Virtual digital human interaction service method and device

Abstract

The invention relates to a virtual digital human interaction service method and device, and relates to the technical field of computers. The method comprises the following steps: acquiring a user data set of a first interaction platform, summarizing high-frequency attribute features in the user data set, and determining a plurality of mainstream user categories; creating a virtual digital human account for each mainstream user category, and establishing a personal friend relationship with the user account under the mainstream user category based on the virtual digital human account; setting a virtual attribute feature of the virtual digital human account based on the attribute feature of the friend user account established by the virtual digital human account; obtaining friend dynamic information published by a friend user account through the virtual digital person account, and identifying effective dynamic information matched with the virtual attribute characteristics; and performing data processing on the effective dynamic information to obtain secondary dynamic information, and issuing the secondary dynamic information through a virtual digital human account. By adopting the method, the information sharing efficiency and the user experience of the interaction platform can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a virtual digital human interactive service method and device. Background Art

[0002] In existing product shopping guide platforms, common user traffic diversion methods usually include content push based on recommendation algorithms and interactive marketing in social communities. Specifically, the platform will collect data such as users' browsing history and shopping habits, use recommendation algorithms to generate product information that meets user preferences, and push it to the user's personal page; at the same time, the platform will also encourage users to share shopping experiences or dynamics in the community, and through social behaviors such as comments and likes, form interactions between users to enhance platform activity. Some platforms will introduce brand promotion accounts to directly publish product-related promotional content to attract user attention. However, these methods usually rely on the active participation of users and cannot fully tap the potential value of user-posted content. The form of interaction between users and the platform is also relatively simple.

[0003] There are still many problems with the existing technical solutions in practical applications. First, the accuracy of the recommendation algorithm is limited by the amount of data and model capabilities, resulting in the incomplete match between the recommended content and the actual needs of users, thereby reducing the drainage effect. Secondly, the interactive behavior of social communities depends on the activity of users, and the participation of different users varies greatly, making it difficult to continuously form efficient content output and interactive links. In addition, the dynamic information posted by users on the platform usually only stays at the level of one-way communication and is not further processed and utilized, resulting in the platform's failure to fully tap the value of these contents. Finally, the commercial attributes of brand promotion accounts are too obvious, which can easily cause user disgust, thereby reducing user experience. Therefore, there is an urgent need for a new interactive service method that can solve the above-mentioned problems of the existing technology through innovative interactive forms and content processing processes, and enhance the platform's drainage capabilities and user stickiness. Summary of the invention

[0004] Based on this, it is necessary to provide a virtual digital human interactive service method and device, computer equipment, storage medium and computer program product to address the above technical problems.

[0005] In a first aspect, the present application provides a virtual digital human interactive service method, the method comprising: Acquire a user data set of the first interactive platform, summarize high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories; Creating a virtual digital human account for each mainstream user category, and establishing a personal friendship relationship with the user accounts under the mainstream user category based on the virtual digital human account; Based on the attribute characteristics of the friend user accounts that have been established by the virtual digital human account, set the virtual attribute characteristics of the virtual digital human account; Obtaining the dynamic information of friends published by the friend user account through the virtual digital human account, and identifying the valid dynamic information matching the virtual attribute characteristics; The effective dynamic information is processed to obtain secondary dynamic information, and the secondary dynamic information is published through the virtual digital human account.

[0006] In a second aspect, the present application further provides a virtual digital human interactive service device, the device comprising: A feature aggregation and classification module, used to obtain a user data set of the first interactive platform, aggregate high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories; A digital human account creation module, used to create a virtual digital human account for each mainstream user category, and to establish a personal friendship relationship with the user accounts under the mainstream user category based on the virtual digital human account; A virtual attribute setting module, used to set the virtual attribute characteristics of the virtual digital human account based on the attribute characteristics of the friend user accounts that have been established by the virtual digital human account; A dynamic information identification module, used to obtain the dynamic information of friends published by the friend user account through the virtual digital human account, and identify the valid dynamic information matching the virtual attribute characteristics; The secondary dynamic publishing module is used to process the effective dynamic information to obtain secondary dynamic information, and publish the secondary dynamic information through the virtual digital human account.

[0007] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in the first aspect are implemented.

[0008] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0009] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0010] By adopting the virtual digital human interactive service method disclosed in this application, by obtaining the user data set of the first interactive platform, summarizing the high-frequency attribute features in the user data set and determining multiple mainstream user categories, it is possible to accurately characterize the behavioral characteristics of the target user group and provide data support for subsequent interactions. A virtual digital human account is created for each mainstream user category, and a personal friend relationship is established with the user account to simulate the interaction scenario in a real social network. This method can effectively cover individual users in the mainstream user category, significantly improve the relevance and closeness between the virtual digital human and the user, and enhance the participation and credibility of the virtual digital human in the social network.

[0011] By setting the virtual attribute characteristics of the virtual digital human account based on the attribute characteristics of the friend user account, the virtual digital human can present an image and characteristics that match the target user group, thereby improving the matching and attractiveness of the interactive content. At the same time, the virtual digital human account can obtain the dynamic information of friends posted by its friend users, and identify the valid dynamic information that matches the virtual attribute characteristics. This precise screening mechanism can avoid the interference of redundant information or invalid content, and improve the efficiency and quality of dynamic information processing.

[0012] In addition, by processing effective dynamic information to generate secondary dynamic information, and publishing the secondary dynamic information through virtual digital human accounts, the content can be re-disseminated and diffused among user groups. This method not only fully taps the potential value of user dynamic information, but also improves the activity and influence of virtual digital human accounts, allowing them to form a continuous interactive chain in social networks, enhancing users' sense of participation and interactive experience.

[0013] This method realizes the automation and intelligence of virtual digital human interaction by automatically creating virtual digital human accounts based on the attribute characteristics of user data, screening effective dynamic information, and processing and publishing secondary dynamic information. This method reduces the cost and complexity of manual operation while ensuring the accuracy and pertinence of the content. In practical applications, this method can adapt to large-scale user scenarios, improve the platform's responsiveness to user needs, and provide technical support for content distribution and user activity improvement on social networks, forming an efficient, intelligent, and continuously optimized virtual digital human interaction service process. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of a virtual digital human interactive service method in one embodiment; Figure 2 A schematic diagram of the architecture of a virtual digital human interactive service device in one embodiment; Figure 3 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0016] The embodiment of the present application provides a virtual digital human interactive service method, which can be applied to the server of the interactive platform and is applicable to various types of interactive platforms, such as online shopping platforms, live broadcast platforms, dating platforms, video playback platforms, and knowledge sharing platforms. In these platforms, users usually post dynamic content, interact with comments, and browse content through their personal accounts, while the platform improves user experience, enhances user stickiness, and achieves commercial traffic in various ways.

[0017] The server of the interactive platform can obtain user data sets and analyze user behavior characteristics to aggregate and form multiple mainstream user categories. For example, in an online shopping platform, mainstream user categories may include high-frequency purchasers, potential users, or users interested in specific types of goods; in a live broadcast platform, mainstream user categories may include viewers of content in a specific field or users who actively interact; in a dating platform, mainstream user categories may be divided according to user interests or social preferences. Based on these mainstream user categories, the server creates virtual digital human accounts and establishes friendships with actual users as virtual users, providing users with more personalized interactive services.

[0018] The server of the interactive platform can serve as the core execution body of the entire system, and is used to complete the creation of virtual digital human accounts, the establishment of friend relationships, the identification and processing of friend dynamic information, and the publication of virtual digital human dynamic information. Through this server, the platform can further process the dynamic information published by actual users to meet more diverse content display needs. For example, on an online shopping platform, a virtual digital human can publish secondary dynamic information combined with user shopping records to recommend products to other users; on a knowledge sharing platform, a virtual digital human can convert the professional knowledge content published by users into more easy-to-understand dynamic information, thereby attracting more user attention and improving the platform's content dissemination effect.

[0019] In the following, the interactive platform is an online shopping platform or a live shopping guide platform, and the actual execution subject of the method is the server of the interactive platform. Figure 1 A virtual digital human interactive service method is shown and explained, comprising the following steps: Step 101: obtain a user data set of a first interactive platform, summarize high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories.

[0020] Among them, the first interactive platform can be the main service platform for implementing the solution, specifically an online shopping platform or a live shopping guide platform. In relevant scenarios, users can register an account on the first interactive platform and register relevant user data. In addition, the first interactive platform can request to obtain the content interaction records and behavior data of the user on the first interactive platform. In this way, after obtaining the user's authorization, the first interactive platform can collect and summarize the content interaction records and behavior data of a user account with the user account as a unit, and extract attribute features from it and analyze them, so as to form a user data set of the first interactive platform. The user data set may include user data units at user granularity, and a user data unit may correspond to a user account, and specifically may include multiple attribute features that can reflect user data. User data can come from multiple dimensions, including but not limited to: basic user information, such as age, gender, location, etc.; behavioral data, such as the types of browsed goods, search keywords, shopping cart addition records, payment records, return records, etc.; content interaction records, which may include the frequency of participating in comments, likes, and rewards in live shopping guides and content preferences, etc.

[0021] Furthermore, the first interactive platform can extract and summarize high-frequency attribute features from the acquired user data set, such as "electronic product enthusiasts", "beauty experts", "home experts", "foodies", "secondary gamers", "sports enthusiasts", etc. Here, the high-frequency attribute features can be judged based on the number of users with the attribute features, the total amount of consumption, or the degree of activity. Afterwards, the first interactive platform can determine multiple mainstream user categories based on the high-frequency attribute features, so that users can be divided into different categories through clustering algorithms. Here, the mainstream user category can be a category with a certain number of users. A mainstream user category can correspond to one high-frequency attribute feature or multiple high-frequency attribute features. Different mainstream user categories can correspond to repeated high-frequency attribute features. For example, the mainstream user category can be "young women who love beauty and makeup", "foodies who are active late at night", "frequent live shopping", "budget-conscious users", etc. Through the above process, the first interactive platform can provide data support for the creation of subsequent virtual digital human accounts, so that the attributes and behaviors of virtual digital humans are more in line with the needs of target users, so as to improve their drainage effect and user acceptance.

[0022] Step 102: Create a virtual digital human account for each mainstream user category, and establish a personal friendship relationship with the user accounts under the mainstream user category based on the virtual digital human account.

[0023] In implementation, the first interactive platform can create a virtual digital human account for each mainstream user category. The role and attributes of each virtual digital human account can be associated with the mainstream user category. Its specific implementation may include the following aspects: image design. The appearance and style of the virtual digital human can be customized according to the preferences of the target user category. For example, young female users may be more inclined to accept virtual digital humans with a sweet image and who love to share shopping experiences; while male users who are interested in electronic products may prefer professional virtual digital humans; behavioral rules, which may include language style (such as humor, professionalism), interaction frequency (such as regular likes and comments), and ways of sharing content.

[0024] Afterwards, the first interactive platform can establish personal friendships with user accounts under the mainstream user category based on the virtual digital human account. Specifically, the virtual digital human can be recommended to the user account through the platform recommendation algorithm. For example, the platform can recommend a virtual digital human who is good at sharing coupons to a "price-sensitive user"; or the virtual digital human can actively initiate a friend request to the user account, accompanied by a personalized welcome message (such as "Hello, I am an assistant who focuses on electronic product recommendations, and I will bring you more discounts and new product recommendations in the future"). Of course, the virtual digital human can serve as an auxiliary role of the anchor, a product explainer, a content sharer, etc., and attract the active attention of the user account through speech, interaction, etc. Through this process, the virtual digital human account can effectively integrate into the user's social circle and gradually accumulate an initial friend base, laying the foundation for subsequent traffic diversion and content dissemination.

[0025] Step 103, based on the attribute characteristics of the friend user accounts that have been established for the virtual digital human account, set the virtual attribute characteristics of the virtual digital human account.

[0026] In practice, the virtual attribute characteristics of the virtual digital human account are the core of its drainage effect, and its setting can fully consider the attribute characteristics of the friend user account. It can be understood that the setting of mainstream user categories is more for the clustering of user accounts, so that potential friends of a virtual digital human account all potentially have certain identical or similar attribute characteristics. Due to the selection of different user accounts, there is also a large uncertainty in the friend addition results of the virtual digital human account. Therefore, after adding friends, the virtual attribute characteristics of the virtual digital human account can be set based on the attribute characteristics of the friend user account established by the virtual digital human account. Specifically, the platform can extract the attribute characteristics of the friend user account by analyzing the interest preferences and behavioral characteristics of the virtual digital human friends, and then set the virtual attribute characteristics of the virtual digital human account according to the attribute characteristics of all friend user accounts. For online shopping platforms or live shopping guide platforms. Attribute characteristics can be the commodity categories and price ranges that the user account mainly focuses on. For example, on an online shopping platform, if the virtual digital person's friends have repeatedly purchased low- and medium-priced skin care products, their attributes can be set as "cheap skin care experts". If the virtual digital person's friends often watch live broadcasts of digital products, their attributes can be set as "digital product experts". Of course, the operation of adding friends is actually a continuous operation after the virtual digital person account is generated. After adding friends, there will be a loss of friends. Therefore, it can be set to continuously update the virtual attribute characteristics at a certain frequency; at the same time, the virtual attribute characteristics of the virtual digital person account can also be adjusted in real time according to the changes in the friend user's behavior. For example, when a large number of friend users begin to pay attention to different commodity categories or participate in new promotional activities, that is, when the attribute characteristics of the friend user account change significantly, the virtual attribute characteristics of the virtual digital person account can also be updated accordingly. In this way, by setting virtual attribute characteristics that fit the needs of friend users, the virtual digital person can form a more distinct role positioning in the eyes of friend users, thereby improving the user's sense of trust and acceptance.

[0027] Step 104, obtaining the friend dynamic information published by the friend user account through the virtual digital human account, and identifying the valid dynamic information matching the virtual attribute characteristics.

[0028] In implementation, the friend user account can publish relevant dynamic information on the first interactive platform. The virtual digital human account can obtain the relevant dynamic information because of the friend relationship, and can further analyze and identify the relevant dynamic information to determine whether the dynamic information is valid dynamic information matching the virtual attribute characteristics. Specifically, the first interactive platform can obtain the dynamic information published by the friend user through the virtual digital human account, such as the shopping sharing and evaluation content published by the user on the online shopping platform, or the interactive comments participated in on the live shopping guide platform. The first interactive platform can perform semantic analysis on the dynamic information through natural language processing technology and match it with the virtual attribute characteristics of the virtual digital human. For example: if the attribute of the virtual digital human is "affordable skin care expert", it can identify whether there is skin care related content in the friend's dynamic information; if the attribute of the virtual digital human is "digital product expert", it can identify whether the friend's dynamic information involves digital product content. Of course, it is also possible to set weights and scores for dynamic information. For example, if the friend's dynamic involves high-frequency interactive behaviors (such as a large number of comments), or involves popular products or topics, or the dynamic information content is better, the score is higher. When selecting valid dynamic information, the friend's dynamic information with a higher score is preferred. In this way, by effectively identifying dynamic information, virtual digital humans can avoid responding to irrelevant content and further improve interaction efficiency.

[0029] Step 105, data processing is performed on the effective dynamic information to obtain secondary dynamic information, and the secondary dynamic information is published through the virtual digital human account.

[0030] In implementation, the first interactive platform can process the effective dynamic information and publish the processed secondary dynamic information in the form of account dynamics through the virtual digital human account. Afterwards, the first interactive platform can publish the generated secondary dynamic information to the friend dynamic circle through the social function of the virtual digital human account, attracting other friend users to view or further interact. In this way, by publishing secondary dynamic information, the virtual digital human can not only effectively expand the scope of content dissemination, but also guide friend users to participate in the interaction, thereby improving the platform's drainage effect and commercial conversion rate.

[0031] Among them, data processing of effective dynamic information can include data extraction, reorganization, optimization, desensitization and re-creation, etc. Specifically: On the one hand, the content of dynamic information can be adjusted in a targeted manner based on virtual attribute characteristics from the content dimension to improve the attractiveness and relevance of secondary dynamics. For example, if the dynamic information of a friend is "getting up early and running in the morning to help the vitality of the day", and the virtual attribute characteristic is "small home appliance enthusiast", then the processed dynamic information can be added with relevant content of home appliances such as "breakfast machine", "smart alarm clock", and "winter electric heater". At the same time, keyword embedding can also be used to implant keywords that are attractive to friends in effective dynamic information. For example, if the virtual attribute characteristic is "coupon enthusiast", words such as "exclusive discount" and "limited purchase" can be added to the dynamic to increase the attention of friends. It can also be processed in the form of optimization, such as adding pictures, short videos or dynamic special effects to dynamic information to enrich the information expression form to avoid the monotonous text dynamic being easily ignored. It can also optimize the length of dynamic information according to the information content and friend preferences, such as long information is suitable for popular science content, and short information is suitable for promotional and reminder content. You can also inject emotional language into dynamic information to increase the sense of involvement and willingness to interact with friends. For example, if the original dynamic information is "xx new product is on the market, buy it and try it out", the processed secondary dynamic information can be "Guys, try it out, xx new product, guess whether it is a pitfall or a prize this time!" Emotional language needs to be consistent with the role setting of the virtual digital person to avoid contradictions or discomfort. In addition, dynamic information can be integrated into specific scenes to enhance its appeal. For example, if the original dynamic information is "the sound quality of this headset is amazing" and the virtual attribute feature is "game enthusiast", the processed secondary dynamic information can be "wear this headset to play xx game, the enemy character's footsteps and breathing can be heard clearly, and 10 consecutive wins are in progress."

[0032] On the other hand, the content of dynamic information can be adjusted from the time dimension to optimize the timing and frequency of release. For example, it can include publishing dynamic information at a scheduled time based on the active time period analysis of the friend group (such as morning, noon, and night) to avoid releases during inactive periods that may cause dynamic information to be ignored; the release frequency of dynamic information can also be flexibly controlled, such as maintaining a moderate push frequency, for example once a day or three times a week, to avoid too high a frequency that may cause friends to be disgusted or too low a frequency that reduces the sense of presence; the release frequency can also be adjusted based on the interactive performance of the friend group (such as the number of likes, comment activity). On the other hand, dynamic information can also be adjusted from the visibility dimension to accurately target the audience. First, specific content can be published for different friend groups to avoid showing irrelevant information to irrelevant friend users; for example, science and technology information can be published to the "digital enthusiasts" group; and catering-related dynamics can be published to the "food lovers" group; second, specific dynamic information can be opened only to target friends with frequent interactions to avoid redundant dissemination, so as to avoid excessive dissemination of dynamic information and information interference to non-target users.

[0033] On the other hand, data processing can also be performed from the data processing algorithm dimension to improve processing accuracy. Specifically, the following points can be included: First, for hot topics, multiple friend accounts are very likely to publish dynamic information related to the same topic. At this time, key elements can be extracted from multiple dynamic information to generate synthetic secondary dynamic information. Secondly, you can use the form of adding tags or titles to generate labels that fit the virtual attribute characteristics for dynamic information, such as "#fitness check-in", "#smart home", etc., so that friend accounts can understand the main content of the secondary dynamic information more quickly.

[0034] On the other hand, data processing can also be performed from the interactive dimension to enhance the sense of participation and interaction of friend users. For example, interactive content can be introduced in secondary dynamic information, such as voting, Q&A, comment incentives, etc.; or content that guides behavior can be added to secondary dynamic information, such as "Here is the link, everyone can see it for themselves", "Friends are welcome to forward", etc. In addition, when processing data, it is also necessary to consider issues such as friend privacy, positive values, content copyright, and information legitimacy. Specifically, when processing dynamic information, sensitive data and privacy information about friends should be eliminated, and the processed content should be ensured to be in line with mainstream values ​​to avoid disputes caused by inappropriate content. At the same time, when dynamic information involves some creative and high-value content, it is necessary to request friend users to authorize and / or add information sources, or choose to remove relevant content. In addition, other information channels should be combined to ensure that the key content in the dynamic information is true and reliable to avoid a crisis of trust caused by false information. In one embodiment, the virtual attribute feature includes a primary attribute feature and multiple secondary attribute features; the recognition process of the effective dynamic information can be adjusted to: identify the effective dynamic information that matches the primary attribute feature; the processing of the secondary dynamic information can be adjusted to: randomly select a secondary attribute feature, and perform data processing on the effective dynamic information based on the selected secondary attribute feature to obtain secondary dynamic information.

[0035] Among them, virtual attribute characteristics refer to the attribute characteristics of virtual digital human accounts, which are used to imitate the preferences and behaviors of mainstream user categories and friend users. Virtual attribute characteristics include primary attribute characteristics and multiple secondary attribute characteristics. The primary attribute characteristics correspond to the overall characteristics of the mainstream user categories, such as "keen on discounts and promotions", "focusing on electronic products", etc., and are mainly used to match virtual digital people with a wide range of user groups. Secondary attribute characteristics are characteristics that are further subdivided on the basis of the primary attribute characteristics, such as "focusing on smart home devices in electronic products", "preferring activity time to be late at night", etc., which are used to more accurately simulate the needs or interests of different friend users.

[0036] In implementation, virtual digital human accounts can use the main attribute characteristics to conduct preliminary screening with dynamic information posted by friend users. Specifically, the key content of dynamic information, such as keywords, tags, semantic information, etc. in the dynamic information, can be analyzed to determine whether it meets the main attribute characteristics of the virtual digital human. For example, if the main attribute characteristic is "keen on discounts and promotions", keywords in dynamic information such as "limited time discounts" and "big promotions" can be used as matching basis; if the main attribute characteristic is "paying attention to electronic products", the content mentioned in the dynamic information such as "new headphones" and "smart bracelets" can be identified as matching information. After confirming the match, the dynamic information is marked as valid dynamic information and enters the next processing link.

[0037] During the data processing process, the virtual digital human account first selects one from multiple secondary attribute features through random selection to process effective dynamic information. The random selection method can prevent the secondary attributes from being too concentrated on a certain feature, thereby covering more areas of interest of friend users; and if the effective dynamic information is randomly processed, different secondary dynamic information can be generated, which can avoid the content being too single, and also avoid the same effective dynamic information being processed and published multiple times, resulting in frequent screen swiping, affecting the experience of friend users.

[0038] Then, the virtual digital human account can process the valid dynamic information in combination with the selected secondary attribute features to generate secondary dynamic information. For the specific processing method, please refer to the content of step 105. In this way, the random selection of secondary attribute features avoids the monotony of the processed content, making the dynamics of the virtual digital human more diverse, thereby covering more points of interest of friend users; if the same valid dynamic information generates multiple secondary dynamic information, the use of random selection of secondary attributes can also ensure that these secondary dynamic information have different styles, avoiding "repeated screen swiping" from interfering with friend users. Of course, for a valid dynamic information, only one secondary attribute feature can be selected to generate a secondary dynamic information to further ensure the rationality of the release of dynamic information and avoid "screen swiping behavior" that affects the experience of friend users.

[0039] Of course, for specific scenarios, effective dynamic information can be processed jointly based on multiple secondary attributes. For example, the effective dynamic information is "This smart speaker is now half price", the secondary attribute feature 1 is "Prefer high cost performance products", and the secondary attribute feature 2 is "Focus on smart device recommendations". The processed secondary dynamic information can be "High cost performance smart speaker recommendation, now half price, experience smart life, today!" In one embodiment, the method further includes: counting the interaction data of all friend user accounts for secondary dynamic information; assigning a first weight to the corresponding secondary attribute feature according to the interaction data; extracting the attribute feature of the secondary dynamic information as a candidate secondary attribute feature, and assigning a second weight to the candidate secondary attribute feature according to the interaction data; and regularly re-determining the secondary attribute features and the corresponding selection probability of the virtual digital human account according to the historical weights of the secondary attribute features and the candidate secondary attribute features.

[0040] Among them, interactive data can refer to the behavioral data generated by the friend user account in response to the secondary dynamic information released by the virtual digital human account, including likes, comments, forwarding, collections, etc. Interactive data reflects the friend user's attention and interest in dynamic information, and is an important basis for optimizing the secondary attribute characteristics of the virtual digital human account.

[0041] In implementation, the first interactive platform can regularly count the interactive data of the secondary dynamic information published by the virtual digital human account, and associate the statistical results with the secondary attribute characteristics as the basic data for the next step of weighting. For each secondary attribute characteristic, the total interactive data of all secondary dynamic information related to it can be counted. For example, the secondary attribute characteristic A is "pay attention to cost-effective products", and the interactive data of the secondary dynamic information 1 associated with it is 10 likes and 5 comments; the interactive data of the secondary dynamic information 2 associated with it is 20 likes and 3 forwardings; then the first weight of the secondary attribute characteristic A = 10·k1+ 5·k2 + 20·k1+ 3·k3, where k1, k2, and k3 are the unit coefficients corresponding to the like, comment, and forwarding behaviors, respectively. Of course, in different secondary dynamic information, the same interactive behavior can also correspond to different unit coefficients. The first weight can reflect the importance of the existing secondary attribute characteristics in the overall preference of friend users, and reflect the adaptability of the characteristics of the current virtual digital human account to the needs of friend users.

[0042] In addition, the attribute features extracted from the secondary dynamic information may not belong to the existing secondary attribute features, but may reflect the potential interests of the friend user. Therefore, other attribute features contained in the secondary dynamic information can be extracted as candidate secondary attribute features, and the candidate secondary attribute features can be assigned a second weight based on the interactive data. For example, the secondary dynamic information is "Smart lighting equipment is on sale!", and the attribute feature of the dynamic information can be extracted: "Smart lighting equipment", and if "Smart lighting equipment" does not belong to the existing secondary attribute features, it can be used as a candidate secondary attribute feature; combined with the interactive data of the secondary dynamic information, such as 15 likes and 10 comments, the second weight B=15·k1 + 10·k2 is assigned to the candidate secondary attribute feature "Smart lighting equipment". The second weight reflects the potential importance of the candidate secondary attribute feature among friend users, and is used to identify the potential interests or emerging needs of friend users.

[0043] Based on the above settings, the first interactive platform can periodically (such as weekly or monthly) analyze the historical weights of the secondary attribute features and candidate secondary attribute features, and can adjust the secondary attribute features of the virtual digital person according to the following rules: if the second weight of a candidate secondary attribute feature exceeds the lower limit of the first weight of the existing secondary attribute feature, it is added to the secondary attribute feature set; if the first weight of an existing secondary attribute feature is lower than a certain threshold for multiple consecutive periods, it is removed from the secondary attribute feature set. Of course, this embodiment does not limit the adjustment rules of the secondary attribute features.

[0044] Afterwards, the selection probability of each secondary attribute feature can be allocated according to the ratio of the first weight to the second weight. For example, the higher the weight of the secondary attribute feature, the greater the selection probability, thereby improving the accuracy of the virtual digital human account. In this way, by assigning weights to secondary attribute features through interactive data, the changing trend of the friend user's interest points can be reflected; regular analysis of the weights can timely discover new interest points of friend users, while eliminating secondary attribute features that are no longer of interest, so that the dynamic release of the virtual digital human account is more in line with the overall dynamic preferences of the friend user.

[0045] Of course, you can also extract positive or high-frequency keywords from your friends' interaction data as a reference for candidate features. You can also set weights for different groups of friend users. For example, for friend users with higher attention, give their interaction data a higher weight; for friend users with lower interaction frequency, you can attract them to participate in the interaction by appropriately increasing the weights of the features corresponding to their points of interest. When using weights to redefine secondary attribute features and selection probabilities, you can also use the trend of weight changes to predict the future needs of friend users. For example, by analyzing the gradual upward trend of the weight of a candidate secondary attribute feature, you can predict that the friend user's interest in the candidate secondary attribute feature will continue to grow, so that it can be included in the secondary attribute feature set with priority.

[0046] In one embodiment, the method further includes: when the secondary attribute feature is determined to be a high-frequency secondary attribute feature according to the historical weight, determining the target friend user account corresponding to the high-frequency secondary attribute feature; performing data processing on the effective dynamic information based on the high-frequency secondary attribute feature at a fixed frequency to obtain specific secondary dynamic information; publishing the specific secondary dynamic information through the virtual digital human account, and setting the specific secondary dynamic information to be visible to the target friend user account.

[0047] In implementation, the secondary attribute features corresponding to the secondary dynamic information with which friend users frequently interact can be determined based on the historical weight statistics, and defined as high-frequency secondary attribute features. High-frequency secondary attribute features can reflect the main interests or focus of the friend user group in a certain field. Specifically, they can be secondary attribute features that are related to multiple interactive data items and whose weights are continuously high or rising; for example, if the historical weight of a certain secondary attribute feature "smart wearable device" is in the top three for several consecutive periods, it can be determined to be a high-frequency secondary attribute feature. Furthermore, through statistical analysis, target friend user accounts that are highly matched with high-frequency secondary attribute features and frequently interact can be obtained. Specifically, they can be friends who have interacted with the dynamic information under the high-frequency secondary attribute features by multiple likes, comments, forwarding, and other interactive behaviors; for example, if friend user A frequently likes and comments on dynamics related to "smart wearable devices", user A can be determined as the target friend user account of the high-frequency secondary attribute feature of "smart wearable devices".

[0048] Based on the above settings, in addition to randomly selecting secondary attribute features to generate secondary dynamic information, data processing can be performed on effective dynamic information based on high-frequency secondary attribute features at a fixed frequency to obtain specific secondary dynamic information, and the specific secondary dynamic information can be published through a virtual digital human account, and set to be visible only to the target friend user account to avoid interference with irrelevant users; if the interactive behavior of the target friend user account changes, the visibility of the specific secondary dynamic information can be dynamically adjusted; for example, when a friend user's interest in "smart wearable devices" decreases, the friend user's visibility of the relevant specific secondary dynamic information can be cancelled.

[0049] It can be understood that the screening of high-frequency secondary attribute features and the determination of target friend user accounts can ensure accurate push and achieve precise positioning through historical weights and interaction data; the data processing link using high-frequency secondary attribute features can ensure that the content of specific secondary dynamic information is highly consistent with the needs of target users, thereby increasing the attractiveness of the content; and the use of visibility settings can reduce interference from irrelevant users and achieve differentiated content dissemination.

[0050] In one embodiment, the method also includes: when the friend interaction frequency of the virtual digital human account meets the preset standard, allocating account registration resources to the virtual digital human account; requesting the account information of other interactive platforms of the friend user account of the established first interactive platform through the virtual digital human account; registering the virtual digital human associated account on other interactive platforms based on the account registration resources, and establishing friend relationships on other interactive platforms through the account information; obtaining the friend dynamic information published by the friend user account on other interactive platforms through the virtual digital human associated account, and identifying the valid dynamic information matching the virtual attribute characteristics; performing data processing on the valid dynamic information to obtain secondary dynamic information, and publishing the secondary dynamic information on the first interactive platform through the virtual digital human account.

[0051] In implementation, when the frequency of interaction between the virtual digital human account and the friend user account meets the preset standard (for example, the frequency of likes, comments, and private messages reaches a certain threshold), the first interactive platform can determine that the virtual digital human account has the potential to further expand the social circle. Based on this determination, the first interactive platform can allocate account registration resources to the virtual digital human account. Account registration resources can be a fixed number of registration opportunities, an extended quota, or an identity information template generated by the virtual digital human system for the creation of cross-platform accounts. Based on the relationship established between the virtual digital human account and the friend user account on the first interactive platform, the virtual digital human account initiates a request through the first interactive platform to obtain the account information of the friend user account on other interactive platforms. This information may include user name, user ID, platform identifier, etc., all of which must be obtained under the premise of authorization by the friend user account to ensure privacy and security. Based on the allocated account registration resources, the first interactive platform can automatically or semi-automatically register the virtual digital human's associated account in other interactive platforms. After the registration is completed, the first interactive platform can use the obtained friend user account information to establish an associated relationship with the original friend user account in other interactive platforms to expand the social network scope of the virtual digital human.

[0052] Furthermore, the dynamic information published by the target friend user account can be obtained from other interactive platforms through the virtual digital human associated account. The first interactive platform can identify and screen these dynamic information, and select effective dynamic information that matches the virtual attribute characteristics of the virtual digital human account. For example: if the virtual attribute characteristics of the virtual digital human account tend to be "technology information", the technology-related dynamics published by the target friend can be screened. For the effective dynamic information identified, further data processing can be performed, such as reorganizing the content, optimizing the form of expression, adding multimedia content, etc., to enhance the attractiveness and dissemination effect of the dynamic information. The processed dynamic information is published to the first interactive platform through the virtual digital human account in the form of secondary dynamic information.

[0053] Based on the above scheme, through the creation of virtual digital human associated accounts, the friend relationship that was originally limited to the first interactive platform is extended to other interactive platforms. This cross-platform friend relationship expansion can effectively expand the influence of virtual digital human accounts and break the data island problem between different platforms. Using virtual digital human associated accounts to actively obtain and integrate dynamic information from different platforms, cross-platform data can be seamlessly connected and presented in a unified form, improving the integrity and expressiveness of dynamic content. By screening effective dynamic information and processing it, more targeted and attractive content can be provided to friend users of the first interactive platform, which not only avoids repeated low-quality information swiping the screen, but also further fits the interests and preferences of friend users. The above-mentioned integrated release mechanism of dynamic information not only enriches the content system of virtual digital human accounts, but also enhances user experience, further improving users' dependence on virtual digital people and interaction frequency.

[0054] It is worth mentioning that when implementing the above plan, you need to pay attention to the data privacy and content authorization of friend users. At the same time, when registering accounts on other platforms, you should follow the account management rules of other platforms to ensure compliance, or establish a cooperation mechanism with other platforms.

[0055] In one embodiment, after identifying the valid dynamic information matching the virtual attribute characteristics, it also includes: determining the platform attribute characteristics corresponding to the first interactive platform and other interactive platforms respectively; performing data processing on the valid dynamic information according to the platform attribute characteristics to obtain secondary dynamic information matching the target interactive platform; and publishing the secondary dynamic information on the target interactive platform through the virtual digital human account and / or the virtual digital human associated account.

[0056] In implementation, through the virtual digital human associated account, effective dynamic information that matches the virtual attribute characteristics of the virtual digital human account can be identified and matched from the dynamic information released by the target friend user account. Product release, evaluation and other content. For the first interactive platform and other interactive platforms, the server determines the platform attribute characteristics of each interactive platform respectively. For example, the video interactive platform can be biased towards short videos and content with strong audio and video performance; the e-commerce platform focuses on product recommendations, prices and promotional information; the knowledge sharing platform is mainly based on the knowledge explanation of pictures, texts or long content. After determining these platform attribute characteristics, a rule basis for dynamic information processing for different interactive platforms can be formed, so that the effective dynamic information can be processed according to the platform attribute characteristics to obtain secondary dynamic information adapted to the target interactive platform. The form of data processing can be, for example, converting text dynamics into short videos suitable for video platforms, or adding product information tags to e-commerce platforms, or supplementing background knowledge or detailed explanations related to dynamics in knowledge sharing platforms. After processing different secondary dynamic information, it can be published on the target interactive platform respectively through the virtual digital human account (first interactive platform) and / or the virtual digital human associated account (other interactive platforms). For example, the virtual digital human account displays relevant product links on the first interactive platform, and at the same time publishes video version dynamic content on the video platform through the associated account, and publishes short text updates on the knowledge sharing platform.

[0057] Based on the above content, by identifying the attribute characteristics of different interactive platforms, the processing form of dynamic information can be targeted to adapt to platform needs. For example: enhancing product information and shopping guide functions on e-commerce platforms, so that users can more conveniently obtain product-related content and directly complete purchases; presenting dynamic information in the form of short videos on video platforms is not only more in line with user habits, but also can significantly enhance the attractiveness of content. According to the content consumption habits of users on the target platform, the processing of dynamic information improves the dissemination effect of secondary dynamic information, avoiding user loss or decreased interaction rate due to format or content mismatch. By forming a unified and flexible content publishing strategy between different platforms, the interactive stickiness of virtual digital human accounts and their associated accounts can be enhanced.

[0058] In one embodiment, after the target interactive platform publishes the secondary dynamic information, it also includes: counting the cross-platform interactive data of the secondary dynamic information of the same valid dynamic information in each target interactive platform; extracting the valid information feature set in each cross-platform interactive data, and determining the set similarity between the valid information feature sets; updating the user feature similarity of the friend user account to which the corresponding cross-platform interactive data belongs according to the set similarity; and performing a friend recommendation process between user accounts on the first interactive platform based on the user feature similarity.

[0059] In practice, after publishing secondary dynamic information on the target interactive platform, the interactive data of the same effective dynamic information on each target interactive platform can be counted: for example, the user's likes, comments, forwarding and other behaviors are aggregated to form an interactive data set; according to the characteristics of different target platforms, weights are assigned according to the behavior type. For example, on a video platform, the weight of "comments" may be higher than that of "likes", while on an e-commerce platform, the weight of "click to buy" is higher.

[0060] Based on the statistical interaction data, extract the effective information feature set of each platform, including: user behavior preference features: such as interaction frequency, behavior type distribution; content preference features of dynamic information: such as keywords and topic tags involved (such as "electronic products" and "outdoor activities"); user account behavior history association features: such as whether the user continues to interact with a certain type of dynamic information. For the effective information feature sets extracted from different target platforms, calculate the set similarity between feature sets: measure the similarity of different users in content preferences and behavior patterns through the intersection size or weight distribution overlap of feature sets; for example, if user A and user B both frequently participate in dynamic interactions related to "new technology products" and have similar interaction methods (such as comment frequency), the set similarity is high. Based on the set similarity, update the user feature similarity of the friend user account on the first interactive platform: through cross-platform interaction data, infer the user account's interest points or behavior characteristics on the first interactive platform; the update of feature similarity may involve a dynamic weighting mechanism. For example, the latest interaction data has a higher weight and the historical data has a lower weight to reflect the dynamic changes in user interests. On the first interactive platform, a friend recommendation process is initiated based on the similarity of user characteristics: specifically, for two users with similar characteristics on other interactive platforms, if it is possible to clearly identify both of their accounts on the first interactive platform, a friend recommendation process between the two accounts can be performed on the first interactive platform; if it is possible to determine that only one of the users has an account on the first interactive platform, the account information of the other account on other interactive platforms can be recommended to the account, so that the account user can add friends on the other interactive platform.

[0061] It can be understood that since the user data of the target interactive platform is usually not directly available, user characteristics can be indirectly inferred through cross-platform interactive data. For example, users frequently click on a certain type of goods on the e-commerce platform and like related evaluation videos on the video platform, indicating that they have a high interest in this type of goods. This feature inference method based on behavioral data can make up for the problem of direct information loss caused by platform data barriers. By calculating the set similarity and user feature similarity, the interest characteristics of cross-platform users can be fed back to the first interactive platform to achieve friend recommendations. Users with high similarity are more likely to have common topics or interests, which increases the success rate of establishing friend relationships; compared with the random recommendation method, the recommendation results of the above scheme are more accurate and the user experience is better. Furthermore, by using cross-platform data, a comprehensive analysis of user characteristics of different interactive platforms can be achieved, which effectively expands the user's potential friend range, enhances the efficiency of building the user relationship network of the first interactive platform, and also enhances its users' dependence and recognition on the first interactive platform.

[0062] In one embodiment, the method is applied to a distributed system, which includes a central node and at least one first-level edge node. The service areas of multiple edge nodes at the same level are different from each other. The upper-level edge node corresponds to at least one lower-level edge node, and the service area of ​​the upper-level edge node covers the service area of ​​its lower-level edge node; wherein the central node is configured to create a virtual digital human account for each mainstream user category, and send the virtual digital human account to the corresponding secondary edge node respectively; the secondary edge node is configured to filter the optional user accounts whose address information belongs to its service area, and establish personal friend relationships with the optional user accounts under the mainstream user category based on the virtual digital human account; the secondary edge node is also configured to extract regional characteristic information of its service area, and the regional characteristic information is used to generate secondary dynamic information when processing effective dynamic information.

[0063] The distributed system includes a central node and at least one level of edge nodes, with a hierarchical structure. The central node serves as the global control center and is responsible for the creation and configuration of virtual digital human accounts. The edge nodes are divided into one or more levels. Each edge node serves a specific area and is responsible for the interaction with user accounts in the area it covers and the processing and publishing of dynamic information.

[0064] The central node can define mainstream user categories, such as "young consumers", "mother and baby users", and "sports experts" in combination with user behavior data analysis, and configure the attributes of the virtual digital human account according to the typical characteristics of each mainstream user category (such as topics of interest and content preferences), including personal information, interest tags, interaction style, etc. After the creation is completed, the central node can send the virtual digital human account to the secondary edge node to enable it to carry out personalized operations in various regions. Each secondary edge node screens user accounts belonging to the area based on the address information of the service area. Specifically, the secondary edge node can identify users belonging to the current edge node service area based on the registered address, geographic location or interaction record of the user account; and further screen users who meet the mainstream categories of the virtual digital human account from the optional user accounts as potential friends of the virtual digital human account. The screening process can combine multiple matching strategies, such as content interest similarity, interaction frequency, etc. Based on the results of the above screening, the virtual digital human account establishes a personal friend relationship with the optional user account under the same mainstream category. After the friend relationship is established, the virtual digital human account can interact with the user account through dynamic sharing, thereby enhancing user stickiness and activity.

[0065] The secondary edge node can further extract regional characteristic information from its service area, which may include user interest characteristics, such as analyzing the preferences of users in the service area for specific topics, such as "electronic product discounts" and "festival activities" based on the behavioral data of users in the service area; it can also include language and cultural characteristics, such as the main language used in the service area and popular expressions; it can also include hot events and needs, extracting real-time hot information in the area, such as large-scale events, promotional seasons, or unique consumption habits in the region. In this way, the secondary edge node can use regional characteristic information to process effective dynamic information and generate secondary dynamic information. During the processing process, the dynamic information content can be customized according to the characteristics of the service area. For example, in areas where shopping is preferred, the dynamic information will strengthen the product information (such as price, discount, purchase link); in areas where videos are preferred, the dynamic information can be converted into short video format; in knowledge sharing areas, the dynamic information focuses more on the explanation of core concepts and the improvement of content depth. The processed secondary dynamic information can be published in their respective service areas through virtual digital human accounts. Each edge node independently publishes dynamic information to avoid content duplication in different areas, so as to more accurately meet user needs and improve the dissemination efficiency of dynamic information.

[0066] Based on the above scheme, the creation of virtual digital human accounts is concentrated in the central node, which can ensure the global consistency of virtual digital human accounts; data screening and dynamic processing are dispersed to edge nodes, which reduces the computing and management burden of the central node. The distributed architecture design enhances the scalability of the system, enabling it to flexibly respond to large-scale user distribution and dynamic content needs. At the same time, the regional processing method of edge nodes improves the efficiency of publishing dynamic information and reduces the delay of content distribution. As for the extraction and application of regional characteristic information, it can make the processing of dynamic information more in line with the actual needs of users. On the one hand, users can receive dynamic information that is more in line with their interests and language and cultural characteristics, thereby enhancing the user experience; the regional characteristics of dynamic information can also help virtual digital human accounts more effectively participate in the interaction of local users and enhance the credibility and influence of accounts.

[0067] In one embodiment, the secondary edge node is also configured to select high-quality secondary dynamic information based on the interactive data of the secondary dynamic information it maintains, and send the high-quality secondary dynamic information to its corresponding upper edge node; the upper edge node is configured to process the high-quality secondary dynamic information, generate tertiary dynamic information, and send the tertiary dynamic information to the remaining corresponding secondary edge nodes, so that the remaining secondary edge nodes publish the tertiary dynamic information through the virtual data person account they maintain.

[0068] In implementation, each secondary edge node can screen out high-quality dynamic information based on the interactive data of secondary dynamic information in its service area. First, user interactive data of secondary dynamic information posted on virtual digital human accounts can be collected, including the number of likes, the number of comments, the forwarding rate, etc. Then, by setting evaluation indicators (such as the threshold of interactive data or the weighted scoring model), dynamic information with high user attention and interaction rate can be screened out. Finally, the screened high-quality dynamic information is marked with a high-quality mark for subsequent reception and processing by the upper edge node. After that, the secondary edge node uploads the high-quality dynamic information to the corresponding upper edge node, which further processes the high-quality dynamic information. Specifically, the upper edge node can combine the high-quality dynamic information of multiple secondary edge nodes, remove redundant content, extract core hot information, and form a higher-level content summary; and can regionalize the high-quality dynamic information according to the overall characteristics of the area covered by the upper edge node (such as language habits and cultural commonalities), such as in different language areas, dynamic information can be translated into different languages; for multi-cultural areas, generally applicable content can be extracted from dynamic information to avoid incompatibility caused by cultural differences. After processing the three-dimensional dynamic information, the upper edge node can send the three-dimensional dynamic information to other secondary edge nodes under its jurisdiction, and each secondary edge node can share the three-dimensional dynamic information with users in the region through the virtual digital human account maintained by it.

[0069] Based on the above scheme, on the one hand, the secondary edge nodes can filter the secondary dynamic information, and upload the high-quality content in the local area to the upper node step by step, so as to realize the hierarchical filtering and dissemination of dynamic information. In the local area, the high-quality dynamic information can reflect the main needs of users in the area and ensure the accuracy of the uploaded content; in the global scope, the dynamic information can be disseminated across regions through the processing of the upper node, so as to realize the efficient sharing of high-quality content. On the other hand, the processing link of the upper edge node can balance the regional characteristics of each secondary node, so that the tertiary dynamic information has cross-regional adaptability, avoids the understanding deviation or cultural conflict caused by regional differences in content dissemination, improves the universality of dynamic information, and enables it to be accepted and disseminated by users in a wider area.

[0070] By adopting the virtual digital human interactive service method disclosed in this application, by obtaining the user data set of the first interactive platform, summarizing the high-frequency attribute features in the user data set and determining multiple mainstream user categories, it is possible to accurately characterize the behavioral characteristics of the target user group and provide data support for subsequent interactions. A virtual digital human account is created for each mainstream user category, and a personal friend relationship is established with the user account to simulate the interaction scenario in a real social network. This method can effectively cover individual users in the mainstream user category, significantly improve the relevance and closeness between the virtual digital human and the user, and enhance the participation and credibility of the virtual digital human in the social network.

[0071] By setting the virtual attribute characteristics of the virtual digital human account based on the attribute characteristics of the friend user account, the virtual digital human can present an image and characteristics that match the target user group, thereby improving the matching and attractiveness of the interactive content. At the same time, the virtual digital human account can obtain the dynamic information of friends posted by its friend users, and identify the valid dynamic information that matches the virtual attribute characteristics. This precise screening mechanism can avoid the interference of redundant information or invalid content, and improve the efficiency and quality of dynamic information processing.

[0072] In addition, by processing effective dynamic information to generate secondary dynamic information, and publishing the secondary dynamic information through virtual digital human accounts, the content can be re-disseminated and diffused among user groups. This method not only fully taps the potential value of user dynamic information, but also improves the activity and influence of virtual digital human accounts, allowing them to form a continuous interactive chain in social networks, enhancing users' sense of participation and interactive experience.

[0073] This method realizes the automation and intelligence of virtual digital human interaction by automatically creating virtual digital human accounts based on the attribute characteristics of user data, screening effective dynamic information, and processing and publishing secondary dynamic information. This method reduces the cost and complexity of manual operation while ensuring the accuracy and pertinence of the content. In practical applications, this method can adapt to large-scale user scenarios, improve the platform's responsiveness to user needs, and provide technical support for content distribution and user activity improvement on social networks, forming an efficient, intelligent, and continuously optimized virtual digital human interaction service process.

[0074] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0075] Based on the same inventive concept, Figure 2 As shown, the embodiment of the present application further provides a virtual digital human interactive service device 200, and the device 200 includes: A feature aggregation and classification module 201 is used to obtain a user data set of the first interactive platform, aggregate high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories; A digital human account creation module 202, used to create a virtual digital human account for each mainstream user category, and to establish a personal friend relationship with the user accounts under the mainstream user category based on the virtual digital human account; A virtual attribute setting module 203, used to set the virtual attribute characteristics of the virtual digital human account based on the attribute characteristics of the friend user accounts that have been established by the virtual digital human account; Dynamic information identification module 204, used to obtain the dynamic information of friends published by the friend user account through the virtual digital human account, and identify the valid dynamic information matching the virtual attribute characteristics; The secondary dynamic publishing module 205 is used to process the effective dynamic information to obtain secondary dynamic information, and publish the secondary dynamic information through the virtual digital human account.

[0076] In one of the embodiments, the virtual attribute feature includes a primary attribute feature and a plurality of secondary attribute features; The dynamic information identification module 204 is specifically used for: Identify valid dynamic information matching the primary attribute feature; The secondary dynamic publishing module 205 is specifically used for: One of the secondary attribute features is randomly selected, and data processing is performed on the effective dynamic information based on the selected secondary attribute feature to obtain secondary dynamic information.

[0077] In one embodiment, the virtual attribute setting module 203 is further used to: Counting the interaction data of all friend user accounts with respect to the secondary dynamic information; Assigning a first weight to the corresponding secondary attribute feature according to the interaction data; Extracting attribute features of the secondary dynamic information as candidate secondary attribute features, and assigning a second weight to the candidate secondary attribute features according to the interaction data; The secondary attribute characteristics and corresponding selection probabilities of the virtual digital human account are periodically re-determined based on the historical weights of the secondary attribute characteristics and the candidate secondary attribute characteristics.

[0078] In one embodiment, the secondary dynamic publishing module 205 is further used to: When the secondary attribute feature is determined to be a high-frequency secondary attribute feature according to the historical weight, determining a target friend user account corresponding to the high-frequency secondary attribute feature; Processing the effective dynamic information based on the high-frequency secondary attribute characteristics at a fixed frequency to obtain specific secondary dynamic information; The specific secondary dynamic information is published through the virtual digital human account, and the specific secondary dynamic information is set to be visible to the target friend user account.

[0079] In one embodiment, the secondary dynamic publishing module 205 is further used to: When the friend interaction frequency of the virtual digital human account meets the preset standard, account registration resources are allocated to the virtual digital human account; Requesting, through the virtual digital human account, account information of other interactive platforms of the friend user account of the first interactive platform that has been established; Registering a virtual digital human associated account on the other interactive platform based on the account registration resource, and establishing a friend relationship on the other interactive platform through the account information; By associating the virtual digital human account, obtaining the dynamic information of friends published by the friend user account on the other interactive platform, and identifying the valid dynamic information matching the virtual attribute characteristics; The effective dynamic information is processed to obtain secondary dynamic information, and the secondary dynamic information is published on the first interactive platform through the virtual digital human account.

[0080] In one embodiment, the secondary dynamic publishing module 205 is further used to: Determining platform attribute characteristics corresponding to the first interactive platform and the other interactive platforms respectively; Processing the effective dynamic information according to the platform attribute characteristics to obtain secondary dynamic information matching the target interactive platform; The secondary dynamic information is published on the target interactive platform through the virtual digital human account and / or the virtual digital human associated account.

[0081] In one embodiment, the secondary dynamic publishing module 205 is further used to: Count the cross-platform interaction data of the secondary dynamic information of the same valid dynamic information in each of the target interactive platforms; Extracting valid information feature sets from each of the cross-platform interaction data, and determining set similarities between the valid information feature sets; According to the set similarity, updating the user feature similarity of the corresponding friend user account to which the cross-platform interaction data belongs; Based on the user feature similarity, a friend recommendation process is performed between user accounts on the first interactive platform.

[0082] Based on the same inventive concept, an embodiment of the present application also provides a distributed system, which includes a central node and at least one level of edge nodes. The service areas of multiple edge nodes at the same level are different from each other. The upper-level edge node corresponds to at least one lower-level edge node, and the service area of ​​the upper-level edge node covers the service area of ​​its lower-level edge node; the central node and / or the edge node are configured to perform the processing of the above steps 101 to 105.

[0083] In one embodiment, the central node is configured to create a virtual digital human account for each mainstream user category, and send the virtual digital human account to the corresponding secondary edge node respectively; The secondary edge node is configured to screen optional user accounts whose address information belongs to its service area, and establish personal friend relationships with optional user accounts under the mainstream user category based on the virtual digital human account; The secondary edge node is further configured to extract regional characteristic information of its service area, and the regional characteristic information is used to generate the secondary dynamic information when performing data processing on the effective dynamic information.

[0084] In one embodiment, the secondary edge node is further configured to select high-quality secondary dynamic information based on the interactive data of the secondary dynamic information maintained by it, and send the high-quality secondary dynamic information to its corresponding upper edge node; The upper edge node is configured to process the high-quality secondary dynamic information, generate tertiary dynamic information, and send the tertiary dynamic information to the corresponding remaining secondary edge nodes, so that the remaining secondary edge nodes publish the tertiary dynamic information through the maintained virtual data person account.

[0085] In one embodiment, a computer device is provided, whose internal structure diagram can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange data between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a virtual digital human interactive service method is implemented.

[0086] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0087] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0088] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0089] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0090] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The processor involved in each embodiment provided in this application can be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc., but is not limited thereto.

[0091] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A virtual digital human interactive service method, characterized in that: The method comprises: Acquire a user data set of the first interactive platform, summarize high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories; Creating a virtual digital human account for each mainstream user category, and establishing a personal friendship relationship with the user accounts under the mainstream user category based on the virtual digital human account; Based on the attribute characteristics of the friend user accounts that have been established by the virtual digital human account, set the virtual attribute characteristics of the virtual digital human account; Obtaining the dynamic information of friends published by the friend user account through the virtual digital human account, and identifying the valid dynamic information matching the virtual attribute characteristics; The effective dynamic information is processed to obtain secondary dynamic information, and the secondary dynamic information is published through the virtual digital human account.

2. The method according to claim 1, characterized in that The virtual attribute features include a primary attribute feature and a plurality of secondary attribute features; The identifying of valid dynamic information matching the virtual attribute feature includes: Identify valid dynamic information matching the primary attribute feature; The effective dynamic information is processed to obtain secondary dynamic information, including: One of the secondary attribute features is randomly selected, and data processing is performed on the effective dynamic information based on the selected secondary attribute feature to obtain secondary dynamic information.

3. The method according to claim 2, characterized in that The method further comprises: Counting the interaction data of all friend user accounts with respect to the secondary dynamic information; Assigning a first weight to the corresponding secondary attribute feature according to the interaction data; Extracting attribute features of the secondary dynamic information as candidate secondary attribute features, and assigning a second weight to the candidate secondary attribute features according to the interaction data; The secondary attribute characteristics and corresponding selection probabilities of the virtual digital human account are periodically re-determined based on the historical weights of the secondary attribute characteristics and the candidate secondary attribute characteristics.

4. The method according to claim 3, characterized in that The method further comprises: When the secondary attribute feature is determined to be a high-frequency secondary attribute feature according to the historical weight, determining a target friend user account corresponding to the high-frequency secondary attribute feature; Processing the effective dynamic information based on the high-frequency secondary attribute characteristics at a fixed frequency to obtain specific secondary dynamic information; The specific secondary dynamic information is published through the virtual digital human account, and the specific secondary dynamic information is set to be visible to the target friend user account.

5. The method according to claim 1, characterized in that The method further comprises: When the friend interaction frequency of the virtual digital human account meets the preset standard, account registration resources are allocated to the virtual digital human account; Requesting, through the virtual digital human account, account information of other interactive platforms of the friend user account of the first interactive platform that has been established; Registering a virtual digital human associated account on the other interactive platform based on the account registration resource, and establishing a friend relationship on the other interactive platform through the account information; By associating the virtual digital human account, obtaining the dynamic information of friends published by the friend user account on the other interactive platform, and identifying the valid dynamic information matching the virtual attribute characteristics; The effective dynamic information is processed to obtain secondary dynamic information, and the secondary dynamic information is published on the first interactive platform through the virtual digital human account.

6. The method according to claim 5, characterized in that After the valid dynamic information matching the virtual attribute feature is identified, the method further includes: Determining platform attribute characteristics corresponding to the first interactive platform and the other interactive platforms respectively; Processing the effective dynamic information according to the platform attribute characteristics to obtain secondary dynamic information matching the target interactive platform; The secondary dynamic information is published on the target interactive platform through the virtual digital human account and / or the virtual digital human associated account.

7. The method according to claim 6, characterized in that After the target interactive platform publishes the secondary dynamic information, the method further includes: Count the cross-platform interaction data of the secondary dynamic information of the same valid dynamic information in each of the target interactive platforms; Extracting valid information feature sets from each of the cross-platform interaction data, and determining set similarities between the valid information feature sets; According to the set similarity, updating the user feature similarity of the corresponding friend user account to which the cross-platform interaction data belongs; Based on the user feature similarity, a friend recommendation process is performed between user accounts on the first interactive platform.

8. The method according to claim 1, characterized in that The method is applied to a distributed system, which includes a central node and at least one level of edge nodes, wherein the service areas of multiple edge nodes at the same level are different from each other, an upper-level edge node corresponds to at least one lower-level edge node, and the service area of ​​the upper-level edge node covers the service area of ​​its lower-level edge node; The central node is configured to create a virtual digital human account for each mainstream user category, and send the virtual digital human account to the corresponding secondary edge node respectively; The secondary edge node is configured to screen optional user accounts whose address information belongs to its service area, and establish personal friend relationships with optional user accounts under the mainstream user category based on the virtual digital human account; The secondary edge node is further configured to extract regional characteristic information of its service area, and the regional characteristic information is used to generate the secondary dynamic information when performing data processing on the effective dynamic information.

9. The method according to claim 8, characterized in that The secondary edge node is further configured to select high-quality secondary dynamic information according to the interactive data of the secondary dynamic information maintained by it, and send the high-quality secondary dynamic information to its corresponding upper edge node; The upper edge node is configured to process the high-quality secondary dynamic information, generate tertiary dynamic information, and send the tertiary dynamic information to the corresponding remaining secondary edge nodes, so that the remaining secondary edge nodes publish the tertiary dynamic information through the maintained virtual data person account.

10. A virtual digital human interactive service device, characterized in that: The device comprises: A feature aggregation and classification module, used to obtain a user data set of the first interactive platform, aggregate high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories; A digital human account creation module, used to create a virtual digital human account for each mainstream user category, and to establish a personal friendship relationship with the user accounts under the mainstream user category based on the virtual digital human account; A virtual attribute setting module, used to set the virtual attribute characteristics of the virtual digital human account based on the attribute characteristics of the friend user accounts that have been established by the virtual digital human account; A dynamic information identification module, used to obtain the dynamic information of friends published by the friend user account through the virtual digital human account, and identify the valid dynamic information matching the virtual attribute characteristics; The secondary dynamic publishing module is used to process the effective dynamic information to obtain secondary dynamic information, and publish the secondary dynamic information through the virtual digital human account.

Images