Digital human interaction system and method based on artificial intelligence large model

Abstract

The invention belongs to the technical field of internet virtual reality, and particularly relates to a digital human interaction system and method based on an artificial intelligence large model. The digital human interaction method based on the artificial intelligence large model comprises the following steps: a target platform creates a first address according to a live broadcast schedule uploaded by a first terminal, generates a digital human video according to a business range of a merchant and plays the digital human video on the first address in real time, and a second terminal obtains the digital human video played in real time after accessing the first address; the second terminal sends request data to the target platform through the first address, the target platform generates a customized digital human video according to the content of the request data, creates a second address as an access address of the second terminal, and plays the customized digital human video to the second terminal through the second address, and taking the last digital human video node acquired by the second terminal as a pause node. According to the scheme, the problem that zero-delay interaction with the user cannot be realized during digital human live broadcast is solved.

Description

Technical Field

[0001] This solution belongs to the field of Internet virtual reality technology, and specifically relates to a digital human interaction system and method based on an artificial intelligence large model. Background Art

[0002] With the increase in computing power and the upgrade of multimodal models, the simulation, interaction and real-time feedback capabilities of digital human technology have been significantly improved. The "SenseTime: AI 2.0 Digital Human Empowered by Big Models White Paper" believes that digital humans empowered by AI have entered the L4 / L5 stage. Among them, digital humans in the L4 stage can interact intelligently, but require manual control in certain situations; digital humans in the L5 stage fully realize intelligent interaction and can handle full-cycle and universal problems.

[0003] In the past two years, with the continuous development of mobile Internet, online live broadcast technology has also been rapidly improved. As a new e-commerce business model that promotes sales through live video, e-commerce live broadcast has been favored by various businesses for its good communication effect.

[0004] At present, the intelligent interactive services provided by digital humans are still in the transformation period from live broadcast to digital human live broadcast (i.e., the L4 stage shown above). At present, digital human live broadcast and live broadcast coexist on the market. Digital human live broadcast is superior to most live broadcast in appearance and physical strength, but in the actual user experience, the degree of intelligence of digital human live broadcast is insufficient, and the low efficiency of interaction is highlighted when compared with live broadcast in the user experience process. The main reason why users choose to shop in the live broadcast room is that they can interact with the live broadcast object with zero delay. However, when there are many people in the live broadcast, the interaction efficiency of the live broadcast object is low (at present, the main way of interaction is that the live broadcast object interacts by reading the information sent by the user and answering it, but because only one message can be read at a time, the efficiency is low). Therefore, there is an urgent need for a digital human interaction system and method based on a large artificial intelligence model that can interact with zero delay and have high interaction efficiency. Summary of the invention

[0005] The purpose of this solution is to provide a digital human interaction system and method based on an artificial intelligence large model to solve the problem that digital humans cannot interact with users with zero delay during live broadcasts.

[0006] In order to achieve the above purpose, this solution provides a digital human interaction method based on an artificial intelligence large model, comprising the following steps: The target platform creates a first address according to the live broadcast schedule uploaded by the first terminal, generates a digital human video according to the business scope of the merchant and plays it in real time on the first address, and the second terminal obtains the real-time played digital human video after accessing the first address; The second terminal sends request data to the target platform through the first address. The target platform generates a customized digital human video according to the content of the request data, creates a second address as the access address of the second terminal, plays the customized digital human video to the second terminal through the second address, and uses the last digital human video node obtained by the second terminal as the pause node; After the customized digital human video is played, the digital human video after the pause node is obtained and played at an accelerated speed until the playback node of the digital human video at the second address is the same as the first address, and the access address of the second terminal is changed to the first address.

[0007] A digital human interaction system based on an artificial intelligence large model, comprising: The first terminal unit: used to formulate and upload the live broadcast schedule and provide the business scope of the merchant; Target platform unit: used to create a first address according to the live broadcast schedule, generate a digital human video according to the business scope of the merchant, and play it in real time on the first address; The second terminal unit is used to access the first address and obtain the real-time digital human video, and use the last obtained digital human video node as a pause node to send request data to the target platform; The target platform unit is also used to generate a customized digital human video according to the content of the request data when receiving the request data sent by the second terminal unit, and create a second address as the access address of the second terminal, and play the customized digital human video to the second terminal through the second address; after the customized digital human video is played, the digital human video after the pause node is obtained and played at an accelerated speed until the playback node of the digital human video at the second address is the same as the first address, and the access address of the second terminal is changed to the first address.

[0008] The principle and effect of this scheme are as follows: the target platform generates a first address (live broadcast address) according to the live broadcast schedule uploaded by the first terminal (merchant), generates a digital human video according to the business scope of the merchant, limits the publication time of the first address according to the live broadcast schedule provided by the merchant, and the target platform plays the digital human video in real time through the first address. The second terminal (audience) obtains the digital human video through the first address on the target platform (enters the live broadcast room to watch the digital human live broadcast). When the audience sends a bullet screen to the target platform (i.e., sends a data request), the second terminal (audience) transmits the data request to the target platform through the first address. The target platform extracts and analyzes the content of the bullet screen (data request) sent by the audience, and generates a customized digital human video and a second address (a second live broadcast room for answering the audience's bullet screen content) to answer the bullet screen (data request). The target platform transfers the audience who sends the bullet screen from the first address to the second address (transfers the audience from the first live broadcast room to the second live broadcast room), and transmits (plays) the customized digital human video to the audience through the second address. When the target platform transfers the audience, the pause node when the audience watches the digital human video in the first live broadcast room is used as the pause node. After the audience finishes watching the customized digital human video in the second live broadcast room, they start to watch the content missed in the first live broadcast room (that is, the content after the pause node of the digital human video) at an accelerated speed from the pause node until the content being played in the first live broadcast room is consistent with the content watched by the audience in the second live broadcast room (that is, the digital human video that the audience is watching at an accelerated speed and the digital human video being played in the first live broadcast room are the same playback nodes), and the audience is transferred from the second live broadcast room to the first live broadcast room, and the second live broadcast room is closed.

[0009] This solution uses digital human videos to explain the business content of the merchant to the audience, and receives questions (bullet screens) from the audience during the explanation process. After receiving the audience's questions (bullet screens), it provides exclusive Q&A services for the audience's questions, and continues to explain the merchant's business content to the audience after the Q&A service ends. Compared with traditional live broadcasts of real people, live broadcasts of real people can only answer one question during the live broadcast (that is, only one bullet screen can be answered at the same time), and when the real anchor answers, all the audience in the live broadcast room need to watch the whole process of the real anchor answering the question, and the anchor can continue the live broadcast after answering the question. The digital human anchor provided by this solution does not have the problem of being unable to cope with multiple roles (facing multiple bullet screens from multiple viewers, the real anchor can only answer one by one). This solution can answer inquiries from multiple viewers at the same time (replying to multiple data requests at the same time). Secondly, when answering questions, this solution will transfer the audience who need consulting services (audience who send barrage) to other live broadcast rooms (second live broadcast rooms), and provide answers to the audience in other live broadcast rooms (different second live broadcast rooms). This will not affect the viewing experience of audiences who have not asked questions (the viewing in the first live broadcast room will not be disturbed), and can provide question and answer services for audiences who ask questions (send barrage). It can also open up multiple live broadcast rooms at the same time to answer different questions from different audiences (the digital human anchor answers questions for multiple live broadcast rooms at the same time, and audiences in different second live broadcast rooms will not disturb each other). This method can not only prevent audiences from affecting each other when watching the live broadcast, but also improve the service quality of the digital human live broadcast, provide exclusive services for audiences with purchasing intentions, and increase the audience's willingness to buy. Furthermore, after answering the audience's questions, the digital human continues to play the explanation content (digital human video), which can continue to provide the audience with complete explanation services, and the audience will not be unable to watch the complete explanation content because the explanation time is occupied by answering questions; at the same time, it can also create a preferential experience for the audience to insert exclusive answers to questions in the complete explanation, further improve the audience's viewing experience and purchase intention, and improve the authenticity of the digital human live broadcast (the digital human in this solution is flexible and more like the mode of inserting answers to questions during real-life explanations). Finally, this solution accelerates the digital human video and transfers the audience back to the first live broadcast room, which can release more resources occupied by the second live broadcast room and reduce the operating burden of the target platform.

[0010] In summary, this solution solves the problem that digital humans cannot interact with users with zero delay during live broadcasts.

[0011] Furthermore, the request data includes a user identifier and request content, the user identifier includes account information and account category, and the account category is divided into a special category and a general category.

[0012] Furthermore, the target platform performs semantic extraction and analysis on the request content in the received request data, and records the account information of the special category, and connects the second terminal corresponding to the request content with the same semantics to the same second address; accumulates the number of request data for the same account information, and sends a customer service invitation to the second terminal of the account information with three request data items. After the second terminal accepts the invitation, it creates a third address as the access address of the second terminal, and provides digital human customer service to the second terminal through the third address.

[0013] Providing different services to different accounts can use computing resources on accounts with a greater likelihood of transaction.

[0014] Furthermore, when the target platform plays the customized digital human video to the second terminal through the second address, it receives the request data of the second terminal whose account category is a special category, creates a third address as the access address of the second terminal, and provides digital human customer service to the second terminal through the third address. When the digital human customer service first accesses the second terminal, it sends a reply to the request data to the second terminal and receives the request data input by the second terminal.

[0015] This approach can provide more accurate purchasing services to viewers with greater purchasing intentions.

[0016] Furthermore, the second terminal inputs or sends request data to the digital human customer service through a dialog box. When the second terminal adds text in the dialog box, each time the text in the dialog box changes, the second terminal generates a request data for all the text content in the dialog box and sends it to the target platform. The digital human customer service on the target platform processes the request data and generates a reply text and sends it to the second terminal. The second terminal generates a temporary reply box to display the received reply text word for word. The second terminal displays the most recently received reply text word for word in the temporary reply box, and clears the temporary reply box before displaying the most recently received reply text.

[0017] Further, when the second terminal sends request data or touches the temporary reply box to view the reply text, the dialog box is cleared, and the temporary reply box is displayed as a fixed reply box after the temporary reply box displays the reply text.

[0018] By monitoring the content input in the dialog box, the system can provide users with more relevant responses multiple times and quickly, which can not only enhance the interaction with the audience, but also allow the audience to learn more about the product during multiple interactions, thereby increasing the audience's desire to buy. At the same time, this zero-delay interaction of input and quick reply can not only increase the audience's interactive experience, but also reduce the time required for the audience to input complete content, thereby increasing the efficiency of the interaction.

[0019] Furthermore, the second terminal inputs voice in the dialog box through keyword triggering or key triggering, recognizes the input voice text as temporary dialogue text and displays it, after the temporary dialogue text changes, generates request data from the temporary dialogue text and sends it to the target platform, the digital human customer service on the target platform processes the request data and generates a reply text and sends it to the second terminal, the second terminal generates a temporary reply box and displays the received reply text word for word, the second terminal displays the most recently received reply text word for word in the temporary reply box, and clears the temporary reply box before displaying the most recently received reply text; When the second terminal pauses inputting voice, the temporary dialogue text is deleted by backspace triggered by keywords or keys; When the second terminal pauses inputting voice, it continues to input voice through keyword triggering or key triggering, recognizes the inputted voice text and appends it to the temporary dialogue text for display; When the second terminal finishes inputting voice, it confirms sending by triggering through keywords or pressing keys or touching the temporary reply box to view the reply text, clears the reply box, and the second terminal sends the complete voice and temporary conversation text to the target platform. After the temporary reply box displays the reply text, it displays the temporary reply box as a fixed reply box.

[0020] The display of temporary dialogue text and the reply text that appears word by word provide real-time feedback to users. Users can intuitively see that their voice is accurately recognized and converted into text, and can also immediately understand the response progress of the digital human customer service, which greatly enhances the user's sense of participation and immersion in the interaction with the system. Moreover, this real-time interactive effect prevents users from feeling anxious while waiting for a reply, because they can clearly see that the system is actively processing their requests. Users can also continue to input voice through keywords or key triggers, and append new content to the temporary dialogue text, which meets the user's needs to gradually improve their expression during the thinking process and avoids the trouble of having to start over due to incomplete input. After finishing the voice input, users can confirm the sending in a variety of ways to ensure the accuracy and completeness of the information. Before confirming the sending, users have the opportunity to check their input and the system's reply again, avoiding incorrect interactions caused by misoperation or misunderstanding. After displaying the reply text, the temporary reply box is transformed into a fixed reply box, providing users with a clear and stable result display area. Users can easily review the previous interaction content, which helps to better understand and remember the conversation process.

[0021] This voice input method can not only increase the interactive experience for the audience, but also reduce input operations and reduce the time for the audience to input text. At the same time, after pausing the voice, continuing to input additional voice is more in line with the scenario of thinking while speaking during actual use, which further increases the interactive experience and improves the interactive efficiency.

[0022] Furthermore, when the number of fixed reply boxes displayed on the second terminal reaches three, the digital human customer service calls the manual customer service to take over the customer service work.

[0023] When there are too many questions to answer, the purchase intention is extremely high, but the audience's patience is running out. Automatic transfer to manual customer service can better retain audiences with high purchase intention.

[0024] Furthermore, the digital human video is divided into several small segments according to the number of products included in the merchant's business. When the number of customized digital human videos generated when the target platform plays a small segment of the digital human video is greater than 1, after the small segment of the digital human video is played, a part of the customized digital human video that has not been played on the second terminal is selected and played to the second terminal through the second address, and the playback speed is adjusted so that all customized digital human videos played by the second terminal are played at the same time, the access address of the second terminal is changed to the first address, and the second address is deleted.

[0025] Since the digital human video may be played multiple times during live broadcast, when the number of viewers is small or the purchase intention is small, too many resources will be spent to avoid repeated playback and waste. When the number of inquiries is large, the audience's purchase intention is also greater. At this time, the process of the digital human replying to other viewers will be broadcast to the audience who have consulted. This can not only avoid excessive repetitions, but also answer more of the audience's questions and create a hot atmosphere with a large number of inquiries. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a structural diagram of a digital human interaction system based on an artificial intelligence large model in one embodiment of the present invention.

[0027] Figure 2 The figure is a flowchart of a digital human interaction method based on an artificial intelligence big model in one embodiment.

[0028] Figure 3 This is a flow chart related to the second terminal sending voice to the target platform in one embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the concept and technical effects of the present invention in combination with the embodiments, so as to fully understand the purpose, features and effects of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention: Figure 1 FIG. 1 is a system structure diagram of digital human interaction based on an artificial intelligence large model in an embodiment. Figure 1The digital human interaction method based on artificial intelligence big model is applied to the digital human interaction system based on artificial intelligence big model. The digital human interaction system based on artificial intelligence big model includes a second terminal 111, a first terminal 112 and a target platform 120.

[0030] The first terminal 112 is equipped with a first terminal unit, which is used to formulate and upload a live broadcast schedule and provide the business scope of the merchant. The target platform 120 is equipped with a target platform unit, which is used to create a first address according to the live broadcast schedule, generate a digital human video according to the business scope of the merchant, and play it in real time at the first address. The second terminal 111 is equipped with a second terminal unit, which is used to obtain the real-time digital human video after accessing the first address, and use the last obtained digital human video node as a pause node to send request data to the target platform.

[0031] The target platform unit is also used to generate a customized digital human video according to the content of the request data when receiving the request data sent by the second terminal unit, and create a second address as the access address of the second terminal, and play the customized digital human video to the second terminal through the second address; after the customized digital human video is played, the digital human video after the pause node is obtained and played at an accelerated speed until the playback node of the digital human video at the second address is the same as the first address, and the access address of the second terminal is changed to the first address.

[0032] In one embodiment, the first terminal 112, the second terminal 111 and the target platform 120 are used to perform the digital human interaction method based on the artificial intelligence large model. Among them, the second terminal 111 and the first terminal 112 can be desktop terminals or mobile terminals. The mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The second terminal 111 is a terminal used by the audience to watch the live broadcast, and the first terminal 112 is a terminal used by the merchant to connect with the target platform 120. The target platform 120 can be any existing live broadcast platform, or any number of independent live broadcast platforms. Among them, the first terminal 112 is also connected to the live broadcast team, which is used to formulate product promotion content based on the product information provided by the merchant, and input the product promotion content to the first terminal 112. The first terminal 112 outputs the video data of the digital human explaining the product according to the product promotion content, and transmits the video data to the target platform 120.

[0033] Figure 2 FIG. 1 is a flow chart of a digital human interaction method based on an artificial intelligence large model in one embodiment. Figure 2 The digital human interaction method based on the artificial intelligence large model specifically includes the following steps: S10: The target platform creates a first address according to the live broadcast schedule uploaded by the first terminal, generates a digital human video according to the business scope of the merchant and plays it in real time on the first address, and the second terminal obtains the real-time played digital human video after accessing the first address.

[0034] Before the live broadcast begins, the live broadcast team or merchant uploads the live broadcast schedule and the merchant's business scope to the target platform through the first terminal. The target platform creates a first address according to the live broadcast schedule, that is, opens a first live broadcast room and records the network address of the live broadcast room. The live broadcast platform generates a digital human video according to the merchant's business scope, and plays the digital human video in real time in the first live broadcast room through the first address. The second terminal enters the first live broadcast room through the first address on the target platform, obtains the real-time digital human video through the first address, and plays it on the second terminal after obtaining it.

[0035] When the live broadcast team produces product promotion content, they produce it in segments according to the number of products included in the merchant's business; the target platform produces the digital human video into several segments based on several small segments of product promotion content. In one embodiment of this solution, the merchant produces three small segments of product promotion content for three commodities to be sold and uploads them to the live broadcast platform. The live broadcast platform produces the three small segments of product promotion content into three digital human videos, sorts the three digital human videos, and plays the three digital human videos in the first live broadcast room through the first address.

[0036] S20: The second terminal sends request data to the target platform through the first address. The target platform generates a customized digital human video according to the content of the request data, creates a second address as the access address of the second terminal, plays the customized digital human video to the second terminal through the second address, and uses the last digital human video node obtained by the second terminal as the pause node.

[0037] After the second terminal inputs the text data, it generates request data from the text data, and the request data includes the user identification and the request content. The user identification includes the account information and the account category, and the account category is divided into a special category and a general category. The request data is transmitted to the target platform through the first address. In one embodiment of this scheme, the audience inputs a text message in the first live broadcast room playing the digital human video through the mobile phone terminal, and the mobile phone terminal generates request data from the text message, and the request data includes the audience's account information, account category and the specific content of the text message, and the request data is transmitted to the live broadcast platform through the first address.

[0038] Among them, the target platform reviews the received request data, reviews the request data related to product consultation as intended request data, reviews the request data related to atmosphere as atmosphere request data, reviews the request data that does not meet the platform requirements as illegal request data, and reviews other request data as ignored request data. The atmosphere request data is accelerated and superimposed on the digital human video, a violation prompt is given to the second terminal that issued the illegal request data, the ignored request data is deleted, the intention request data is subjected to natural language semantic extraction and analysis, and the second terminal with the same semantic request content is connected to the same second address.

[0039] In one embodiment of the present scheme, the request data received related to the sale of goods are reviewed as intention request data, and the intention request data is subjected to natural language semantic extraction analysis to extract the core of the question (i.e., what question is being asked and what kind of answer is needed), and viewers who ask the same question (with the same answer content) are connected to the same second address.

[0040] When the target platform records the account type in the received request data as special account information, it accumulates the number of request data for the same account information and sends a customer service invitation to the second terminal of the account information with three request data. After the second terminal accepts the invitation, it creates a third address as the access address of the second terminal and provides the digital human customer service to the second terminal through the third address.

[0041] In one embodiment of the present scheme, if the live broadcast platform receives request data (excluding illegal request data) sent by a special category of viewers in the first live broadcast room for three times, it will send a customer service invitation to the viewer (the special category of viewers) and create a third address, and provide customer service to the viewer through the third address.

[0042] S30: After the customized digital human video is played, the digital human video after the pause node is obtained and played at an accelerated speed until the playback node of the digital human video at the second address is the same as the first address, and the access address of the second terminal is changed to the first address.

[0043] Among them, when the number of customized digital human videos generated when the target platform plays a short digital human video is greater than 1, after the short digital human video is played, a part of the customized digital human videos that have not been played on the second terminal is selected and played to the second terminal through the second address, and the playback speed is adjusted so that all customized digital human videos played by the second terminal are played at the same time, the access address of the second terminal is changed to the first address, and the second address is deleted.

[0044] Specifically, when the target platform plays a short segment of digital human video and the number of customized digital human videos generated is not greater than 1, the target platform plays the next segment of digital human video after playing the short segment of digital human video. After the customized digital human video is played, if the playback node of the digital human video at the second address is still different from the first address after the acceleration playback of the short segment of digital human video after the pause node, the second address will play the next segment of digital human video until the playback node of the digital human video at the second address is the same as the first address.

[0045] In one embodiment, when the live broadcast platform is playing the digital human video of product A, it receives an inquiry from viewer A, opens a second live broadcast room for viewer A, generates a customized digital human video to answer viewer A's inquiry, and plays the customized digital human video to the viewer in the second live broadcast room. Before the live broadcast platform transfers viewer A from the first live broadcast room to the second live broadcast room, the viewer has watched 95% of the digital human video of product A. When viewer A is in the second live broadcast room, the first live broadcast room has played to product B. After playing the customized digital human video, the second live broadcast room accelerates the playback of the remaining 5% of the digital human video of product A, and then accelerates the playback of the digital human video of product B until the playback nodes of the first live broadcast room and the second live broadcast room are the same, then the viewer A is transferred back to the first live broadcast room, and the second live broadcast room is closed.

[0046] In another embodiment, when the live broadcast platform plays the digital human video of product A, it receives consultation 1 from viewer A, consultation 2 from viewer B, and consultation 3 from viewer C. The live broadcast platform starts a second live broadcast room 1 for viewer A, a second live broadcast room 2 for viewer B, and a second live broadcast room 3 for viewer C, and generates customized digital human video 1, customized digital human video 2, and customized digital human video 3 for consultation 1, consultation 2, and consultation 3, respectively. After playing customized digital human video 1, the first live broadcast room accelerates the playback of a short segment of digital human video after the pause node, and then plays customized digital human video 2 (do not select customized digital human video 1, which has been played in the second live broadcast room 1). After playing customized digital human video 2, the second live broadcast room accelerates the playback of a short segment of digital human video after the pause node, and then plays customized digital human video 3. After playing customized digital human video 3, the third live broadcast room accelerates the playback of a short segment of digital human video after the pause node, and then plays customized digital human video 1. Since the accelerated content played in the three second live broadcast rooms has different durations, the acceleration is also adjusted differently. At this time, the three live broadcast rooms finish playing their respective content at the same time, and viewers A, B and C are transferred back to the first live broadcast room, and the three second live broadcast rooms are closed.

[0047] When the target platform plays the customized digital human video to the second terminal through the second address, it receives the request data of the second terminal whose account category is a special category, creates a third address as the access address of the second terminal, and provides digital human customer service to the second terminal through the third address. When the digital human customer service first accesses the second terminal, it sends a reply to the request data to the second terminal and receives the request data input by the second terminal.

[0048] In one embodiment of the present solution, when a special category of viewers sends a text inquiry to the live broadcast platform in the second live broadcast room, the digital human customer service is accessed for the viewer, and when accessed, the digital human customer service responds to the inquiry raised by the viewer, and the inquiry is displayed in a fixed reply box.

[0049] Among them, the second terminal inputs or sends request data to the digital human customer service through a dialog box. When the second terminal adds text in the dialog box, every time the text in the dialog box changes, the second terminal generates a request data with all the text content in the dialog box and sends it to the target platform. The digital human customer service on the target platform processes the request data and generates a reply text and sends it to the second terminal. The second terminal generates a temporary reply box to display the received reply text word by word. The second terminal displays the most recently received reply text word by word in the temporary reply box, and clears the temporary reply box before displaying the most recently received reply text. When the second terminal sends request data or touches the temporary reply box to view the reply text, the dialog box is cleared, and the temporary reply box is displayed as a fixed reply box after the temporary reply box displays the reply text.

[0050] In one embodiment of the present scheme, when a special category of viewers inputs the text content they want to consult in the dialog box (via a mobile phone), each time a text is added to the text box, the second terminal obtains all the information in the dialog box, generates request data, and sends it to the live broadcast platform. After receiving the request data, the live broadcast platform generates a reply text and sends the reply text to the second terminal. After receiving the reply text, the second terminal displays the reply text word by word. After viewing the reply text, the viewer finds that it is not what he wants, so he continues to input text into the dialog box. After the dialog box receives the text, the mobile phone generates request data again and sends it to the live broadcast platform. The live broadcast platform makes a reply again. When the second terminal displays the reply word by word, the viewer enters text again. When the second terminal displays the text word by word, it receives the reply, so it clears the temporary reply box (that is, clears the reply that is being displayed word by word), and displays the latest reply word by word. The user confirms the sending, or slides to view the second terminal, or clicks the second terminal to view the temporary text box and then clears the dialog box. After the temporary reply box displays the reply text, the temporary reply box is displayed as a fixed reply box.

[0051] Among them, Figure 3As shown, the second terminal inputs voice in the dialog box through keyword triggering or key triggering, recognizes the input voice text as temporary dialogue text and displays it, and after the temporary dialogue text changes, generates request data from the temporary dialogue text and sends it to the target platform. The digital human customer service on the target platform processes the request data and generates a reply text and sends it to the second terminal. The second terminal generates a temporary reply box to display the received reply text word by word. The second terminal displays the most recently received reply text word by word in the temporary reply box, and clears the temporary reply box before displaying the most recently received reply text. When the second terminal pauses inputting voice, the temporary dialogue text is deleted by backspace triggered by keywords or keys; When the second terminal pauses inputting voice, it continues to input voice through keyword triggering or key triggering, recognizes the inputted voice text and appends it to the temporary dialogue text for display; When the second terminal finishes inputting voice, it confirms sending by triggering through keywords or pressing keys or touching the temporary reply box to view the reply text, clears the reply box, and the second terminal sends the complete voice and temporary conversation text to the target platform. After the temporary reply box displays the reply text, it displays the temporary reply box as a fixed reply box.

[0052] In one embodiment of this solution, the audience triggers voice input by pressing a button. When the audience speaks, the mobile phone recognizes and displays the text of what the audience has said. When the audience finds that the recognition is wrong, they pause speaking, press the backspace key to delete the wrong part, and continue to input voice. The mobile phone will append the text recognition content of the voice input to the previous content (the response made by the digital human customer service of the live broadcast platform when the audience inputs text is the same as the response made by the audience when the audience types text, which will not be repeated here). When the audience triggers the reply text by a keyword or a button or touches the temporary reply box to view the reply text, confirm the sending and clear the reply box.

[0053] In another embodiment, when voice is converted to text, if backspace is pressed multiple times at the same position and voice is input again, the digital human customer service calls the manual audience to take over the customer service work and reminds the manual audience to provide service in text mode first.

[0054] Among them, when the number of fixed reply boxes displayed on the second terminal reaches three, the digital human customer service calls the manual audience to take over the customer service work.

[0055] In one embodiment of the present solution, after the digital human customer service answers the audience's questions three times, the digital human customer service calls the manual customer service to take over the customer service work.

[0056] The above is only an embodiment of the present invention, and the common knowledge such as the known specific structure and characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several deformations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A digital human interaction method based on an artificial intelligence large model, characterized in that: The following steps are involved: The target platform creates a first address according to the live broadcast schedule uploaded by the first terminal, generates a digital human video according to the business scope of the merchant and plays it in real time on the first address, and the second terminal obtains the real-time played digital human video after accessing the first address; The second terminal sends request data to the target platform through the first address. The target platform generates a customized digital human video according to the content of the request data, creates a second address as the access address of the second terminal, plays the customized digital human video to the second terminal through the second address, and uses the last digital human video node obtained by the second terminal as the pause node; After the customized digital human video is played, the digital human video after the pause node is obtained and played at an accelerated speed until the playback node of the digital human video at the second address is the same as the first address, and the access address of the second terminal is changed to the first address.

2. The digital human interaction method based on artificial intelligence large model according to claim 1 is characterized by: The request data includes a user identifier and request content. The user identifier includes account information and account categories. The account categories are divided into special categories and general categories.

3. The digital human interaction method based on artificial intelligence large model according to claim 2 is characterized by: The target platform performs semantic extraction and analysis on the request content in the received request data, and records the account information of the special category, and connects the second terminal with the same semantic request content to the same second address; accumulates the number of request data for the same account information, and sends a customer service invitation to the second terminal with three request data. After the second terminal accepts the invitation, it creates a third address as the access address of the second terminal, and provides digital human customer service to the second terminal through the third address.

4. The digital human interaction method based on artificial intelligence large model according to claim 3 is characterized by: When the target platform plays the customized digital human video to the second terminal through the second address, it receives the request data of the second terminal whose account category is a special category, creates a third address as the access address of the second terminal, and provides digital human customer service to the second terminal through the third address. When the digital human customer service first accesses the second terminal, it sends a reply to the request data to the second terminal and receives the request data input by the second terminal.

5. The digital human interaction method based on artificial intelligence large model according to claim 4 is characterized by: The second terminal inputs or sends request data to the digital human customer service through a dialog box. When the second terminal adds text in the dialog box, each time the text in the dialog box changes, the second terminal generates a request data with all the text content in the dialog box and sends it to the target platform. The digital human customer service on the target platform processes the request data and generates a reply text and sends it to the second terminal. The second terminal generates a temporary reply box to display the received reply text word for word. The second terminal displays the most recently received reply text word for word in the temporary reply box, and clears the temporary reply box before displaying the most recently received reply text.

6. The digital human interaction method based on artificial intelligence large model according to claim 5 is characterized by: When the second terminal sends request data or touches the temporary reply box to view the reply text, the dialog box is cleared, and the temporary reply box is displayed as a fixed reply box after the temporary reply box displays the reply text.

7. The digital human interaction system and method based on artificial intelligence big model according to claim 6 is characterized by: The second terminal inputs voice in the dialog box through keyword triggering or key triggering, recognizes the input voice text as temporary dialogue text and displays it, and after the temporary dialogue text changes, generates request data from the temporary dialogue text and sends it to the target platform. The digital human customer service on the target platform processes the request data and generates a reply text and sends it to the second terminal. The second terminal generates a temporary reply box to display the received reply text word by word. The second terminal displays the most recently received reply text word by word in the temporary reply box, and clears the temporary reply box before displaying the most recently received reply text; When the second terminal pauses inputting voice, the temporary dialogue text is deleted by backspace triggered by keywords or keys; When the second terminal pauses inputting voice, it continues to input voice through keyword triggering or key triggering, recognizes the inputted voice text and appends it to the temporary dialogue text for display; When the second terminal finishes inputting voice, it confirms sending by triggering through keywords or pressing keys or touching the temporary reply box to view the reply text, clears the reply box, and the second terminal sends the complete voice and temporary conversation text to the target platform. After the temporary reply box displays the reply text, it displays the temporary reply box as a fixed reply box.

8. The digital human interaction method based on artificial intelligence large model according to claim 7 is characterized by: When the number of fixed reply boxes displayed on the second terminal reaches three, the digital customer service calls the manual customer service to take over the customer service work.

9. The digital human interaction system and method based on artificial intelligence big model according to claim 8 is characterized by: The digital human video is divided into several small segments according to the number of products included in the merchant's business. When the number of customized digital human videos generated when the target platform plays a small segment of the digital human video is greater than 1, after the small segment of the digital human video is played, a part of the customized digital human video that has not been played on the second terminal is selected through the second address and played to the second terminal, and the playback speed is adjusted so that all customized digital human videos played by the second terminal are played at the same time, the access address of the second terminal is changed to the first address, and the second address is deleted.

10. A digital human interaction system based on an artificial intelligence large model, characterized in that: include: The first terminal unit: used to formulate and upload the live broadcast schedule and provide the business scope of the merchant; Target platform unit: used to create a first address according to the live broadcast schedule, generate a digital human video according to the business scope of the merchant, and play it in real time on the first address; The second terminal unit is used to access the first address and obtain the real-time digital human video, and use the last obtained digital human video node as a pause node to send request data to the target platform; The target platform unit is also used to generate a customized digital human video according to the content of the request data when receiving the request data sent by the second terminal unit, and to create a second address as the access address of the second terminal, and to play the customized digital human video to the second terminal through the second address; After the customized digital human video is played, the digital human video after the pause node is obtained and played at an accelerated speed until the playback node of the digital human video at the second address is the same as the first address, and the access address of the second terminal is changed to the first address.

Images