1605 is further used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there is one display screen 1605, which is set on the front panel of the terminal 1600; in other embodiments, there are at least two display screens 1605, which are respectively set on different surfaces of the terminal 1600 or are in a folding design; in still other embodiments, the display screen 1605 is a flexible display screen, which is set on the curved surface or the folding surface of the terminal 1600. Even more optionally, the display screen 1605 is set to an irregular non-rectangular shape, that is, a special-shaped screen. Optionally, the display screen 1605 is prepared using materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0318] The camera component 1606 is used to collect images or videos. Optionally, the camera component 1606 includes a front camera and a rear camera. Generally, the front camera is disposed on the front panel of the terminal, and the rear camera is disposed on the back of the terminal. In some embodiments, there are at least two rear cameras, which are any one of a main camera, a depth camera, a wide-angle camera, and a telephoto camera, so as to implement the function of background blurring by fusing the main camera and the depth camera, panoramic shooting and VR (Virtual Reality) shooting functions or other fusion shooting functions by fusing the main camera and the wide-angle camera. In some embodiments, the camera component 1606 further includes a flash. Optionally, the flash is a single-color temperature flash or a dual-color temperature flash. The dual-color temperature flash refers to a combination of a warm light flash and a cold light flash for light compensation under different color temperatures.

[0319] In some embodiments, the audio circuit 1607 includes a microphone and a speaker. The microphone is used to collect sound waves of the user and the environment, and convert the sound waves into electrical signals and input them to the processor 1601 for processing, or input them to the radio frequency circuit 1604 to implement voice communication. For the purpose of stereo collection or noise reduction, there are multiple microphones, which are respectively disposed at different parts of the terminal 1600. Optionally, the microphone is an array microphone or an omnidirectional collection type microphone. The speaker is used to convert the electrical signal from the processor 1601 or the radio frequency circuit 1604 into sound waves. Optionally, the speaker is a traditional thin film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert the electrical signal into sound waves audible to humans, but also convert the electrical signal into sound waves inaudible to humans for uses such as ranging. In some embodiments, the audio circuit 1607 further includes a headphone jack.

[0320] The power supply 1608 is used to supply power to each component in the terminal 1600. Optionally, the power supply 1608 is alternating current, direct current, a primary battery or a rechargeable battery. When the power supply 1608 includes a rechargeable battery, the rechargeable battery supports wired charging or wireless charging. The rechargeable battery is also used to support fast charging technology.

[0321] In some embodiments, the terminal 1600 further includes one or more sensors 1610. The one or more sensors 1610 include but are not limited to: an acceleration sensor 1611, a gyroscope sensor 1612, a pressure sensor 1613, an optical sensor 1614, and a proximity sensor 1615.

[0322] In some embodiments, the acceleration sensor 1611 detects the magnitudes of accelerations on the three coordinate axes of the coordinate system established by the terminal 1600. For example, the acceleration sensor 1611 is used to detect the components of the gravitational acceleration on the three coordinate axes. Optionally, the processor 1601 controls the display screen 1605 to display the user interface in a landscape view or a portrait view according to the gravitational acceleration signal collected by the acceleration sensor 1611. The acceleration sensor 1611 is also used to collect game or user motion data.

[0323] In some embodiments, the gyroscope sensor 1612 detects the body direction and rotation angle of the terminal 1600. The gyroscope sensor 1612 cooperates with the acceleration sensor 1611 to collect the 3D actions of the user on the terminal 1600. The processor 1601 implements the following functions according to the data collected by the gyroscope sensor 1612: motion sensing (such as changing the UI according to the user's tilting operation), image stabilization during shooting, game control, and inertial navigation.

[0324] Optionally, the pressure sensor 1613 is disposed on the side frame of the terminal 1600 and / or the lower layer of the display screen 1605. When the pressure sensor 1613 is disposed on the side frame of the terminal 1600, it can detect the holding signal of the user on the terminal 1600, and the processor 1601 performs left / right hand recognition or shortcut operations according to the holding signal collected by the pressure sensor 1613. When the pressure sensor 1613 is disposed on the lower layer of the display screen 1605, the processor 1601 controls the operable controls on the UI interface according to the pressure operation of the user on the display screen 1605. The operable controls include at least one of a button control, a scroll bar control, an icon control, and a menu control.

[0325] The optical sensor 1614 is used to collect the ambient light intensity. In one embodiment, the processor 1601 controls the display brightness of the display screen 1605 according to the ambient light intensity collected by the optical sensor 1614. Specifically, when the ambient light intensity is high, the display brightness of the display screen 1605 is increased; when the ambient light intensity is low, the display brightness of the display screen 1605 is decreased. In another embodiment, the processor 1601 also dynamically adjusts the shooting parameters of the camera module 1606 according to the ambient light intensity collected by the optical sensor 1614.

[0326] The proximity sensor 1615, also known as the distance sensor, is usually disposed on the front panel of the terminal 1600. The proximity sensor 1615 is used to collect the distance between the user and the front of the terminal 1600. In one embodiment, when the proximity sensor 1615 detects that the distance between the user and the front of the terminal 1600 is gradually decreasing, the processor 1601 controls the display screen 1605 to switch from the lit state to the off state; when the proximity sensor 1615 detects that the distance between the user and the front of the terminal 1600 is gradually increasing, the processor 1601 controls the display screen 1605 to switch from the off state to the lit state.

[0327] Those skilled in the art can understand that Figure 16 the structure shown in does not constitute a limitation on the terminal 1600, and it can include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0328] Figure 17 is a schematic structural diagram of another computer device provided by an embodiment of the present application. As Figure 17 shown, taking the computer device as the server 1700 as an example for illustration, the server 1700 may vary greatly due to different configurations or performances. The server 1700 includes one or more processors (Central Processing Units, CPUs) 1701 and one or more memories 1702. Among them, at least one computer program is stored in the memory 1702, and the at least one computer program is loaded and executed by the one or more processors 1701 to implement the recommended method for business videos provided in the above various embodiments. Optionally, the server 1700 also has components such as a wired or wireless network interface, a keyboard, and an input / output interface for input and output. The server 1700 also includes other components for implementing the functions of the device, which will not be elaborated here.

[0329] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including at least one computer program. The at least one computer program can be executed by a processor in the terminal to complete the recommended method for business videos or the display method for business videos in the above various embodiments. For example, the computer-readable storage medium includes ROM (Read-Only Memory), RAM (Random-Access Memory), CD-ROM (Compact Disc Read-Only Memory), magnetic tape, floppy disk, and optical data storage devices, etc.

[0330] In an exemplary embodiment, a computer program product or a computer program is further provided, including one or more program codes stored in a computer-readable storage medium. One or more processors of a computer device can read the one or more program codes from the computer-readable storage medium, and the one or more processors execute the one or more program codes, so that the computer device can execute to complete the recommendation method or the display method of the business video in the above embodiments.

[0331] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware, or can be completed by instructing relevant hardware through a program. Optionally, the program is stored in a computer-readable storage medium. Optionally, the above-mentioned storage medium is a read-only memory, a disk, an optical disc, etc.

[0332] The above are only optional embodiments of the present application, and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for recommending business videos, characterized in that, The method includes: In response to a video recommendation request of a business account, obtaining business behavior information of the business account, where the business behavior information is information on business behaviors performed by the business account in a business client; Based on the business behavior information, extracting business behavior features of the business account, where the business behavior features characterize the characteristics of the business behaviors performed by the business account in the business client; Based on video attribute information of multiple business videos, extracting video attribute features of each of the multiple business videos, where the video attribute features characterize the characteristics of the video attributes of the business videos; Based on the business behavior features and the video attribute features, recommending a target video for the business account from the multiple business videos, where the matching degree between the target video and the business account meets the recommendation conditions.

2. The method according to claim 1, wherein The recommending a target video for the business account from the multiple business videos based on the business behavior features and the video attribute features includes: Inputting the business behavior features into a video recommendation model, where the video recommendation model is used to predict videos to be recommended for the business account based on the business behavior features of the business account; Through the video recommendation model, determining the target video to be recommended from the multiple business videos based on the video attribute features of each of the multiple business videos; Recommending the target video to the business account.

3. The method according to claim 2, wherein The determining the target video to be recommended from the multiple business videos based on the video attribute features of each of the multiple business videos includes: Based on the business behavior features and the video attribute features of each business video, determining a matching probability between the business account and each business video, where the matching probability indicates the matching degree between the business account and the business video; Based on the matching probability between the business account and each business video, determining the target video from the multiple business videos.

4. The method according to claim 2, characterized in that The recommending the target video to the business account includes: Sending video recommendation information of the target video to the terminal logged in by the business account, so that the terminal pulls the target video based on the video recommendation information; or, sending the target video to the terminal.

5. The method according to claim 1, characterized in that The extracting business behavior features of the business account based on the business behavior information includes: Inputting the business behavior information into a behavior feature extraction model, and through the behavior feature extraction model, performing feature extraction on the business behavior information to obtain the business behavior features, where the behavior feature extraction model is used to extract business behavior features of a business account.

6. The method according to claim 1, characterized in that, The method further includes: When a role identifier of a virtual character is carried in the video recommendation request, querying a candidate video list associated with the virtual character based on the role identifier; Determining the multiple business videos from the multiple candidate videos included in the candidate video list.

7. The method according to claim 6, characterized in that, The determining the multiple business videos from the multiple candidate videos included in the candidate video list includes: Creating a video pulling timing task for the virtual character, where the video pulling timing task is used to pull the candidate video list associated with the virtual character at regular intervals; Based on the video pulling timing task, every target duration, pull the candidate video list, where the candidate video list contains multiple candidate videos associated with the virtual character; From the candidate video list, determine the multiple candidate videos that have passed the review as the multiple service videos.

8. The method according to claim 1, characterized in that The extracting the video attribute features of each of the multiple service videos based on the video attribute information of the multiple service videos includes: Input the video attribute information of each service video into a video feature extraction model, and through the video feature extraction model, perform feature extraction on the video attribute information to obtain the video attribute features of the service video. The video feature extraction model is used to extract the video attribute features of service videos.

9. The method according to any one of claims 1 to 8, characterized in that The service behavior information includes at least one of the operation behavior information of the service account or the historical video records. The operation behavior information indicates the accumulated game behavior information of the service account in the virtual game, and the historical video records indicate the interaction behavior information of the service account for the recommended historical videos.

10. The method according to any one of claims 1 to 8, characterized in that, The video attribute information includes at least one of the publisher account information, video popularity information, video details information, or associated video information of the service video.

11. A method for displaying a service video, characterized in that, The method includes: When the interactive interface of the service client includes a service video, send a video recommendation request for the service account. The video recommendation request is used to obtain a target video recommended for the service account. The service account is the account logged in to the service client, and the matching degree between the target video and the service account meets the recommendation conditions; Receive the video recommendation information returned based on the video recommendation request. The video recommendation information indicates the target video; Based on the video recommendation information, pull the target video; Display the target video in the interactive interface.

12. The method according to claim 11, wherein The interactive interface includes at least one of the following: The function interface of the virtual character, which is used to introduce the virtual character or configure the equipment of the virtual character; The preparation interface before the start of the practice mode or the virtual environment after the start; The notification interface of the client version, which is used to prompt the updated content of the client version.

13. The method according to claim 12, characterized in that, When the interactive interface is the function interface of the virtual character, the role identifier of the virtual character is carried in the video recommendation request; or, when the interactive interface is the preparation interface before the start of the practice mode or the virtual environment after the start, the role identifier of the virtual character controlled by the service account is carried in the video recommendation request.

14. A recommendation device for business videos, characterized in that, The device includes: An acquisition module, configured to acquire the service behavior information of the service account in response to a video recommendation request of the service account. The service behavior information is the information of the service behavior performed by the service account in the service client; A first extraction module, configured to extract the service behavior features of the service account based on the service behavior information. The service behavior features characterize the characteristics of the service behavior performed by the service account in the service client; A second extraction module, configured to extract video attribute features of each of the multiple service videos based on video attribute information of the multiple service videos, where the video attribute features characterize characteristics of video attributes of the service videos; A recommendation module, configured to recommend a target video for the service account from the multiple service videos based on the service behavior features and the video attribute features, where the matching degree between the target video and the service account meets the recommendation conditions.

15. A display device for business videos, characterized in that, The apparatus includes: A sending module, configured to send a video recommendation request of a service account when an interactive interface of a service client includes a service video, where the video recommendation request is used to obtain a target video recommended for the service account, the service account is an account logged in to the service client, and the matching degree between the target video and the service account meets the recommendation conditions; A receiving module, configured to receive video recommendation information returned based on the video recommendation request, where the video recommendation information indicates the target video; A pulling module, configured to pull the target video based on the video recommendation information; A display module, configured to display the target video in the interactive interface.

16. A computer device, characterized in that, The computer device includes one or more processors and one or more memories, and at least one computer program is stored in the one or more memories, and the at least one computer program is loaded and executed by the one or more processors to implement the recommendation method for service videos according to any one of claims 1 to 10; or, the display method for service videos according to any one of claims 11 to 13.

17. A storage medium, characterized in that, At least one computer program is stored in the storage medium, and the at least one computer program is loaded and executed by a processor to implement the recommendation method for service videos according to any one of claims 1 to 10; or, the display method for service videos according to any one of claims 11 to 13.

18. A computer program product, characterized in that, The computer program product includes at least one computer program, and the at least one computer program is loaded and executed by a processor to implement the recommendation method for service videos according to any one of claims 1 to 10; or, the display method for service videos according to any one of claims 11 to 13.

Images