Marketing training system based on digital human

Abstract

The invention discloses a marketing training system based on digital people. Comprising the following modules: a user interaction module, an interaction data module, a data processing module, a marketing knowledge base module, a marketing expert base module, an action sequence base module, a user answer data analysis module, a user portrait module, a digital real person module, a professional large model module and a diffusion model module. The method has the main advantages that the existing marketing training is generally on-line or off-line curriculum teaching, the personnel cost is reduced, the teaching does not need preparation, and the emotional problem does not exist; and 2, the method is not limited by time length, and has pertinence and diversity, and compared with the prior art, the method has the following main creativity: 1, real person live broadcast and human input are not needed, and marketing knowledge is trained by digital person live broadcast; 2, establishing a marketing knowledge base, and calculating similarity of knowledge points according to user reply by a marketing expert base; 3, a lecturer action sequence is preloaded, marketing training based on the digital person is not limited by duration, and human input is not needed.

Description

Technical Field

[0001] The present invention relates to the field of AI technology, and in particular to a marketing training system based on digital humans. Background Art

[0002] At present, many traditional industries and Internet industries need to market their products, so it is very necessary to provide marketing training for employees. In traditional marketing training, professional lecturers teach marketing knowledge, and the teaching form is generally offline, or online recording and broadcasting, live broadcasting, etc. Offline requires advance preparation, venue costs, team organization, face-to-face training and many other requirements. Online recording and broadcasting or live broadcasting also requires advance preparation, team operation, lecturers’ personal participation, and time constraints. These two methods are relatively troublesome to organize, with high personnel costs, lecturers worry about labor, emotional problems, and training teaching is not targeted and diverse. At present, digital humans are relatively rare in marketing training, and most of them are only 2D digital humans, without real human features and not too many movements; currently, the neural radiation field NeRF is used in digital human marketing training technology to build a real 3D model, and then combined with facial and body capture technology, animation and simulation, and real physical rendering technology to shape the real effect, and then use RIFE video interpolation multi-insertion point technology to achieve real action effects. In the process of RIFE video interpolation, due to inaccurate intermediate stream estimation or complex movements, motion boundary distortion and unnatural movements may occur during video interpolation; and high frame rate refresh is required, which requires high computing resources. In terms of marketing training, digital humans only learn knowledge points through the marketing knowledge base, with a shallow learning level, lack of interactivity, pertinence, diversity, and low training effect. In order to solve the above two drawbacks, it is necessary to invent a marketing training system and system based on digital humans. It is necessary to use RGB images without deep input to build a realistic digital real person with natural and smooth movements, and a high-quality and high-level digital marketing training expert with interactivity, pertinence, and diversity. Using this system for marketing training does not require organizing activities, has low personnel costs, does not require venues, and is not limited by time; it only needs one-click deployment, live broadcast anytime, anywhere, and the instructor will not be tired, and can broadcast training 24 hours a day without stopping. Summary of the invention

[0003] The present invention provides a method and system for digital human marketing training. In the field of digital marketing, it is necessary to train marketing knowledge. The present invention implements marketing training for digital humans by establishing marketing knowledge, a marketing expert database, a large marketing professional model and user portraits, calculates similarity scores for user replies, and uses a diffusion model to pre-generate digital human action sequences for pre-loading, thereby solving the above problems.

[0004] To achieve the above object, the present invention provides the following technical solutions: A marketing training system based on digital human, comprising the following steps: S1: User interaction module: used for language interaction between users and digital humans; S2: Data processing module: module for processing user interaction language; S3: Digital Human Lecturer Module: Digital Human Lecturer Model; S4: Marketing knowledge base: a module that stores basic marketing knowledge; S5: Marketing expert database: a module for storing marketing expert knowledge datasets; S6: Diffusion model: used to generate the action sequence specified by the instructor; S7: Action sequence library: stores the instructor’s action sequences; S8: Marketing professional big model: interact with users in real time and reply to users; S9: User answer data analysis module: analyzing user answer data; S10: User portrait module: build user portraits and provide targeted training.

[0005] Preferably, the user interaction module in S1 includes user voice input recognition, digital human voice generation and voice feature extraction, etc.

[0006] Preferably, the data processing module in S2 includes processing input natural language.

[0007] Preferably, the digital human lecturer module in S3 is mainly used to display a real person model, and can be constructed using neural radiation field NeRF, assisted by multimodal intelligent technology.

[0008] Preferably, the marketing knowledge base in S4 includes marketing knowledge called up by real lecturers for training.

[0009] Preferably, the marketing expert database in S5 includes real lecturers for training and training a large marketing professional model.

[0010] Preferably, the user answer analysis module in S9: performs similarity calculation on the knowledge points of the user's answer, first performs N-gram model statistical value calculation, and counts the frequency of occurrence of each bigram in the text (if it is 2-grammo), wherein count(n) represents the number of occurrences of N-gram n, and N represents the total number of N-grams in the text, wherein A and B respectively represent the N-gram sets of the two texts, and then combines the Jaccard similarity calculation.

[0011] Preferably, the user portrait module in S10 constructs a learning portrait of each user through similarity results of user responses.

[0012] In summary, due to the adoption of the above technology, the beneficial effects of the present invention are: Compared with existing technologies, the main advantages are: existing marketing training is generally taught through online or offline courses, 1. It reduces personnel costs, does not require preparation for teaching, and has no emotional problems; 2. It is not limited by time and is targeted and diverse.

[0013] Compared with existing technologies, the main creativity lies in: 1. No need for real-person live broadcast, no manpower investment, digital people live broadcast to train marketing knowledge; 2. Establish a marketing knowledge base and a marketing expert database to calculate the similarity of knowledge points based on user responses; 3. Pre-load the instructor's action sequence.

[0014] 3. Marketing training based on digital people is not limited by time and does not require human investment. It builds user portraits by establishing a marketing knowledge base and expert database combined with a marketing professional model and performing data calculation and analysis based on user questions and answers. This method can be used not only in marketing training, but also in many scenarios, such as course teaching, tour guide introductions, medical assistants, etc. It only requires the establishment of a corresponding knowledge system and the use of corresponding methods for data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention, so that other features, purposes and advantages of the present invention become more obvious. The accompanying drawings of the exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings: Figure 1 The present invention is a marketing training system diagram based on digital human. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0017] In the description of the present invention, it is necessary to understand that the terms indicating orientation or positional relationship are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0018] In the present invention, unless otherwise clearly specified and limited, the terms "install", "connect", "connect", "fix" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances of the specification.

[0019] The present invention provides Figure 1 A marketing training system based on digital humans is shown, a marketing training system based on digital humans, the system of the present invention comprises: a user interaction module, an interaction data module, a data processing module, a marketing knowledge base module, a marketing expert library module, an action sequence library module, a user answer data analysis module, a user portrait module, a digital real person module, a professional large model module and a diffusion model module.

[0020] 1. User interaction module: Users interact with digital humans, transmit data to the interactive data module, and analyze data in real time through the professional large model module.

[0021] 2. Interactive data module: transfer simple data to the data processing module, professional large model module and digital real person, and distribute data.

[0022] 3. Data processing module: perform logical processing on complex interactive data, and connect with professional large models, digital real people, marketing knowledge base and marketing expert database.

[0023] 4. Marketing knowledge base module: stores basic knowledge of the marketing industry and is used for digital live-action basic teaching and training.

[0024] 5. Marketing expert library module: A large number of marketing expert knowledge data sets for training professional large models, stored for advanced digital real-person teaching and training.

[0025] 6. Action sequence library module: Use the instructor action sequence set generated by the diffusion model to preload the action sequence when the instructor performs a specified action at a certain moment to ensure that the action is natural and smooth.

[0026] 7. User answer data analysis module: used to analyze user answer data and build user portraits.

[0027] 8. User portrait module: stores many user portraits to understand the user training level.

[0028] 9. Digital real-person lecturer module: The 3D model is constructed by neural radiation field NeRF, multimodal intelligent technology rendering, and action sequences are generated based on the diffusion model to realize the digital real-person lecturer. It combines the marketing knowledge base, marketing expert library and professional large models to conduct training and teaching with users, and is assisted by the user answer analysis module to build user portraits and understand the user's training level.

[0029] 10. Professional big model module: A professional marketing big model trained with marketing knowledge and marketing expert data sets. It analyzes user interactions in real time and connects with digital real-person lecturers to impart marketing knowledge or answer user questions.

[0030] 11. Diffusion model module: Based on the input of a lecturer picture, it imitates the specified teaching action video. First, it generates a 3D model of the lecturer, renders the clothing and different character postures, and then imitates different actions according to the specified video to generate an action sequence and store it in the action sequence library.

[0031] Main processing flow: In the digital human marketing training process, first, a real person 3D model is constructed through the neural radiation field, combined with multimodal intelligent technology, and a diffusion model is run to generate a variety of action sequences for the digital real person to choose, and a digital real person lecturer model is constructed. Secondly, marketing training first passes through the user interaction module, through the digital real person lecturer and data processing module, and then the marketing knowledge base module for basic training, and advanced targeted training is carried out through a professional large model obtained by training the knowledge data set of many marketing experts; users can flexibly adjust the knowledge points of the marketing knowledge base and the marketing expert library through interactive questions and answers with the lecturer to adapt to user targeted training, and calculate the mastery of each knowledge point based on the similarity between the user's input and response and the knowledge points of the marketing knowledge base and the marketing expert library to determine whether to learn new knowledge points or relearn. The above similarity calculation uses the N-gram and Jaccard similarity calculation to first use the N-gram ancestor statistics to calculate the frequency of occurrence of words in the input text and the reference text. The formula is as follows:

[0032] Here, count(n) represents the number of occurrences of N-tuple n, and N represents the total number of N-tuples in the text.

[0033] Combined with Jaccard similarity calculation, the formula is as follows:

[0034] Among them, A and B represent the N-tuple sets of two texts respectively.

Claims

1. A marketing training system based on digital human, characterized by: Mainly include: S1: User interaction module: used for language interaction between users and digital humans; S2: Data processing module: module for processing user interaction language; S3: Digital Human Lecturer Module: Digital Human Lecturer Model; S4: Marketing knowledge base: a module that stores basic marketing knowledge; S5: Marketing expert database: a module for storing marketing expert knowledge datasets; S6: Diffusion model: used to generate the action sequence specified by the instructor; S7: Action sequence library: stores the instructor’s action sequences; S8: Marketing professional big model: interact with users in real time and reply to users; S9: User answer data analysis module: analyzing user answer data; S10: User portrait module: build user portraits and provide targeted training.

2. According to claim 1, a marketing training system based on digital human, characterized in that: The user interaction module in S1 includes user voice input recognition, digital human voice generation, and voice feature extraction.

3. The marketing training system based on digital human according to claim 1, characterized in that: The data processing module in S2 includes processing input natural language.

4. The digital human-based marketing training system according to claim 1, characterized in that: The digital human lecturer module in the S3 is mainly used to display a real-life model, and can be constructed using the neural radiation field NeRF and assisted by multimodal intelligent technology.

5. The marketing training system based on digital human according to claim 1, characterized in that: The marketing knowledge base in S4 includes marketing knowledge called up by real lecturers for training.

6. The marketing training system based on digital human according to claim 1, characterized in that: The marketing expert database in S5 includes real lecturers for training and training of marketing professional models.

7. The digital human-based marketing training system according to claim 1, characterized in that: The user answer analysis module in S9: performs similarity calculation on the knowledge points of the user's answer, first performs N-gram model statistical value calculation, and counts the frequency of occurrence of each bigram in the text (if it is 2-grammo), where count(n) represents the number of occurrences of N-gram n, N represents the total number of N-grams in the text, where A and B represent the N-gram sets of the two texts respectively, and then combines the Jaccard similarity calculation.

8. The marketing training system based on digital human according to claim 1, characterized in that: The user portrait module in S10 is used to construct a learning portrait of each user based on the similarity results of the user's replies.

Images