A virtual digital human interaction service method and device

Abstract

The application 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; a virtual digital human account is created for each mainstream user category, a personal friend relationship is established between the virtual digital human account and a user account under the mainstream user category based on the virtual digital human account; virtual attribute features of the virtual digital human account are set based on attribute features of friend user accounts established by the virtual digital human account; friend dynamic information published by the friend user accounts of the virtual digital human account is acquired through the virtual digital human account, and effective dynamic information matching the virtual attribute features is identified; secondary dynamic information is obtained by performing data processing on the effective dynamic information, and the secondary dynamic information is published through the virtual digital human account. The method can improve the information sharing efficiency and user experience of the interaction platform.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a virtual digital human interaction service method and device. BACKGROUND

[0002] In the existing commodity shopping guide platform, the common user flow attracting method usually includes content pushing based on recommendation algorithm and interactive marketing of social community. Specifically, the platform will collect the browsing records, shopping habits and other data of the users, generate commodity information meeting the user's preferences by using the recommendation algorithm, 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, form the interaction between users through comments, likes and other social behaviors to improve the platform activity. Some platforms will introduce brand promotion accounts to directly publish promotion content related to commodities to attract users' attention. However, these methods usually rely on the active participation of users and cannot fully tap the potential value of the user's published content. The interaction form between users and the platform is also relatively single.

[0003] The existing technical solutions still have many problems in actual application. First, the accuracy of the recommendation algorithm is limited by the amount of data and the model capability, which leads to the mismatch between the recommended content and the real needs of the users, thereby reducing the flow attracting effect. Second, the interactive behavior of the social community depends on the activity of the users, and the participation of different users is quite different, making it difficult to continuously form an efficient content output and interaction link. In addition, the dynamic information published by the users on the platform usually only stays at the level of one-way propagation and is not further processed and utilized, resulting in the platform failing to fully tap the value of these contents. Finally, the commercial attribute of the brand promotion account is too obvious, which easily causes the users to feel disgusted and thus reduces the user experience. Therefore, there is an urgent need for a new type of interactive service method to solve the above problems of the existing technology through innovative interactive forms and content processing procedures, and to improve the flow attracting ability and user stickiness of the platform. SUMMARY

[0004] Therefore, it is necessary to provide a virtual digital human interaction service method and device, computer equipment, storage medium and computer program product in view of the above technical problems.

[0005] In a first aspect, the present application provides a virtual digital human interaction service method, which comprises:

[0006] obtaining 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;

[0007] 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;

[0008] set virtual attribute features of the virtual digital human account based on attribute features of friend user accounts that have established a personal friend relationship with the virtual digital human account;

[0009] obtain friend dynamic information published by friend user accounts of the virtual digital human account, and identify valid dynamic information matching the virtual attribute features;

[0010] process the valid dynamic information to obtain secondary dynamic information, and publish the secondary dynamic information through the virtual digital human account.

[0011] In a second aspect, the present application further provides a virtual digital human interaction service device, which comprises:

[0012] a feature summary classification module configured to obtain a user data set of a first interaction platform, summarize high-frequency attribute features in the user data set, and determine a plurality of mainstream user categories;

[0013] a digital human account creation module configured to create a virtual digital human account for each mainstream user category, and establish a personal friend relationship between the virtual digital human account and a user account in the mainstream user category;

[0014] a virtual attribute setting module configured to set virtual attribute features of the virtual digital human account based on attribute features of friend user accounts that have established a personal friend relationship with the virtual digital human account;

[0015] a dynamic information identification module configured to obtain friend dynamic information published by friend user accounts of the virtual digital human account, and identify valid dynamic information matching the virtual attribute features;

[0016] a secondary dynamic information publishing module configured to process the valid dynamic information to obtain secondary dynamic information, and publish the secondary dynamic information through the virtual digital human account.

[0017] In a third aspect, the present application further provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method of the first aspect when executing the computer program.

[0018] In a fourth aspect, the present application further provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program implements the steps of the method of the first aspect when executed by a processor.

[0019] In a fifth aspect, the present application further provides a computer program product. The computer program product comprises a computer program, and the computer program implements the steps of the method of the first aspect when executed by a processor.

[0020] By adopting the virtual digital human interaction service method disclosed in the present application, the user data set of the first interaction platform is obtained, the high-frequency attribute features in the user data set are summarized, and a plurality of mainstream user categories are determined, so that the behavior characteristics of the target user group can be accurately described, and data support can be provided for subsequent interaction. 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 scene in the real social network. This way can effectively cover individual users in the mainstream user category, significantly improve the association 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.

[0021] By setting the virtual attribute features of the virtual digital human account based on the attribute features 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 degree and attractiveness of the interaction content. At the same time, the virtual digital human account can obtain the friend dynamic information published by its friend users, and identify effective dynamic information matching the virtual attribute features. This precise screening mechanism can avoid the interference of redundant or invalid content, and improve the efficiency and quality of dynamic information processing.

[0022] In addition, by data processing the effective dynamic information to generate secondary dynamic information, and publishing the secondary dynamic information by the virtual digital human account, the content can be re-transmitted and diffused in the user group. This method not only fully taps the potential value of user dynamic information, but also improves the activity and influence of the virtual digital human account, so that it forms a continuous interaction chain in the social network, enhancing the user's sense of participation and interactive experience.

[0023] The method automatically creates a virtual digital human account based on the attribute features of the user data, filters effective dynamic information, and processes and publishes secondary dynamic information, thereby realizing the automation and intelligentization of virtual digital human interaction. This way reduces the cost and complexity of manual operation, while ensuring the accuracy and relevance of the content. In practical applications, the method can adapt to large-scale user scenarios, improve the platform's response to user needs, and provide technical support for content distribution and user activity improvement in social networks, forming an efficient, intelligent, and continuously optimized virtual digital human interaction service process. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A virtual digital human interaction service method flowchart in an embodiment;

[0025] Figure 2 A virtual digital human interaction service device architecture diagram in an embodiment;

[0026] Figure 3 Fig. 1 is a schematic diagram of an internal structure of a computer device according to an embodiment. DETAILED DESCRIPTION

[0027] For the purposes of the present application, the technical solutions and advantages thereof will be more clearly apparent from the following detailed description of embodiments, given by way of example only, to be considered together with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended to be illustrative only and are not intended to limit the scope of the present application.

[0028] The embodiment of the present application provides a virtual digital human interaction service method, which can be applied to the server of an interaction platform and is suitable for various types of interaction platforms, such as online shopping platforms, live streaming platforms, friend-making platforms, video playing platforms and knowledge sharing platforms, etc. In these platforms, users usually publish dynamics, interact with comments and browse content through personal accounts, while platform parties improve user experience, enhance user stickiness and realize commercial diversion through various ways.

[0029] The server of the interaction platform can obtain a user data set and analyze user behavior characteristics to summarize and form multiple mainstream user categories. For example, in an online shopping platform, the mainstream user categories can include high-frequency purchasing users, potential users or users interested in specific product types; in a live streaming platform, the mainstream user categories can include viewers of specific field content or users with active interaction; in a friend-making platform, the mainstream user categories can be divided according to the interests or social preferences of users. Based on these mainstream user categories, the server creates a virtual digital human account and establishes a friend relationship with actual users as a virtual user to provide more personalized interaction services for users.

[0030] The server of the interaction platform can serve as the core executive body of the entire system to complete the creation of a virtual digital human account, the establishment of a friend relationship, the identification and processing of friend dynamic information, and the publication of virtual digital human dynamic information. Through the server, the platform can further process the dynamic information published by actual users to meet more diversified content display needs. For example, in an online shopping platform, a virtual digital human can publish secondary dynamic information combined with user shopping records to recommend goods to other users; in a knowledge sharing platform, a virtual digital human can convert professional knowledge content published by users into more popular dynamic information to attract more users' attention and improve the content dissemination effect of the platform.

[0031] Hereinafter, taking an online shopping platform or a live streaming shopping guide platform as an example, a virtual digital human interaction service method as shown in Fig. 1 will be explained, which includes the following steps: Figure 1

[0032] ​In step 101, a user data set of the first interaction platform is obtained, high-frequency attribute features in the user data set are summarized, and a plurality of mainstream user categories are determined.

[0033] The first interaction platform can be a main service platform implementing the scheme, and can be an online shopping platform or a live shopping guide platform. In related scenarios, a user can register an account on the first interaction platform and register relevant user data. In addition, the first interaction platform can request to obtain content interaction records and behavior data of the user on the first interaction platform. Thus, after obtaining authorization from the user, the first interaction platform can collect and summarize the content interaction records and behavior data of a user account, extract attribute features therefrom, and analyze them, thereby forming a user data set of the first interaction platform. The user data set can include user data units at the user granularity, and each user data unit can correspond to a user account and specifically include a plurality of attribute features that can reflect user data. User data can come from multiple dimensions, including but not limited to: user basic information such as age, gender, location, etc.; behavior data such as the types of browsed goods, search keywords, shopping cart addition records, payment records, return records, etc.; and content interaction records, which can include the frequency and content preferences of participating in comments, likes, and tips in live shopping guides, etc.

[0034] Further, the first interaction platform can extract and summarize high-frequency attribute features from the obtained user data set, such as “electronics enthusiasts”, “makeup experts”, “home improvement experts”, “foodies”, “second-dimensional players”, and “sports enthusiasts”. Here, high-frequency attribute features can be determined based on the number of users with the attribute feature, the total consumption, or the level of activity. Then, the first interaction platform can determine a plurality of mainstream user categories based on the high-frequency attribute features, thereby dividing users into different categories through clustering algorithms. Here, a mainstream user category can be a category with a certain number of users, and one 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, mainstream user categories can be “young women who love makeup”, “late-night active foodies”, “live shopping regulars”, and “frugal users”. Through the above process, the first interaction platform can provide data support for the subsequent creation of virtual digital person accounts, making the attributes and behaviors of virtual digital persons more in line with the needs of target users to improve their lead generation effects and user acceptance.

[0035] In step 102, a virtual digital person account is created for each mainstream user category, and a personal friend relationship is established between the virtual digital person account and the user account in the mainstream user category.

[0036] In implementation, the first interaction platform can create a virtual digital person account for each mainstream user category, the role and attributes of each virtual digital person account can be associated with the mainstream user category, and the specific implementation can include the following aspects: image design, the appearance and style of the virtual digital person can be customized according to the preferences of the target user category, for example, young female users may prefer virtual digital persons with sweet appearance and love to share shopping experience; while male users interested in electronic products may prefer professional virtual digital persons; behavior rules, which can include language style (such as humor, professionalism), interaction frequency (such as regular likes, comments) and sharing content methods, etc.

[0037] After that, the first interaction platform can establish a personal friendship relationship with the user account under the mainstream user category based on the virtual digital person account. Specifically, the virtual digital person can be recommended to the user account through the platform recommendation algorithm, for example, the platform can recommend a virtual digital person who is good at sharing coupons to the "price sensitive user"; or the virtual digital person can actively initiate a friendship request to the user account, accompanied by personalized greetings (such as "Hello, I am a small helper who specializes in electronic product recommendations, and will bring you more discounts and new product recommendations in the future"). Of course, the virtual digital person can be an auxiliary role of the host, a product explainer, a content sharer, etc., attracting the active attention of the user account through speaking, interacting, etc. Through this process, the virtual digital person account can effectively integrate into the user's social circle and gradually accumulate an initial friend base, laying a foundation for subsequent lead generation and content dissemination.

[0038] Step 103, based on the attribute characteristics of the friend user account established by the virtual digital person account, setting the virtual attribute characteristics of the virtual digital person account.

[0039] In implementation, the virtual attribute feature of the virtual digital human account is the core of its lead generation effect, and its setting can fully consider the attribute features of the friend user accounts. It can be understood that the setting of the mainstream user category is more for clustering of user accounts, so that the potential friends of a virtual digital human account all have certain same or similar attribute features. Due to the selection of different user accounts, the friend adding result of the virtual digital human account also has great uncertainty, so the virtual attribute feature of the virtual digital human account can be set based on the attribute features of the friend user accounts established by the virtual digital human account after the friend adding. Specifically, the platform can analyze the interest preferences and behavior characteristics of the virtual digital human friends, extract the attribute features of the friend user accounts, and then set the virtual attribute feature of the virtual digital human account according to the attribute features of all friend user accounts. For e-commerce platforms or live shopping platforms, the attribute features can be the main product categories and price ranges that the user accounts focus on, etc. For example, in an e-commerce platform, if the friends of the virtual digital human frequently purchase low-price skincare products, their attribute can be set as "affordable skincare expert", and if the friends of the virtual digital human often watch live streaming of digital products, their attribute can be set as "digital product expert". Of course, the operation of adding friends is essentially a continuous process after the generation of the virtual digital human account, and after the friends are added, there will also be friend loss, so the virtual attribute feature can be set to be updated constantly at a certain frequency; at the same time, the virtual attribute feature of the virtual digital human account can also be adjusted in real time according to the changes in the behavior of the friend users, for example, when a large number of friend users start to focus on different product categories or participate in new promotional activities, i.e. when the attribute features of the friend user accounts change significantly, the virtual attribute feature of the virtual digital human account can also be updated. In this way, by setting the virtual attribute feature that fits the needs of the friend users, the virtual digital human can form a more distinctive role positioning in the eyes of the friend users, thereby improving the trust and acceptance of the users.

[0040] Step 104, obtaining the friend dynamic information published by the friend user accounts of the virtual digital human account through the virtual digital human account, and identifying the effective dynamic information matching the virtual attribute feature.

[0041] In implementation, the friend user account can publish relevant dynamic information on the first interaction platform, and the virtual digital person account can obtain the relevant dynamic information due to the friendship relationship, and can further analyze and identify the relevant dynamic information to determine whether the dynamic information is effective dynamic information matching the virtual attribute feature. Specifically, the first interaction platform can obtain the dynamic information published by the friend user through the virtual digital person account, such as the shopping sharing, evaluation content published by the user in the online shopping platform, or the interactive comments participated in in the live shopping guide platform, etc. The first interaction platform can perform semantic analysis on the dynamic information through natural language processing technology, and match with the virtual attribute feature of the virtual digital person. For example: if the attribute of the virtual digital person is “affordable skincare expert”, it can identify whether there is content related to skincare in the friend dynamic information; if the attribute of the virtual digital person is “digital product expert”, it can identify whether the friend dynamic information involves content related to digital products. Of course, the dynamic information can also be set with weight score, etc., such as the friend dynamic involving high-frequency interactive behavior (such as more comments), or involving popular goods or topics, or the dynamic information content is more high-quality, and the score is higher, and when selecting effective dynamic information, the friend dynamic information with higher score is preferentially selected. In this way, through effective identification of the dynamic information, the virtual digital person can avoid responding to irrelevant content, and further improve the interactive efficiency.

[0042] In step 105, the effective dynamic information is processed to obtain secondary dynamic information, and the secondary dynamic information is published through the virtual digital person account.

[0043] In implementation, the first interaction platform can process the effective dynamic information, and publish the processed secondary dynamic information through the virtual digital person account in the form of account dynamic. Then the first interaction platform can publish the generated secondary dynamic information to the friend dynamic circle through the social function of the virtual digital person account, to attract other friend users to check or further interact. In this way, by publishing the secondary dynamic information, the virtual digital person can not only effectively expand the content dissemination range, but also guide the friend users to participate in the interaction, and further improve the platform's flow effect and business conversion rate.

[0044] Among them, the data processing on the effective dynamic information can include data extraction, reorganization, optimization, desensitization, and re-creation, etc. Specifically: on the one hand, the dynamic information content can be adjusted based on the virtual attribute characteristics from the content dimension, so as to improve the attractiveness and relevance of the secondary dynamics, such as the friend dynamic information is “morning run helps a day of vitality”, and the virtual attribute characteristic is “small household appliance enthusiast”, then the relevant content of “breakfast machine”, “smart alarm clock”, “electric heating” and other household appliances can be added to the processed dynamic information; at the same time, the key word embedding method can be used to implant the key words attractive to the friend in the effective dynamic information, such as the virtual attribute characteristic is “coupon enthusiast”, the words such as “exclusive discount” and “limited purchase” can be added in the dynamic to improve the friend's attention. It can also be processed from the form optimization method, such as adding pictures, short videos or dynamic special effects to the dynamic information, enriching the information form to avoid the monotonous text dynamic being easily ignored, and the dynamic information length can also be optimized according to the information content and friend preference, such as long information is suitable for popular science content, and short information is suitable for promotion and reminder content. Emotional language can also be injected into the dynamic information to increase the friend's account of the sense of substitution and interactive willingness, for example, the original dynamic information is “xx new product is on sale, try it out”, and the processed secondary dynamic information can be “fellows, try it out, xx new product, guess if it is a pit or a prize! ”, the emotional language needs to conform to the role setting of the virtual digital person, avoiding contradiction or discomfort. In addition, the dynamic information can also be integrated into specific scenes to enhance the attractiveness, for example, the original dynamic information is “the sound quality of this earphone is amazing”, and the virtual attribute characteristic is “game enthusiast”, the processed secondary dynamic information can be “with this earphone, the footsteps and breathing sound of the enemy character in xx game can be clearly heard, 10 consecutive victories”.

[0045] On the other hand, the dynamic information content can be adjusted from the time dimension to optimize the release time and frequency, such as it can include analyzing the active period of the friend group (such as morning, noon, and night) to release the dynamic information at a certain time, so as to avoid the release in the non-active period which may lead to the dynamic information being ignored; the release frequency of the 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 the high frequency leading to the friend's aversion or the low frequency reducing the presence; the release frequency can also be adjusted based on the interaction performance of the friend group (such as the number of likes and the comment activity).

[0046] In another aspect, the dynamic information can also be adjusted from the visibility dimension to accurately locate the audience. Firstly, specific content can be published to different friend groups to avoid irrelevant information being displayed to irrelevant friend users; for example, technology information can be published to a "digital enthusiasts" group, and dining-related dynamic information can be published to a "food enthusiasts" group. Secondly, specific dynamic information can be opened to target friends with frequent interactions to avoid redundant dissemination, so as to avoid non-target users being disturbed by the widespread dissemination of dynamic information.

[0047] In another aspect, data processing can also be performed from the data processing algorithm dimension to improve processing accuracy. Specifically, the following points can be included: firstly, for hot topics, multiple friend accounts have a high probability of publishing dynamic information related to the same topic, at which time key elements can be extracted from multiple dynamic information to generate synthesized secondary dynamic information. Secondly, labels or titles can be used to generate labels that match virtual attribute features for dynamic information, such as "# fitness check-in" and "# smart home", so that friend accounts can quickly understand the main content of the secondary dynamic information.

[0048] In another aspect, data processing can also be performed from the interaction dimension to enhance the sense of participation and interaction of friend users, such as introducing interactive content, such as voting, question and answer, comment incentives, etc. in the secondary dynamic information; or adding content that guides behavior, such as "link here, everyone can see it", "welcome partners to forward", etc. in the secondary dynamic information.

[0049] In addition, when processing data, friend privacy, positive values, content copyright, information legality, etc. need to be considered. Specifically, sensitive data and private information about friends should be excluded when processing dynamic information, and the processed content should conform to mainstream values to avoid disputes due to inappropriate content. At the same time, when dynamic information involves some creative and high-value content, the friend user's authorization and / or information source needs to be requested, or the relevant content needs to be removed. In addition, the key content in the dynamic information should be ensured to be true and reliable by combining other information channels to avoid trust crises caused by false information.

[0050] In one embodiment, the virtual attribute feature includes a primary attribute feature and a plurality of secondary attribute features; the identification processing of the effective dynamic information can be adjusted to: identify the effective dynamic information matching the primary attribute feature; and the processing of the secondary dynamic information can be adjusted to: randomly select one secondary attribute feature, and process the effective dynamic information based on the selected secondary attribute feature to obtain the secondary dynamic information.

[0051] The virtual attribute feature refers to the attribute feature possessed by the virtual digital person account, and is used to imitate the preferences and behaviors of the mainstream user category and the friend users. The virtual attribute feature includes a primary attribute feature and multiple secondary attribute features. The primary attribute feature corresponds to the overall feature of the mainstream user category, such as “keen on discount promotion”, “concerned about electronic products”, etc., and is mainly used for matching the virtual digital person with a large range of user groups. The secondary attribute feature is a feature further subdivided on the basis of the primary attribute feature, such as “concerned about smart home devices among electronic products”, “prefer active time at midnight”, etc., and is used to more accurately simulate the needs or interests of different friend users.

[0052] In implementation, the virtual digital person account can preliminarily filter the dynamic information published by the friend users through the primary attribute feature. Specifically, the key content of the dynamic information, such as keywords, tags, semantic information, etc. in the dynamic, can be analyzed to determine whether it conforms to the primary attribute feature of the virtual digital person. For example, if the primary attribute feature is “keen on discount promotion”, keywords such as “limited-time offer” and “big promotion” in the dynamic information can be used as matching basis; if the primary attribute feature is “concerned about electronic products”, mentioning “new earphones” and “smart bracelets” in the dynamic information 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.

[0053] In the data processing process, the virtual digital person account selects one of the multiple secondary attribute features through a random selection method for processing the valid dynamic information. The random selection method can avoid the secondary attribute being too concentrated in a certain feature, thereby covering more interest areas of the friend users; and if the valid dynamic information is processed randomly, different secondary dynamic information can be generated, which can avoid the content being too single and also avoid the same valid dynamic information being processed and published multiple times, causing frequent screen flashing and affecting the experience of the friend users.

[0054] Then, the virtual digital person account can process the valid dynamic information to generate secondary dynamic information in combination with the selected secondary attribute feature. The specific processing method can refer to the content of step 105. In this way, the random selection of the secondary attribute feature avoids the single processing content, making the dynamic of the virtual digital person more diverse and rich, thereby covering more interest points of the friend users; if multiple secondary dynamic information is generated from the same valid dynamic information, the random selection of the secondary attribute can also ensure that the secondary dynamic information has different styles and avoids the “repeated screen flashing” from interfering with the friend users. Of course, for one valid dynamic information, only one secondary attribute feature can be selected to generate one secondary dynamic information, so as to further ensure the rationality of the dynamic information publishing and avoid the “screen flashing behavior” affecting the experience of the friend users.

[0055] Of course, for specific scenarios, the effective dynamic information can be jointly processed 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 goods", and the secondary attribute feature 2 is "pay attention to smart device recommendation". The processed secondary dynamic information can be "high cost performance smart speaker recommendation, now half price, experience smart life, today!"

[0056] In one embodiment, the method further comprises: counting interaction data of all friend user accounts on the secondary dynamic information; assigning a first weight value to the corresponding secondary attribute feature according to the interaction data; extracting the attribute features of the secondary dynamic information as candidate secondary attribute features, and assigning a second weight value to the candidate secondary attribute features according to the interaction data; periodically re-determining the secondary attribute features of the virtual digital person account and the corresponding selection probability according to the historical weight values of the secondary attribute features and the candidate secondary attribute features.

[0057] The interaction data can be the behavior data generated by the friend user account on the secondary dynamic information published by the virtual digital person account, including likes, comments, forwards, and collections. The interaction data reflects the attention and interest of the friend user on the dynamic information, and is an important basis for optimizing the secondary attribute features of the virtual digital person account.

[0058] In implementation, the first interaction platform can periodically count the interaction data of the secondary dynamic information published by the virtual digital person account, and associate the statistical results with the secondary attribute features as the basis data for assigning weight values in the next step. For each secondary attribute feature, the total amount of interaction data of all secondary dynamic information related to it can be counted. For example, the secondary attribute feature A is "pay attention to high cost performance goods", the interaction data of the secondary dynamic information 1 associated with it is 10 likes and 5 comments; the interaction data of the secondary dynamic information 2 associated with it is 20 likes and 3 forwards; then the first weight value A of the secondary attribute feature = 10·k1+ 5·k2 + 20·k1+ 3·k3, where k1, k2, k3 are the unit coefficients corresponding to the like, comment and forward behaviors, respectively. Of course, the same kind of interaction behavior in different secondary dynamic information can also correspond to different unit coefficients. The first weight value can reflect the importance of the existing secondary attribute feature in the overall preference of the friend user, and reflect the adaptation degree of the current virtual digital person account features to the friend user's demand.

[0059] In addition, the attribute feature extracted from the secondary dynamic information may not belong to the existing secondary attribute feature, but may reflect the potential interest of the friend user. Therefore, other attribute features contained in the secondary dynamic information can be extracted as candidate secondary attribute features, and a second weight value can be assigned to the candidate secondary attribute feature according to the interaction data. For example, the secondary dynamic information is “smart lighting device is on sale! ”, the attribute feature of the dynamic information “smart lighting device” can be extracted, and if “smart lighting device” does not belong to the existing secondary attribute feature, it can be used as a candidate secondary attribute feature; in combination with the interaction data of the secondary dynamic information, such as 15 likes and 10 comments, the second weight value B = 15 · k1 + 10 · k2 is assigned to the candidate secondary attribute feature “smart lighting device”. The second weight value reflects the potential importance of the candidate secondary attribute feature in the friend user, and is used to identify the potential interest or emerging needs of the friend user.

[0060] Based on the above settings, the first interaction platform can periodically (such as every week, every month) analyze the historical weight values of the secondary attribute features and the 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 value of a candidate secondary attribute feature exceeds the lower limit of the first weight value of the existing secondary attribute feature, it is added to the secondary attribute feature set; if the first weight value of a certain existing secondary attribute feature is below a certain threshold for consecutive multiple periods, it is removed from the secondary attribute feature set. Of course, the present embodiment does not limit the adjustment rules of the secondary attribute features.

[0061] Thereafter, the selection probability of each secondary attribute feature can also be allocated according to the proportion of the first weight value and the second weight value, for example, the secondary attribute feature with a higher weight value can be set to have a higher selection probability, thereby improving the accuracy of the virtual digital person account. In this way, by assigning weight values to the secondary attribute features based on the interaction data, the changing trend of the interest points of the friend user can be reflected; by periodically analyzing the weight values, the emerging interest points of the friend user can be timely discovered, and the secondary attribute features that are no longer of interest can be removed, so that the dynamic publication of the virtual digital person account is more in line with the overall dynamic preferences of the friend user.

[0062] Of course, positive or high-frequency keywords can also be extracted from the interaction data of the friend as a reference basis for candidate features. Different friend user groups can also be set with weights, such as for friend users with higher attention, higher weights can be assigned to their interaction data; for friend users with lower interaction frequency, the weight values of the features corresponding to their interest points can be appropriately increased to attract their participation in the interaction. When using weight values to re-determine secondary attribute features and selection probabilities, the future needs of the friend user can also be predicted by using the weight value trend, such as by analyzing the gradual upward trend of the weight value of a candidate secondary attribute feature, it can be predicted that the interest of the friend user in the candidate secondary attribute feature will continue to grow, so that it can be preferentially included in the secondary attribute feature set.

[0063] In an embodiment, the method further comprises: when the secondary attribute feature is determined to be a high-frequency level attribute feature according to the historical weight, determining a target friend user account corresponding to the high-frequency level attribute feature; performing data processing on the effective dynamic information based on the high-frequency level attribute feature at a fixed frequency to obtain specific secondary dynamic information; and 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.

[0064] In implementation, the secondary attribute feature corresponding to the secondary dynamic information frequently interacted by the friend user can be determined according to the historical weight statistical result, and is defined as a high-frequency level attribute feature. The high-frequency level attribute feature can reflect the main interest point or attention direction of the friend user group in a certain field, which can be a secondary attribute feature related to multiple interaction data items and having a high or rising weight for a continuous period of time; for example, if the historical weight of a secondary attribute feature “smart wearable device” is in the top three for a continuous period of time, it can be determined to be a high-frequency level attribute feature. Further, the target friend user account highly matched with the high-frequency level attribute feature and frequently interacted can be analyzed, which can be a friend who frequently likes, comments, forwards, and performs other interaction behaviors on the dynamic information under the high-frequency level attribute feature; for example, if a friend user A frequently likes and comments on the dynamic information related to “smart wearable device”, the user A can be determined to be the target friend user account of the high-frequency level attribute feature of “smart wearable device”.

[0065] Based on the above setting, in addition to randomly selecting a secondary attribute feature to generate secondary dynamic information, data processing can be performed on the effective dynamic information based on the high-frequency level attribute feature at a fixed frequency to obtain specific secondary dynamic information, and the specific secondary dynamic information can be published through the virtual digital human account and set to be visible only to the target friend user account, avoiding interference with irrelevant users; if the interaction 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 device” decreases, the visibility of the relevant specific secondary dynamic information for the friend user can be cancelled.

[0066] It can be understood that the screening of the high-frequency level attribute feature and the determination of the target friend user account can ensure the accuracy of the push, and the high-frequency level attribute feature and the interaction data can be used to accurately locate; the data processing link of the high-frequency level attribute feature can ensure that the content of the specific secondary dynamic information is highly matched with the demand of the target user, thereby increasing the attraction of the content; and the visibility setting can reduce the interference of irrelevant users and realize differentiated content dissemination.

[0067] In an embodiment, the method further comprises: when the frequency of interaction of the virtual digital person account with the friend account meets a preset standard, allocating account registration resources to the virtual digital person account; requesting, by the virtual digital person account, account information of the friend user account of the first interaction platform on other interaction platforms; registering, based on the account registration resources, a virtual digital person associated account on the other interaction platforms, and establishing a friend relationship on the other interaction platforms based on the account information; obtaining, by the virtual digital person associated account, friend dynamic information published by the friend user account on the other interaction platforms, and identifying effective dynamic information matching the virtual attribute characteristics; and performing data processing on the effective dynamic information to obtain secondary dynamic information, and publishing the secondary dynamic information on the first interaction platform by the virtual digital person account.

[0068] In implementation, when the frequency of interaction of the virtual digital person account with the friend user account meets a preset standard (for example, the frequency of likes, comments, and private messages reaches a certain threshold), the first interaction platform can determine that the virtual digital person account has the potential to further expand the social circle. Based on this determination, the first interaction platform can allocate account registration resources to the virtual digital person account. The account registration resources can be a fixed number of registration opportunities, an expansion quota, or an identity information template generated by the virtual digital person system for the creation of cross-platform accounts. Based on the established relationship with the friend user account on the first interaction platform, the virtual digital person account initiates a request through the first interaction platform to obtain the account information of the friend user account on other interaction platforms. These information may include usernames, user IDs, platform identifiers, etc., and must be obtained on the premise of authorization by the friend user account to protect privacy and security. Based on the allocated account registration resources, the first interaction platform can automatically or semi-automatically register the associated account of the virtual digital person in other interaction platforms. After registration is completed, the first interaction platform can use the obtained friend user account information to establish an association relationship with the original friend user account in other interaction platforms, expanding the social network range of the virtual digital person.

[0069] Further, the dynamic information published by the target friend user account on the other interaction platforms can be obtained through the virtual digital person associated account. The first interaction platform can identify and screen these dynamic information, and select effective dynamic information matching the virtual attribute characteristics of the virtual digital person account. For example, if the virtual attribute characteristics of the virtual digital person account are inclined to "technology information", the focus can be on screening technology-related dynamics published by the target friend. For the identified effective dynamic information, further data processing can be performed, such as reorganizing the content, optimizing the presentation form, adding multimedia content, etc., to improve the attractiveness and propagation effect of the dynamic information. The processed dynamic information is in the form of secondary dynamic information, which is published to the first interaction platform through the virtual digital person account.

[0070] Based on the above scheme, through the creation of the virtual digital person associated account, the friend relationship originally limited to the first interaction platform is extended to other interaction platforms. This cross-platform friend relationship expansion can effectively expand the influence range of the virtual digital person account and break the data silo problem between different platforms. By actively obtaining and integrating dynamic information of different platforms using the virtual digital person associated account, cross-platform data can be seamlessly connected and presented in a unified form, improving the completeness and expressiveness of dynamic content. By filtering effective dynamic information and processing, more targeted and attractive content can be provided for friend users of the first interaction platform, avoiding repeated and low-quality information flooding and further fitting the interests and preferences of friend users. The above dynamic information integration and publishing mechanism not only enriches the content system of the virtual digital person account, but also enhances the user experience, further improving the user's dependence on the virtual digital person and the interaction frequency.

[0071] It is worth mentioning that when implementing the above scheme, attention should be paid to the data privacy and content authorization of friend users, and when registering accounts on other platforms, the account management rules of other platforms should be followed to ensure compliance, or a cooperation mechanism should be established with other platforms.

[0072] In one embodiment, after identifying the effective dynamic information matching the virtual attribute features, it further includes: determining the platform attribute features corresponding to the first interaction platform and the other interaction platforms respectively; processing the effective dynamic information according to the platform attribute features to obtain secondary dynamic information matching the target interaction platform; publishing the secondary dynamic information on the target interaction platform through the virtual digital person account and / or the virtual digital person associated account.

[0073] 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 published by the target friend user account. The product release, evaluation and other content. For the first interaction platform and other interaction platforms, the server respectively determines the platform attribute characteristics of each interaction platform, for example, the video interaction platform can be biased towards short video, audio-visual performance strong content; the e-commerce platform focuses on commodity recommendation, price and promotion information; the knowledge sharing platform mainly uses text or long content to explain knowledge, after determining these platform attribute characteristics, the rule basis for dynamic information processing of different interaction 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 interaction platform. The form of data processing can be, for example, converting text dynamic into short video suitable for the video platform, or adding commodity information tags for the e-commerce platform, or supplementing background knowledge related to the dynamic or detailed explanation in the knowledge sharing platform. After processing different secondary dynamic information, it can be published through the virtual digital human account (first interaction platform) and / or the virtual digital human associated account (other interaction platforms) in the target interaction platform. For example, the virtual digital human account displays the related commodity link in the first interaction platform, and at the same time publishes the video version of the dynamic content in the video platform through the associated account, and publishes a brief text dynamic in the knowledge sharing platform.

[0074] Based on the above content, by identifying the attribute characteristics of different interaction platforms, the processing form of dynamic information can be adapted to the platform demand. For example: in the e-commerce platform, the commodity information and shopping guide function are enhanced, so that the user can more conveniently obtain the product related content and directly complete the purchase; in the video platform, the dynamic information is presented in the form of short video, which not only conforms to the user habit, but also can significantly improve the attraction of the content. According to the content consumption habit of the target platform user, the processing of the dynamic information improves the propagation effect of the secondary dynamic information, and avoids the user loss or interaction rate decrease caused by the format or content mismatch. Through the formation of unified and flexible content publishing strategy between different platforms, the interaction stickiness of the virtual digital human account and its associated account is enhanced.

[0075] In one embodiment, after publishing the secondary dynamic information in the target interaction platform, it further includes: counting the cross-platform interaction data of the secondary dynamic information of the same effective dynamic information in each target interaction platform; extracting the effective information feature set in each cross-platform interaction data, determining the set similarity between the effective information feature sets; according to the set similarity, updating the user feature similarity of the friend user account corresponding to the cross-platform interaction data; based on the user feature similarity, performing a friend recommendation process between user accounts on the first interaction platform.

[0076] In implementation, after publishing the secondary dynamic information on the target interaction platform, the interaction data of the same effective dynamic information on each target interaction platform can be counted: such as collecting behaviors of the user, such as likes, comments, and forwards, to form an interaction data set; according to the characteristics of different target platforms, the weight is assigned according to the behavior type. For example, the weight of "comment" on the video platform may be higher than "like", and the weight of "click to buy" on the e-commerce platform is higher.

[0077] Based on the statistical interaction data, the effective information feature set of each platform is extracted, including: user behavior preference features: such as interaction frequency, behavior type distribution; content preference features of dynamic information: such as keywords involved, topic labels (such as "electronic products", "outdoor activities"); behavior history association features of user accounts: such as whether the user continuously interacts with a certain type of dynamic information. For the effective information feature set extracted in different target platforms, the set similarity between the feature sets is calculated: through the intersection size or weight distribution overlap degree of the feature set, the similarity of different users in content preference and behavior mode is measured; for example, user A and user B both frequently participate in dynamic interaction about "new technology products", and the interaction mode (such as comment frequency) is similar, then the set similarity is high. According to the set similarity, the user feature similarity of the friend user account on the first interaction platform is updated: through cross-platform interaction data, the interest points or behavior characteristics of the user account on the first interaction platform are inferred; the update of the feature similarity may involve a dynamic weighting mechanism. For example, the weight of the latest interaction data is high, and the weight of the historical data is low, to reflect the dynamic change of user interest. On the first interaction platform, according to the user feature similarity, the friend recommendation process is started: specifically, for two users with similar features on other interaction platforms, if the accounts of the two on the first interaction platform can be determined, then the friend recommendation process between the two accounts on the first interaction platform can be performed; if only one of the users has an account on the first interaction platform, the account information of the other account on the other interaction platform can be recommended to the account, so that the account user can add friends on the other interaction platform.

[0078] It can be understood that, since the user data of the target interaction platform is generally not directly available, the user characteristics can be indirectly inferred through the cross-platform interaction data, for example, a user frequently clicks on a certain type of goods on an e-commerce platform, and likes a related review video on a video platform, indicating that he has a high interest in this type of goods. This feature inference method based on behavior data can make up for the lack of direct information due to platform data barriers. By calculating the similarity of the set and the similarity of the user characteristics, the interest characteristics of the cross-platform users can be fed back to the first interaction platform to realize friend recommendation. Users with high similarity are more likely to have common topics or interests, increasing the success rate of establishing a friend relationship; compared with the random recommendation method, the recommendation result of the above scheme is more accurate, and the user experience is better. Further, by using cross-platform data, comprehensive analysis of user characteristics of different interaction platforms can be realized, effectively expanding the potential friend range of users, enhancing the construction efficiency of the user relationship network of the first interaction platform, and also enhancing the dependence and recognition of users on the first interaction platform.

[0079] In one embodiment, the method is applied to a distributed system including a center node and at least one level of edge nodes, multiple edge nodes of the same level serving different areas, an upper level edge node corresponding to at least one lower level edge node, the service area of the upper level edge node covering the service area of the lower level edge node; wherein the center 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; the secondary edge node is configured to filter the optional user account whose address information belongs to its service area, and establish a personal friend relationship with the optional user account under the mainstream user category based on the virtual digital human account; the secondary edge node is also configured to extract the regional characteristic information of its service area, and the regional characteristic information is used to generate secondary dynamic information when processing the effective dynamic information.

[0080] The distributed system includes a center node and at least one level of edge nodes, has a hierarchical structure, the center node serves as a 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 interaction with user accounts in its coverage area and processing and publishing of dynamic information.

[0081] The center node can define mainstream user categories, such as "young consumers", "maternal and infant users", and "sports enthusiasts", in combination with user behavior data analysis, and configure the properties of the virtual digital human account according to the typical characteristics of each mainstream user category (such as the topics of interest and content preferences), including the personal profile, interest tags, interaction style, etc. After creation, the center node can distribute the virtual digital human account to the secondary edge nodes for personalized operation in various regions. Each secondary edge node filters user accounts belonging to the service area according to the address information of the service area. Specifically, the secondary edge node can identify users belonging to the current edge node service area according to the registered address, geographic positioning or interaction record of the user account; further filter users who meet the mainstream category of the virtual digital human account from the optional user accounts as potential friend objects of the virtual digital human account. The screening process can combine various matching strategies, such as content interest similarity and interaction frequency. 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, and after establishing the friend relationship, the virtual digital human account can interact with the user account through dynamic sharing, thereby enhancing user stickiness and activity.

[0082] The secondary edge node can further extract regional feature information from its service area, which can include user interest features, such as analyzing the preferences of users in the service area for specific topics based on their behavior data, such as "electronic product discounts" and "festival activities"; it can also include language and cultural features, such as the main language used in the service area and the popular expression method; it can also include hot events and demands, such as extracting real-time hot information in the region, such as large-scale activities, promotional seasons, or region-specific consumption habits, etc. In this way, the secondary edge node can use regional feature information to process effective dynamic information and generate secondary dynamic information. During processing, the dynamic information content can be customized according to the characteristics of the service area. For example, in a region that prefers shopping, the dynamic information will emphasize product information (such as price, discount, purchase link); in a region that prefers video, the dynamic information can be converted into short video form; in a knowledge sharing region, the dynamic information focuses more on the explanation of core concepts and the depth of content. The processed secondary dynamic information can be published in the respective service areas by the virtual digital human account, and each edge node independently publishes dynamic information to avoid content duplication in different regions, thereby more accurately meeting user needs and improving the dissemination efficiency of dynamic information.

[0083] Based on the above scheme, the creation of virtual digital human accounts is centralized in the central node, which can ensure the global consistency of virtual digital human accounts; the data filtering and dynamic processing are dispersed to the edge nodes, reducing the computational and management burden of the central node. The distributed architecture design enhances the scalability of the system, enabling it to flexibly cope with large-scale user distribution and dynamic content demand. At the same time, the regional processing method of the edge node improves the efficiency of dynamic information release 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 user experience; the regional characteristics of dynamic information can also help virtual digital human accounts more effectively participate in local user interactions, enhancing the credibility and influence of the accounts.

[0084] In one embodiment, the secondary edge node is also configured to select high-quality secondary dynamic information according to the interaction data of the secondary dynamic information it maintains, and send the high-quality secondary dynamic information to its corresponding superior edge node; the superior edge node is configured to process the high-quality secondary dynamic information to generate tertiary dynamic information, and send the tertiary dynamic information to the rest of the secondary edge nodes corresponding to it, so that the rest of the secondary edge nodes publish the tertiary dynamic information through the maintained virtual digital human accounts.

[0085] In implementation, each secondary edge node can filter out high-quality dynamic information according to the interaction data of the secondary dynamic information in its service area. First, the user interaction data of the secondary dynamic information published on the virtual digital human account can be collected, including the number of likes, the number of comments, the forwarding rate, etc. Then, by setting evaluation indicators (such as threshold values of interaction data or weighted score models), dynamic information with high user attention and interaction rate can be selected. Finally, the selected high-quality dynamic information is marked with a high-quality label for subsequent reception and processing by the superior edge node. After that, the secondary edge node uploads the high-quality dynamic information to the corresponding superior edge node, which further processes the high-quality dynamic information. Specifically, the superior edge node can combine the high-quality dynamic information from multiple secondary edge nodes to remove redundant content and extract core hot information, forming a higher level of content summary; and can adjust the high-quality dynamic information according to the overall characteristics of the area covered by the superior edge node (such as language habits, cultural commonalities), such as translating the dynamic information into different languages in different language areas; for multi-cultural areas, the dynamic information can be extracted to be universally applicable, avoiding incompatibility due to cultural differences. After processing the tertiary dynamic information, the superior edge node can send the tertiary dynamic information to other secondary edge nodes under its jurisdiction, and each secondary edge node can share the tertiary dynamic information with users in the region through the virtual digital human account it maintains.

[0086] Based on the above scheme, on the one hand, through the screening of secondary dynamic information by the secondary edge node, high-quality content in the local region can be uploaded to the superior node level by level, realizing hierarchical filtering and propagation of dynamic information. In the local region, high-quality dynamic information can reflect the main needs of users in the region, ensuring the accuracy of the uploaded content. In the global range, through the processing of the superior node, dynamic information can be transmitted across regions, thereby realizing efficient sharing of high-quality content. On the other hand, the processing link of the superior edge node can balance the regional characteristics of each secondary node, making the three dynamic information have cross-regional adaptability, avoiding understanding deviation or cultural conflict caused by regional differences in content transmission, improving the universality of dynamic information, and making it accepted and propagated by users in a wider area.

[0087] By adopting the virtual digital human interaction service method disclosed in the present application, the user data set of the first interaction platform is obtained, the high-frequency attribute features in the user data set are summarized, and a plurality of mainstream user categories are determined, so as to accurately depict the behavior characteristics of the target user group and provide data support for subsequent interaction. 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 scene in the real social network. This way can effectively cover individual users in the mainstream user category, significantly improve the association 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.

[0088] By setting the virtual attribute features of the virtual digital human account based on the attribute features 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 degree and attractiveness of the interaction content. At the same time, the virtual digital human account can obtain the friend dynamic information published by its friend users and identify effective dynamic information matching the virtual attribute features. This precise screening mechanism can avoid the interference of redundant or invalid content, improving the efficiency and quality of dynamic information processing.

[0089] In addition, by processing the effective dynamic information to generate secondary dynamic information and publishing the secondary dynamic information by the virtual digital human account, the content can be re-transmitted and diffused in the user group. This method not only fully taps the potential value of user dynamic information, but also improves the activity and influence of the virtual digital human account, forming a continuous interaction chain in the social network, enhancing the user's sense of participation and interactive experience.

[0090] The method automatically creates a virtual digital person account based on the attribute characteristics of user data, filters effective dynamic information, and processes and publishes secondary dynamic information, thereby realizing the automation and intelligentization of virtual digital person interaction. This approach reduces the cost and complexity of manual operation while ensuring the accuracy and relevance of the content. In practical applications, the method can adapt to large-scale user scenarios, improve the platform's response to user needs, and provide technical support for content distribution and user activity improvement in social networks, forming an efficient, intelligent, and continuously optimized virtual digital person interaction service process.

[0091] It should be understood that, although each step in the flowchart involved in the above embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in the above embodiments can include multiple steps or stages, which 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 sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0092] Based on the same inventive concept, as shown in Figure 2 The embodiments of the present application also provide a virtual digital person interaction service device 200, which comprises:

[0093] a feature summary classification module 201 configured to obtain a user data set of a first interaction platform, summarize high-frequency attribute characteristics in the user data set, and determine a plurality of mainstream user categories;

[0094] a digital person account creation module 202 configured to create a virtual digital person account for each mainstream user category and establish a personal friend relationship between the virtual digital person account and a user account in the mainstream user category;

[0095] a virtual attribute setting module 203 configured to set virtual attribute characteristics of the virtual digital person account based on attribute characteristics of friend user accounts of the virtual digital person account;

[0096] a dynamic information identification module 204 configured to obtain friend dynamic information published by friend user accounts of the virtual digital person account through the virtual digital person account and identify effective dynamic information matching the virtual attribute characteristics;

[0097] The secondary dynamic publishing module 205 is configured to process the effective dynamic information to obtain secondary dynamic information, and publish the secondary dynamic information through the virtual digital person account.

[0098] In one of the embodiments, the virtual attribute features include a primary attribute feature and a plurality of secondary attribute features.

[0099] The dynamic information identifying module 204 is specifically configured to:

[0100] identify effective dynamic information matching the primary attribute feature.

[0101] The secondary dynamic publishing module 205 is specifically configured to:

[0102] randomly select one of the secondary attribute features, and process the effective dynamic information based on the selected secondary attribute feature to obtain secondary dynamic information.

[0103] In one of the embodiments, the virtual attribute setting module 203 is further configured to:

[0104] statistically analyze interaction data of all friend user accounts on the secondary dynamic information.

[0105] assign a first weight value to the corresponding secondary attribute feature according to the interaction data.

[0106] extract attribute features of the secondary dynamic information as candidate secondary attribute features, and assign a second weight value to the candidate secondary attribute features according to the interaction data.

[0107] periodically re-determine the secondary attribute features of the virtual digital person account and the corresponding selection probability according to historical weight values of the secondary attribute features and the candidate secondary attribute features.

[0108] In one of the embodiments, the secondary dynamic publishing module 205 is further configured to:

[0109] when the secondary attribute feature is determined to be a high-frequency secondary attribute feature according to the historical weight value, determine a target friend user account corresponding to the high-frequency secondary attribute feature.

[0110] process the effective dynamic information based on the high-frequency secondary attribute feature at a fixed frequency to obtain specific secondary dynamic information.

[0111] publish the specific secondary dynamic information through the virtual digital person account, and set the specific secondary dynamic information to be visible to the target friend user account.

[0112] In one of the embodiments, the secondary dynamic publishing module 205 is further configured to:

[0113] When the friend interaction frequency of the virtual digital person account meets a preset standard, account registration resources are allocated to the virtual digital person account;

[0114] The virtual digital person account requests account information of a friend user account of the first interaction platform from other interaction platforms;

[0115] Based on the account registration resources, a virtual digital person associated account is registered on the other interaction platforms, and a friend relationship of the other interaction platforms is established through the account information;

[0116] Through the virtual digital person associated account, friend dynamic information published by the friend user account on the other interaction platforms is obtained, and valid dynamic information matching the virtual attribute characteristics is identified;

[0117] The valid dynamic information is processed to obtain secondary dynamic information, and the secondary dynamic information is published on the first interaction platform through the virtual digital person account.

[0118] In one embodiment, the secondary dynamic publishing module 205 is further configured to:

[0119] Determine platform attribute characteristics corresponding to the first interaction platform and the other interaction platforms, respectively;

[0120] According to the platform attribute characteristics, the valid dynamic information is processed to obtain secondary dynamic information matching a target interaction platform;

[0121] The secondary dynamic information is published on the target interaction platform through the virtual digital person account and / or the virtual digital person associated account.

[0122] In one embodiment, the secondary dynamic publishing module 205 is further configured to:

[0123] Statistical cross-platform interaction data of secondary dynamic information of the same valid dynamic information in each target interaction platform;

[0124] Extract an effective information feature set in each cross-platform interaction data, and determine a set similarity between the effective information feature sets;

[0125] According to the set similarity, update a user feature similarity of a friend user account to which the corresponding cross-platform interaction data belongs;

[0126] Based on the user feature similarity, a friend recommendation process between user accounts on the first interaction platform is performed.

[0127] Based on the same inventive concept, this 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. 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 and / or the edge nodes are configured to perform the processing of steps 101-105 above.

[0128] In one embodiment, the central node is configured to create virtual digital human accounts for each mainstream user category and send the virtual digital human accounts to the corresponding secondary edge nodes.

[0129] The secondary edge node is configured to filter address information belonging to its service area as an optional user account, and establish personal friend relationships with the optional user accounts under the mainstream user category based on the virtual digital human account;

[0130] The secondary edge node is also configured to extract regional feature information of its service area, which is used to generate the secondary dynamic information when processing the effective dynamic information.

[0131] 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 it maintains, and send the high-quality secondary dynamic information to its corresponding upper-level edge node.

[0132] The upper-level edge node is configured to process the high-quality secondary dynamic information to generate tertiary dynamic information, and send the tertiary dynamic information to its corresponding other secondary edge nodes, so that the other secondary edge nodes can publish the tertiary dynamic information through the virtual data person accounts they maintain.

[0133] In one embodiment, a computer device is provided, the internal structure of which can be shown as follows: Figure 3As shown in the figure. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. 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 external devices. The communication interface of the computer device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement a virtual digital human interaction service method.

[0134] Those skilled in the art can understand that, Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0135] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by the processor to implement the steps in each of the method embodiments described above.

[0136] In one embodiment, a computer program product is provided, and the computer program product includes a computer program. The computer program is executed by the processor to implement the steps in each of the method embodiments described above.

[0137] 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 for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0138] 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. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric 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 but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The processor involved in the embodiments provided in the present application can be a general processor, central processing unit, graphics processing unit, digital signal processor, programmable logic device, quantum computing-based data processing logic device, etc., without being limited thereto.

[0139] The technical features of the above embodiments can be combined in any way. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present disclosure.

[0140] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of protection of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A virtual digital human interactive service method, characterized in that, The method is applied to a distributed system of a server of a first interactive platform, the first interactive platform being an online shopping platform or a live shopping guide platform, the distributed system including a center node and at least one level of edge nodes, a plurality of edge nodes of the same level serving areas being different from each other; the method includes: obtaining a user data set of the first interactive platform, summarizing high-frequency attribute features in the user data set, and determining a plurality of mainstream user categories, the high-frequency attribute features being determined at least according to total consumption; the center node creates a virtual digital human account for each mainstream user category, and sends the virtual digital human account to a corresponding secondary edge node; the secondary edge node screens selectable user accounts whose address information belongs to the service area thereof, and establishes a personal friend relationship between the virtual digital human account and the selectable user accounts under the mainstream user category; the secondary edge node sets and updates virtual attribute features of the virtual digital human account at a certain frequency based on attribute features of the friend user accounts established by the virtual digital human account and behaviors of the friend user accounts, the attribute features of the friend accounts at least including a main commodity category and a price interval that the friend user accounts mainly focus on, the virtual attribute features including a main attribute feature and a plurality of secondary attribute features, the main attribute feature corresponding to an overall feature of the mainstream user category, and the secondary attribute features being used to simulate different friend users' demands or interests; the secondary edge node obtains friend dynamic information published by the friend user accounts through the virtual digital human account, performs semantic analysis on the friend dynamic information through natural language processing technology, and matches the virtual attribute features of the virtual digital human, to identify effective dynamic information matching the main attribute feature; the secondary edge node randomly selects one of the secondary attribute features, extracts regional feature information of the service area thereof, and based on the selected secondary attribute feature and the regional feature information, processes the effective dynamic information in a manner of adding secondary attribute feature related content and regional feature information of the service area to obtain secondary dynamic information, and publishes the secondary dynamic information through the virtual digital human account, to expand the propagation range of the dynamic information content; the method further includes: when a friend interaction frequency of the virtual digital human account meets a preset standard, allocating account registration resources for the virtual digital human account, the account registration resources being used to create a cross-platform account; requesting, through the virtual digital human account, account information of other interactive platforms of the friend user accounts of the first interactive platform that have been established, the other interactive platforms being cooperative platforms of the first interactive platform; registering a virtual digital human associated account on the other interactive platforms based on the account registration resources, and establishing a friend relationship of the other interactive platforms through the account information; obtaining friend dynamic information published by the friend user accounts of the other interactive platforms through the virtual digital human associated account, and identifying effective dynamic information matching the virtual attribute features. The effective dynamic information is data-processed to obtain secondary dynamic information, and the secondary dynamic information is published on the first interaction platform through the virtual digital person account to enrich the dynamic information content of the virtual digital person account.

2. The method of claim 1, wherein, The method further comprises: statistically obtaining interaction data of all friend user accounts on the secondary dynamic information; assigning a first weight value to a corresponding secondary attribute feature according to the interaction data; extracting an attribute feature of the secondary dynamic information as a candidate secondary attribute feature, and assigning a second weight value to the candidate secondary attribute feature according to the interaction data; periodically re-determining the secondary attribute features of the virtual digital person account and the corresponding selection probability according to the historical weight values of the secondary attribute features and the candidate secondary attribute features.

3. The method of claim 2, wherein, The method further comprises: when the secondary attribute feature is determined to be a high-frequency secondary attribute feature according to the historical weight value, determining a target friend user account corresponding to the high-frequency secondary attribute feature; data-processing 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 person account, and setting the specific secondary dynamic information to be visible to the target friend user account.

4. The method of claim 1, wherein, After the effective dynamic information matching the virtual attribute feature is identified, the method further comprises: determining platform attribute features corresponding to the first interaction platform and the other interaction platforms, respectively; data-processing the effective dynamic information according to the platform attribute features to obtain secondary dynamic information matching a target interaction platform; publishing the secondary dynamic information on the target interaction platform through the virtual digital person account and / or a virtual digital person associated account.

5. The method of claim 4, wherein, After the secondary dynamic information is published on the target interaction platform, the method further comprises: statistically obtaining cross-platform interaction data of the secondary dynamic information of the same effective dynamic information in each target interaction platform; extracting an effective information feature set in each cross-platform interaction data, and determining a set similarity between the effective information feature sets; updating a user feature similarity of a friend user account corresponding to the cross-platform interaction data according to the set similarity; based on the user feature similarity, performing a friend recommendation process between user accounts on the first interaction platform.

6. The method of claim 1, wherein, The upper edge node corresponds to at least one lower edge node, and the service area of the upper edge node covers the service areas of the lower edge nodes thereof.

7. The method of claim 6, wherein, The secondary edge node is further configured to select high-quality secondary dynamic information according to the interaction data of the secondary dynamic information maintained thereby, and send the high-quality secondary dynamic information to the upper edge node corresponding thereto; The upper edge node is configured to data-process the high-quality secondary dynamic information to generate tertiary dynamic information, and send the tertiary dynamic information to the remaining secondary edge nodes corresponding thereto, so that the remaining secondary edge nodes publish the tertiary dynamic information through the maintained virtual digital person accounts.

Images