Human-computer interaction processing method, device, equipment, product and system
By using a large language model to process interactive task information and user input in the human-computer interaction system, and generating task instructions, the problem of complex process design in traditional systems is solved, and the system flexibility and user experience are improved.
Patent Information
- Application Number
- CN202410838459.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-11-26
AI Technical Summary
When achieving specific interaction purposes, traditional human-computer interaction systems rely on complex flowchart designs and a large number of process nodes, resulting in complex configurations and difficulty in flexibly responding to various user inputs, which makes the user experience poor.
By obtaining interactive task information, generating task instructions and triggering the large language model to perform interactive tasks, multiple backbone processes are used to achieve global tracking and control of human-computer interaction status, reducing configuration requirements for process nodes, and increasing the naturalness and personality of dialogue change.
It realizes the flexible response capabilities of the human-computer interaction system, improves the conversion rate of user experience and marketing activities, and reduces the complexity of system configuration.
Smart Images

Figure CN119026685A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and specifically to a human-computer interaction processing method, device, equipment, product and system. Background Art
[0002] With the continuous advancement of technology, human-computer interaction technology has been widely used in many fields, especially in scenarios that require automated and intelligent interaction with users, due to its intelligent and automated characteristics. Taking the intelligent outbound call system as an example, the system can automatically perform outbound call tasks according to business scenarios, conduct product marketing through voice dialogue interaction between people and the system, collect business results, and perform statistical processing on data, which significantly improves processing efficiency and reduces labor costs.
[0003] However, in traditional human-computer interaction systems, in order to achieve specific interaction purposes (such as product promotion, etc.), the system often relies on complex flowchart design and a large number of process nodes to cover all possible user reactions and dialogue paths. This is not only complex to configure, but also difficult to flexibly respond to various user inputs, resulting in a rigid human-computer interaction process and poor user experience. Summary of the invention
[0004] In view of this, the embodiments of the present application are committed to providing a human-computer interaction processing method, device, equipment, product and system, which can adjust the system output information according to the preset main process and real-time human-computer interaction information, without relying on complex flowchart design, reducing the configuration requirements of the human-computer interaction system for each process node, and increasing the naturalness and personalization of the conversation.
[0005] According to a first aspect of an embodiment of the present application, a human-computer interaction processing method is provided, including:
[0006] Acquire interaction task information, wherein the interaction task information includes interaction purpose, user information, product information, and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of the target user;
[0007] Generate a first task instruction according to the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user according to the interactive task information to promote the product corresponding to the product information to the target user;
[0008] The first task instruction is input into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
[0009] Optionally, interacting with the target user according to the interactive task information includes:
[0010] Determine the current main process according to the interaction process and the human-computer interaction information, wherein the human-computer interaction information includes historical interaction content and current user input;
[0011] For the purpose of completing the current main process, first output information is generated according to the interactive task information and the human-computer interaction information, wherein the first output information includes reply information to the current user input and / or guidance information to the target user.
[0012] Optionally, the generating of first output information according to the interactive task information and the human-computer interaction information for the purpose of completing the current main process includes:
[0013] Determine a marketing strategy based on the human-computer interaction information;
[0014] For the purpose of completing the current main process, the first output information is generated according to the marketing strategy, the interactive task information and the human-computer interaction information.
[0015] Optionally, before determining a marketing strategy based on the human-computer interaction information, the method further includes:
[0016] Determine, according to the interactive task information and the human-computer interaction information, a first instruction corresponding to the current user input, wherein the first instruction is used to instruct the object receiving the first instruction to perform a first process corresponding to the current user input according to the first instruction, and return an execution result corresponding to the first instruction, wherein the first process includes at least one of memory information retrieval, knowledge query, and business handling;
[0017] Obtaining an execution result corresponding to the first instruction;
[0018] Determining a marketing strategy according to the human-computer interaction information includes:
[0019] Determining a marketing strategy according to the execution result corresponding to the first instruction and the human-computer interaction information;
[0020] The step of generating the first output information for the purpose of completing the current main process according to the marketing strategy, the interactive task information and the human-computer interaction information includes:
[0021] With the purpose of completing the current main process, the first output information is generated according to the execution result, the marketing strategy, the interactive task information and the human-computer interaction information.
[0022] Optionally, the method further includes:
[0023] Performing marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generating marketing anomaly information when a marketing anomaly event is identified during the interaction process, wherein the marketing anomaly information includes a marketing anomaly handling strategy corresponding to the marketing anomaly event;
[0024] The marketing abnormality information is input into the first large language model, so that the first large language model generates second output information according to the marketing abnormality information.
[0025] Optionally, the interaction process between the first language model and the target user is monitored for marketing anomalies according to a preset marketing anomaly identification strategy, and when a marketing anomaly event is identified in the interaction process, marketing anomaly information is generated, including:
[0026] The pre-trained second language model is triggered to perform marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generates marketing anomaly information when a marketing anomaly event is identified from the interaction process.
[0027] Optionally, the marketing anomaly identification strategy includes:
[0028] When the cumulative number of occurrences of the same abnormal scenario reaches a preset number threshold corresponding to the abnormal scenario, it is determined that a marketing abnormal event corresponding to the abnormal scenario has occurred;
[0029] The step of monitoring the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy and generating marketing anomaly information when a marketing anomaly event is identified during the interaction process includes:
[0030] Matching the latest round of human-computer interaction information with preset abnormal scenario description information, and generating second execution information when the latest round of human-computer interaction information matches the description information corresponding to the first abnormal scenario, wherein the abnormal scenario description information includes description information corresponding to each abnormal scenario type, and the second execution information is used to instruct the object receiving the second execution information to add one to the cumulative number of occurrences of the first abnormal scenario, and return the updated cumulative number of occurrences of the first abnormal scenario;
[0031] Obtaining the updated cumulative number of occurrences of the first abnormal scenario;
[0032] Based on the updated cumulative number of occurrences of the first abnormal scenario and the preset number threshold corresponding to the first scenario, determine whether a marketing abnormal event corresponding to the first abnormal scenario has occurred, and when it is determined that a marketing abnormal event corresponding to the first abnormal scenario has occurred, generate marketing abnormality information, wherein the marketing abnormality information includes description information of the marketing abnormal event corresponding to the first abnormal scenario and a marketing abnormality handling strategy corresponding to the first abnormal scenario.
[0033] According to a second aspect of an embodiment of the present application, a human-computer interaction processing device is provided, including:
[0034] A first unit is used to obtain interaction task information, wherein the interaction task information includes interaction purpose, user information, product information and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of a target user;
[0035] A second unit is configured to generate a first task instruction according to the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user according to the interactive task information to promote the product corresponding to the product information to the target user;
[0036] The third unit is used to input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
[0037] According to a third aspect of an embodiment of the present application, there is provided an electronic device, including a memory and a processor;
[0038] The memory is connected to the processor and is used to store programs;
[0039] The processor is used to implement the human-computer interaction processing method as described in any one of the first aspects of the embodiments of the present application by running the program in the memory.
[0040] According to a fourth aspect of an embodiment of the present application, a computer program product is provided, comprising computer program instructions, which, when executed by a processor, enable the processor to implement a human-computer interaction processing method as described in any one of the first aspects of the embodiments of the present application.
[0041] According to a fifth aspect of the embodiments of the present application, there is provided a human-computer interaction system, comprising: an input-output module and an interaction module, wherein the input-output module is communicatively connected with the interaction module;
[0042] The input-output module is used to receive input information from the user and output output information generated by the interaction module;
[0043] The interaction module is used to obtain interaction task information, wherein the interaction task information includes interaction purpose, user information, product information and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of the target user; generate a first task instruction according to the interaction task information, wherein the first task instruction is used to trigger the large language model to perform the interaction task, and the interaction task is used to interact with the target user according to the interaction task information to promote the product corresponding to the product information to the target user; input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
[0044] Optionally, the system further includes:
[0045] An anomaly monitoring module is used to monitor the interaction process between the first large language model and the target user for marketing anomalies according to a preset marketing anomaly identification strategy, and generate marketing anomaly information when a marketing anomaly event is identified in the interaction process, and input the marketing anomaly information into the first large language model so that the first large language model generates second output information according to the marketing anomaly information, and the marketing anomaly information includes a marketing anomaly handling strategy corresponding to the marketing anomaly event.
[0046] The human-computer interaction processing method provided in the embodiment of the present application can be applied to a human-computer interaction system. The method automatically generates a first task instruction according to the interaction task information including multiple main processes, and the first task instruction automatically triggers the first large language model to execute the interaction task. Human-computer interaction is performed with the target user according to the interaction task information. The interaction process including multiple main processes is used to achieve global tracking and control of the human-computer interaction state, rather than relying on complex flowchart design, thereby reducing the configuration requirements of the human-computer interaction system for each process node, so that the human-computer interaction system can respond to user intentions more flexibly instead of strictly following preset process nodes.
[0047] On the other hand, relying on the excellent natural language processing capabilities of the large language model, applying the present application scheme to the human-computer interaction system can comprehensively and accurately understand the interactive task information and user input information, and adjust the system output information according to the real-time human-computer interaction information, rather than relying on preset response scripts, thereby increasing the naturalness and personalization of the conversation, improving the user experience, and thereby improving the conversion rate of marketing activities and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0049] Figure 1 A schematic diagram of the structure of a human-computer interaction system provided in an embodiment of the present application.
[0050] Figure 2 A schematic diagram of the structure of another human-computer interaction system provided in an embodiment of the present application.
[0051] Figure 3 A flowchart of a human-computer interaction processing method provided in an embodiment of the present application.
[0052] Figure 4 A schematic diagram of a processing flow for interacting with a target user based on interactive task information provided in an embodiment of the present application.
[0053] Figure 5 A flowchart of another human-computer interaction processing method provided in an embodiment of the present application.
[0054] Figure 6 A schematic diagram of the structure of a human-computer interaction processing device provided in an embodiment of the present application.
[0055] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0056] The technical solution of the embodiment of the present application is applicable to various scenarios that require human-computer interaction, such as intelligent product marketing scenarios, intelligent outbound call scenarios, intelligent customer service scenarios, etc. Taking the intelligent outbound call scenario as an example, the intelligent outbound call system in this scenario can automatically perform outbound call tasks according to the business scenario, and conduct product marketing through voice dialogue interaction between people and the system, which significantly improves processing efficiency and reduces labor costs.
[0057] The technical solution provided in the embodiments of the present application can be exemplarily applied to hardware devices such as processors, electronic devices, servers (including cloud servers), or packaged into software programs to be run. When the hardware device executes the processing of the technical solution of the embodiments of the present application, or the above-mentioned software program is run, the automatic splitting of the target task and the automatic calling of the application program interface required for the task can be achieved to complete the purpose of the target task. The embodiments of the present application only exemplarily introduce the specific processing of the technical solution of the present application, and do not limit the specific implementation form of the technical solution of the present application. Any technical implementation form that can execute the processing of the technical solution of the present application can be adopted by the embodiments of the present application.
[0058] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0059] Before introducing this application solution, the relevant technologies are first introduced:
[0060] With the continuous advancement of technology, human-computer interaction technology has been widely used in many fields, especially in scenarios that require automated and intelligent interaction with users, due to its intelligent and automated characteristics. Taking the intelligent outbound call system as an example, the system can automatically perform outbound call tasks according to business scenarios, conduct product marketing through voice dialogue interaction between people and the system, collect business results, and perform statistical processing on data, which significantly improves processing efficiency and reduces labor costs.
[0061] However, in traditional human-computer interaction systems, in order to achieve specific interaction purposes (such as product promotion, etc.), the system often relies on complex flowchart design and a large number of process nodes to cover all possible user reactions and dialogue paths. This is not only complex to configure, but also difficult to flexibly respond to various user inputs, resulting in a rigid human-computer interaction process and poor user experience.
[0062] In view of this, the embodiments of the present application are committed to providing a human-computer interaction processing method, device, equipment, product and system, which can adjust the system output information according to the preset main process and real-time human-computer interaction information, without relying on complex flowchart design, reducing the configuration requirements of the human-computer interaction system for complex process nodes, and increasing the naturalness and personalization of the conversation, which will be described in detail one by one in the following embodiments.
[0063] Exemplary Systems
[0064] For ease of understanding, firstly, an exemplary introduction to the implementation environment of the human-computer interaction processing method provided in the embodiment of the present application is given. Figure 1 , Figure 1 Schematic diagram of the structure of a human-computer interaction system. The human-computer interaction processing method provided in the present application can be exemplarily applied to the human-computer interaction system.
[0065] like Figure 1 As shown, the human-computer interaction system includes an input-output module 110 and an interaction module 120, and the input-output module is communicatively connected with the interaction module.
[0066] The input-output module 110 is the front-end part of the human-computer interaction system, which is used to receive input information from the user and output information generated by the output interaction module 120, and is a bridge for human-computer interaction between the user and the system.
[0067] Optionally, the input-output module 110 includes an input component and an output component, and both the input component and the output component are communicatively connected to the interaction module 120. The input component is used to receive the user's input information and send the received user's input information to the interaction module 120 in real time. Exemplarily, the input component can be a microphone or a pickup (for receiving the user's voice input), a touch screen or a keyboard (for receiving the user's text input or selection operation), a camera (for capturing the user's gestures or facial expressions), etc. The output component is used to output the received output information after receiving the output information generated and sent by the interaction module 120. Exemplarily, the output component can be a speaker (for outputting voice information to the user), a display or a projector (for outputting text, image or video information to the user), etc.
[0068] Exemplarily, the human-computer interaction system is a marketing outbound call system. In the system, the input component includes a voice input component and a keyboard input component. The voice recognition component can accurately capture the user's voice signal, and the keyboard input component supports the user to input information or select options by keystroke. The output component includes a voice output component for playing voice information to the user.
[0069] The interaction module 120 is used to execute the human-computer interaction processing method provided by the embodiment of the present application. Specifically, the interaction task information is first obtained, wherein the interaction task information includes the interaction purpose, user information, product information and interaction process, the interaction process includes multiple trunk processes for achieving the interaction purpose, and the user information includes the identification information of the target user; then, according to the interaction task information, a first task instruction is generated, wherein the first task instruction is used to trigger the large language model to perform the interaction task, and the interaction task is used to interact with the target user according to the interaction task information to promote the product corresponding to the product information to the target user; finally, the first task instruction is input into the first large language model trained in advance, so that the first large language model promotes the product to the target user by executing the first task instruction. As for other specific implementations of the human-computer interaction processing method, they will be specifically expanded in the subsequent content and will not be described in detail here.
[0070] As an optional implementation, the human-computer interaction system is a split system, including a server and at least one client device, and each client device is communicatively connected to the server.
[0071] Each client device is provided with an input / output module 110, which is used to receive the user's input information and send it to the server, and receive the output information returned by the server and display it to the user. Optionally, the client device can be various types of computer devices with input / output capabilities, such as portable handheld devices, general-purpose computers, workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, various message sending and receiving devices, etc.
[0072] The server is integrated or deployed with an interaction module 120 for executing the human-computer interaction processing method provided in the embodiment of the present application. Optionally, the server can also store and manage data such as user information, product information, interaction process, etc., and provide computing resources to support the operation of a large language model.
[0073] The server may be an independent server, a server group consisting of multiple physical servers, or a cloud server providing cloud computing services.
[0074] As an optional implementation, the human-computer interaction system is an integrated system running on a single device, integrating the input-output module 110 and the interaction module 120 in the same device, and the input-output module 110 and the interaction module 120 are directly connected through the hardware and software interfaces inside the device. Exemplarily, the single device can be an intelligent robot, and the human-computer interaction system is a local system running on the intelligent robot, which is equipped with a high-performance processor and a large-capacity storage device to support the operation and data processing of the interaction module 120. In addition, the intelligent robot is integrated with the input-output module 110 to realize direct interaction with the user. For example, the intelligent robot can receive the user's voice input through the built-in microphone, capture the user's gestures or expressions through the camera, and output the output information generated by the interaction module 120 to the user through a speaker or a display screen.
[0075] Figure 2 is a schematic diagram of another human-computer interaction system. As an optional implementation method, Figure 2 As shown, the human-computer interaction system provided in this embodiment further includes an abnormality monitoring module 130 , and the abnormality monitoring module 130 is communicatively connected with the input-output module 110 and the interaction module 120 , respectively.
[0076] The abnormality monitoring module is used to monitor the interaction process between the first language model and the target user for marketing abnormalities according to a preset marketing abnormality identification strategy, and to generate marketing abnormality information when a marketing abnormality event is identified in the interaction process, and to input the marketing abnormality information into the first language model so that the first language model generates second output information according to the marketing abnormality information, and the marketing abnormality information includes the marketing abnormality handling strategy corresponding to the marketing abnormality event. As for the specific implementation of the above functions of the abnormality monitoring module, it will be introduced in subsequent embodiments and will not be described in detail here.
[0077] Exemplary Methods
[0078] Figure 3 This is a flow chart of a human-computer interaction processing method provided in an embodiment of the present application. Figure 3 As shown, the human-computer interaction processing method provided in this embodiment includes steps S301-S303:
[0079] S301: Obtain interactive task information.
[0080] The interactive task information can be understood as relevant information of the human-computer interaction task to be executed, including the basic data and background information required for the human-computer interaction system to execute the human-computer interaction task, which is used to guide the human-computer interaction system to execute the corresponding human-computer interaction task, so that the human-computer interaction system can carry out human-computer dialogue based on a full understanding of the context corresponding to the human-computer interaction task, and determine the pertinence and effectiveness of the human-computer interaction.
[0081] The interactive task information includes the interactive purpose, user information, product information and interactive process.
[0082] The interaction purpose can be understood as the goal or intention to be achieved by performing the human-computer interaction task, which is used to guide the behavior and dialogue content of the human-computer interaction system and ensure that the interaction process meets the preset goals. For example, in the mobile phone package marketing outbound call scenario, the interaction purpose can be: you are a marketing service assistant of the operator company, introduce the company's products to them, and answer their questions that they may have to attract users' purchasing interest.
[0083] The user information can be understood as a data set that describes the characteristics of the target user, including at least the identification information of the target user, which is used to guide the human-computer interaction system to uniquely determine and identify the target user. The identification information of the target user can be a user name, mobile phone number, email address, etc. Exemplarily, in the aforementioned mobile phone package marketing outbound call scenario, the identification information of the target user can be the mobile phone number of the target user, and the human-computer interaction system performs an outbound call operation based on the mobile phone number of the target user to establish a call connection with the target user.
[0084] Optionally, the user information may also include the user's current product information. The user's current product information can be understood as relevant information about the product or service currently used by the target user, including one or more of the attributes such as product name, type, function, and cost. The user's current product information can help the human-computer interaction system understand the target user's existing consumption status. By mastering the user's current product information, the human-computer interaction system can formulate a more personalized marketing strategy. For example, in the aforementioned mobile phone package marketing outbound call scenario, the currently used product information is "4G_59 yuan package, including 200 minutes of free calls and 10G domestic traffic".
[0085] Optionally, the user information may also include user portrait information. The user portrait information can be understood as a comprehensive description of the target user, which may cover one or more of the user's personal characteristics, consumption habits, hobbies, social behaviors and other aspects of data. User portrait information can help the human-computer interaction system understand the target user more accurately, thereby providing more personalized services to the target user. Exemplarily, in the aforementioned mobile phone package marketing outbound call scenario, the user portrait information may be "usually surfing the Internet more, using more data traffic, and exceeding the package usage every month."
[0086] The product information can be understood as relevant information of the product to be promoted, and may include a detailed description and / or feature description of the product to be promoted, such as the product's name, type, function, pricing, and differentiation from competing products. The product information enables the human-computer interaction system to clearly understand the relevant circumstances of the product to be promoted, so as to accurately introduce the product to the target user during the interaction process, answer user questions, and effectively promote the product. It should be noted that the product to be promoted is a plurality of different products. For example, in the aforementioned mobile phone package marketing outbound call scenario, the product information is "4G_99 yuan package, including 500 minutes of free calls and 50G of domestic traffic".
[0087] The interaction process can be understood as a series of main links designed to achieve the purpose of the interaction. These links are interrelated and together constitute the main framework and direction of the entire human-computer interaction task. The interaction process specifies the basic path and behavior norms that should be followed in the human-computer interaction process. The interaction process includes multiple main processes, each of which carries specific interaction tasks and goals. They are carried out in a specific order and logic, aiming to guide users to complete specific operations or provide necessary information to ensure that the human-computer interaction process can proceed in an orderly manner. For example, in the product marketing scenario, the main process may include identity confirmation, product marketing, product handling, etc. These links constitute the backbone of the entire marketing interaction process, and each step is intended to push users closer to the final product handling. Exemplarily, in the aforementioned mobile phone package marketing outbound call scenario, the interaction process is "1. Confirm the last number information of the user's mobile phone owner; 2. Conduct product marketing to the user and verbally confirm the intention to handle; 3. Press the button to confirm the intention to handle with the user again, and press the No. 1 button for the second confirmation before handling."
[0088] As an optional implementation manner, the interaction task information may also include external tool information, third-party service integration information, historical interaction records, and other information related to the human-computer interaction task to be executed.
[0089] Optionally, obtaining the interactive task information includes: obtaining the interactive task information in response to an input operation on the interactive task information, that is, directly inputting the interactive task information into the human-computer interaction system.
[0090] Optionally, acquiring the interactive task information includes: in response to an input operation on the first information, acquiring the first information, analyzing the first information, and determining the interactive task information according to the analysis result. The first information includes content representing the interactive task information.
[0091] Optionally, obtaining the interactive task information includes: calling the interactive task information by accessing a data source storing the interactive task information, wherein the data source may be a database, a cloud storage service, a local file system, a cache system, or other storage space capable of storing and providing the interactive task information.
[0092] S302. Generate a first task instruction based on the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user based on the interactive task information to promote the product corresponding to the product information to the target user.
[0093] The first task instruction can be understood as a clear instruction to trigger the first large language model to perform an interactive task, and can be used to instruct the object receiving the first task instruction to interact with the target user according to the interactive task information to promote the target product corresponding to the product information to the target user.
[0094] The first task instruction may be a sentence, a question or a combination of a series of instructions. The present application does not limit the specific content and format of the first task instruction.
[0095] Optionally, the first task instruction includes interactive task information. Optionally, the first task instruction includes interactive task information and task description information.
[0096] Optionally, the generated first task instruction complies with a preset task instruction format. The task instruction format specifies the structure and representation method of the first task instruction, is preset and stored on the controller or server side, and can be directly retrieved and used when needed.
[0097] Optionally, in order for the large language model receiving the first task execution to correctly understand and execute the first task instruction, the first task instruction should adopt a task instruction format that can be recognized by the large language model executing the first task instruction. Preferably, the first task instruction format should be consistent with the instruction format of the task instruction in the training sample pre-trained by the large language model.
[0098] As an optional implementation manner, each piece of information in the interactive task information is combined according to a preset task instruction format to obtain a first task instruction.
[0099] As an optional implementation method, a task instruction is generated according to the logical relationship and constraints between the interactive task information. Optionally, after obtaining the interactive task information, the logical relationship and constraints between the interactive task information are first analyzed, or the preset logical relationship and constraints are called, and then the first task instruction is generated using the interactive task information according to the logical relationship and constraints. For example, the task instruction is "You are currently a marketing service assistant of an operator company. Please follow the interactive process defined below to introduce the company's products to user A, answer user A's questions, and attract user A's purchasing interest. User A's user information is..., and the company's product information is... Please follow the following interactive process..."
[0100] S303: Input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
[0101] A large language model (LLM) is a generative deep neural network model based on the Transformer structure, which has powerful natural language processing capabilities and multi-round dialogue capabilities.
[0102] As an optional implementation manner, the pre-trained first large language model may be any type of large language model in the prior art.
[0103] As an optional implementation method, the pre-trained first large language model can be a fine-tuned large language model obtained by training any large language model in the prior art using a data set consisting of multiple groups of first task instructions-human-computer interaction content as training samples.
[0104] The pre-trained first language model has the ability to execute the input first task instruction and promote the target product to the target user, that is, it can execute the interactive task corresponding to the first task instruction, interact with the target user according to the interactive task information, so as to promote the product corresponding to the product information in the interactive task information, that is, the target product, to the target user.
[0105] Specifically, after generating the first task instruction, the generated first task instruction is input into the pre-trained first large language model, and the first large language model executes the first task instruction, that is, executes the interactive task corresponding to the first task instruction, and the first large language model interacts with the target user according to the interactive task information.
[0106] As an optional implementation, the training process of the first language model includes:
[0107] A 1. Input a pre-acquired first training sample into a pre-constructed first large language model to obtain human-computer interaction information corresponding to the first training sample.
[0108] This embodiment requires pre-acquisition of a first training sample, which includes: sample interaction task information, sample human-computer interaction information (including multiple rounds of user input information and label system output information corresponding to each round of user input information), wherein the system output information in the sample human-computer interaction information serves as label data.
[0109] The first training sample is input into a pre-built first large language model according to a preset instruction format, and the first large language model generates actual system output information corresponding to each round of user input information.
[0110] A2. Determine a first loss function based on the difference between the label system output information corresponding to each round of user input information in the first training sample and the actual system output information corresponding to each round of user input information generated by the first large language model, and adjust the model parameters of the first large language model based on the first loss function.
[0111] As an optional implementation method, the default instruction format is:
[0112] [system,user,Thought+Action,Observation,Thought+Finish,user,Thought+Action,Observation,Thought+Finish,…user,Thought+Action,Observation,Thought+Finish]
[0113] Among them, system is the interactive task information; user is the user input information in this round of human-computer interaction; Thought in Thought+Action is to integrate the existing information and think about what to do next; Action is a decision based on the thought stage, which determines the specific action information that needs to be executed next, including relevant information without tool call or tool call; Observation is the result obtained after taking action. When the action taken is tool call, Observation is the return result obtained by calling the existing tool; Thought in Thought+Finish is to integrate the obtained action results with the existing information. Next, it is necessary to answer the user input information in this round of human-computer interaction. Finish is the final answer to the question. All the information in the above process needs to be summarized and a reply is generated.
[0114] In this implementation, optionally, the losses of the system and observation parts are not involved in adjusting the model parameters of the first language model.
[0115] Optionally, a first loss function is determined according to the losses corresponding to the Thought+Action and Thought+Finish parts, and model parameters of the first language model are adjusted based on the first loss function.
[0116] Optionally, relevant model parameters of the first language model are adjusted according to the losses corresponding to the Thought+Action and Thought+Finish parts.
[0117] The human-computer interaction processing method provided in the embodiment of the present application can be applied to a human-computer interaction system. The method automatically generates a first task instruction according to the interaction task information including multiple main processes, and the first task instruction automatically triggers the first large language model to execute the interaction task. Human-computer interaction is performed with the target user according to the interaction task information. The interaction process including multiple main processes is used to achieve global tracking and control of the human-computer interaction state, thereby realizing human-computer interaction that matches the interaction process, rather than relying on complex flowchart design, thereby reducing the configuration requirements of the human-computer interaction system for each process node, so that the human-computer interaction system can respond to user intentions more flexibly instead of strictly following preset process nodes.
[0118] On the other hand, relying on the excellent natural language processing capabilities of the large language model, applying the present application scheme to the human-computer interaction system can comprehensively and accurately understand the interactive task information and user input information, and adjust the system output information according to the real-time human-computer interaction information, rather than relying on preset response scripts, thereby increasing the naturalness and personalization of the conversation, improving the user experience, and thereby improving the conversion rate of marketing activities and user satisfaction.
[0119] Figure 4 A schematic diagram of a processing flow for interacting with a target user based on interactive task information provided in an embodiment of the present application.
[0120] As an optional implementation, Figure 4 As shown, interacting with the target user according to the interactive task information includes steps S401-S402:
[0121] S401. Determine a current main process according to the interaction process and human-computer interaction information, wherein the human-computer interaction information includes historical interaction content and current user input.
[0122] According to the foregoing content, the interaction process is a plurality of predefined main processes designed to achieve the purpose of interaction. Each main process carries specific interaction tasks and goals. The multiple main processes are carried out in a specific order and logic, aiming to guide users to complete specific operations or provide necessary information, ensuring that the human-computer interaction process can follow the interaction process in an orderly manner, so as to achieve the ultimate interaction purpose.
[0123] The human-computer interaction information includes historical interaction content and current user input. The historical interaction content records the previous communication between the user and the system, including the user's historical input information, the system's historical output information, and other interactive behaviors of both parties. Optionally, by analyzing the historical interactions, it is possible to determine the completed main processes and the unfinished main processes. The current user input can be understood as the latest information to be processed input by the target user. Optionally, by analyzing the current user input, it is possible to determine the immediate needs and intentions of the target user.
[0124] Specifically, during the human-computer interaction process, when the human-computer interaction system needs to generate output information, the first language model will conduct a comprehensive analysis of the interaction process and the human-computer interaction information to determine the current main process.
[0125] Optionally, a comprehensive analysis is performed on the interaction process and human-computer interaction information to determine the current main process, including: analyzing the human-computer interaction information to obtain analysis results, which include the completion status of the main process and the immediate needs and intentions of the target users; matching the analysis results with the interaction process, and determining the main process corresponding to the current human-computer interaction link from the interaction process, that is, the current main process.
[0126] S402. For the purpose of completing the current main process, generate first output information according to the interactive task information and the human-computer interaction information, wherein the first output information includes reply information to the current user input and / or guidance information to the target user.
[0127] The core purpose of this step is to generate appropriate first output information based on the current main process, combined with the interactive task information, the context of human-computer interaction and the current user input, and the first output information can effectively promote the progress of the current main process.
[0128] The first output information may be a direct response to the user's current input, answering the user's questions or providing the information they need; it may also be guidance information for the user, aiming to push the user to take the next step along the predetermined interaction process, such as providing purchase links, promotional discounts, etc.; in the scenario of selling products, it may also be persuasive words that combine user needs and product features to facilitate transactions.
[0129] Specifically, after determining the current main process, the first language model integrates the current main process, interactive task information and human-computer interaction information, analyzes the integrated comprehensive information, and generates first output information.
[0130] As an optional implementation, step S402 "generating first output information according to the interactive task information and the human-computer interaction information for the purpose of completing the current main process" includes steps A1-A2:
[0131] A1. Determine a marketing strategy based on the human-computer interaction information.
[0132] The marketing strategy can be understood as a series of methods and means to achieve the purpose of interaction. The marketing strategy aims to strengthen the product features that users are interested in, answer users' doubts, and eliminate purchase barriers as much as possible, so as to guide users to continue to interact along the current trunk process and ultimately promote purchase behavior. For example, if it is found that users are particularly interested in a certain function of the product, the marketing strategy can emphasize this function and provide a detailed introduction and demonstration. If users express concerns about prices, the marketing strategy can provide solutions such as preferential activities or installment payments. This marketing strategy aims to meet user needs to the greatest extent, eliminate user doubts, and thus increase the possibility of successful sales.
[0133] Specifically, when the human-computer interaction system needs to generate output information, the first language model will analyze the most recently updated human-computer interaction information to determine user preference information such as user interests, needs, concerns, and possible purchase barriers. After determining the user preference information, the first language model will conduct an in-depth analysis of the user preference information to determine the marketing strategy corresponding to the most recently updated human-computer interaction information.
[0134] A2. For the purpose of completing the current main process, the first output information is generated according to the marketing strategy, the interactive task information and the human-computer interaction information.
[0135] Furthermore, when generating the first output information, the first language model will comprehensively consider the marketing strategy, the current main process, the interactive task information and the human-computer interaction information, so that the generated first output information can not only effectively promote the progress of the current main process, but also take into account the user's latest personalized needs, and be more targeted and persuasive.
[0136] In this implementation, the first language model can instantly capture and analyze human-computer interaction information during the user's interaction with the human-computer interaction system through its real-time data processing and analysis capabilities, dynamically adjust the marketing strategy based on the latest data, and ensure that the marketing strategy is always synchronized with the user's current status and needs, so that the generated first output information is more targeted and persuasive. This implementation not only improves the user experience, but also significantly improves the efficiency of sales conversion, demonstrating its unique advantages in a fast-paced business environment.
[0137] Further, as an optional implementation, step S402 "generating first output information according to the interactive task information and the human-computer interaction information for the purpose of completing the current main process" includes steps B1-B4:
[0138] B1. Determine a first instruction corresponding to the current user input based on the interactive task information and the human-computer interaction information, wherein the first instruction is used to instruct the object receiving the first instruction to perform a first processing corresponding to the current user input according to the first instruction, and return an execution result corresponding to the first instruction, wherein the first processing includes at least one of memory information retrieval, knowledge query and business processing.
[0139] First, the first language model analyzes the human-computer interaction information to determine the current user intention corresponding to the current user input. For example, the current user intention includes wanting to query certain information, handle a certain business, or obtain a certain service.
[0140] Then, the large language model determines the first instruction for the current user's intention based on the interactive task information, and the first instruction is used to instruct the object that receives the first instruction to perform the first processing corresponding to the current user input and return the execution result corresponding to the first instruction. The first instruction includes the object that executes the first instruction and the execution content, that is, the first processing is executed and the first processing result is returned.
[0141] The object that executes the first instruction can be a module inside the human-computer interaction system, or a service or tool outside the human-computer interaction system. For example, if a user wants to query certain professional knowledge or specific data, it is necessary to call an external knowledge base or database for query. For another example, if a user needs to handle a certain business, and the business involves external services or verification, it is necessary to call an external API or service interface to complete the operation.
[0142] The first processing may be at least one of memory information retrieval, knowledge query or business handling. Memory information retrieval may involve querying the user's historical interaction data to retrieve previous discussion content or user preferences. Knowledge query may involve calling a search engine or knowledge base to obtain answers to specific questions or related information. Business handling may involve calling the enterprise's internal business processing system to complete the service requested by the user, such as order processing, account query, etc.
[0143] B2. Obtain the execution result corresponding to the first instruction.
[0144] After issuing the first instruction, the first language model will wait for and receive the execution result corresponding to the first instruction. The execution result may come from the information retrieval result of an external tool, the query feedback of the knowledge base, or the status update of the business processing.
[0145] B3. Determine a marketing strategy according to the execution result corresponding to the first instruction and the human-computer interaction information.
[0146] After obtaining the execution result of the first instruction, the first language model determines the marketing strategy according to the execution result corresponding to the first instruction and the human-computer interaction information. In this process, the execution result of the first instruction provides key data support or reference information for the formulation of the marketing strategy. For example, if a user queries the detailed information of a product, the execution result may include the sales volume, evaluation and other data of the product. The first language model can formulate more targeted marketing strategies based on this data, such as providing promotion information or recommending related products.
[0147] B4. With the purpose of completing the current main process, the first output information is generated according to the execution result, the marketing strategy, the interactive task information and the human-computer interaction information.
[0148] Finally, the first language model will generate the first output information with the purpose of completing the current main process, taking into account the execution result, marketing strategy, interactive task information and human-computer interaction information corresponding to the first instruction.
[0149] This implementation method obtains the information required by users by calling external tools and services, and formulates effective marketing strategies based on this information, and finally generates targeted output information that meets user needs and promotes sales conversion. This processing method improves the intelligence and automation level of the system and provides users with a more personalized and efficient service experience.
[0150] Figure 5 A flowchart of another human-computer interaction processing method provided in an embodiment of the present application. As an optional implementation method, Figure 5As shown, the human-computer interaction processing method provided in this embodiment also includes steps S501-S502:
[0151] S501. Perform marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generate marketing anomaly information when a marketing anomaly event is identified during the interaction process, wherein the marketing anomaly information includes a marketing anomaly handling strategy corresponding to the marketing anomaly event.
[0152] The marketing anomaly identification strategy can be understood as a set of preset rules and conditions for monitoring and identifying marketing anomalies that may occur during the interaction between the first language model and the target user. The marketing anomaly identification strategy is usually formulated based on historical data, market trends, consumer behavior, laws and regulations, and other factors. This application does not limit the specific content of the marketing anomaly identification strategy.
[0153] The marketing abnormal event can be understood as an abnormal situation that occurs during the interaction between the first language model and the target user and is inconsistent with the normal pattern, expected behavior or specified standards, and may include but is not limited to at least one of the following: continuous user rejection, negative user feedback, abnormal user information, and abnormal interaction content violations.
[0154] When a marketing abnormality event is identified, the human-computer interaction system will generate corresponding marketing abnormality information, which includes at least a marketing abnormality handling strategy corresponding to the marketing abnormality event, and may also include description information of the marketing abnormality event and other content.
[0155] Optionally, after identifying a marketing abnormality event, the human-computer interaction system analyzes the marketing abnormality event and dynamically generates a marketing abnormality handling strategy corresponding to the marketing abnormality event based on the analysis result.
[0156] Optionally, a corresponding marketing exception handling strategy is pre-set for each marketing exception event in the marketing exception identification strategy. After identifying a marketing exception event, the human-computer interaction system directly calls the marketing exception handling strategy corresponding to the marketing exception event.
[0157] As an optional implementation, step S501 is implemented based on the processing concept of a large language model. In this implementation, step S501 includes:
[0158] The pre-trained second language model is triggered to perform marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generates marketing anomaly information when a marketing anomaly event is identified from the interaction process.
[0159] As an optional implementation manner, the pre-trained second large language model may be any type of large language model in the prior art.
[0160] As an optional implementation method, the pre-trained second large language model can be a fine-tuned large language model obtained by training any large language model in the prior art using a data set consisting of multiple groups of human-computer interaction content-marketing abnormal information as training samples.
[0161] The pre-trained second language model can monitor the interaction process between the first language model and the target user for marketing anomalies according to a preset marketing anomaly identification strategy, and generate marketing anomaly information when a marketing anomaly event is identified during the interaction process.
[0162] As an optional implementation, the marketing anomaly identification strategy includes: when the cumulative number of occurrences of the same abnormal scenario reaches a preset number threshold corresponding to the abnormal scenario, determining that a marketing abnormal event corresponding to the abnormal scenario has occurred.
[0163] Specifically, when the cumulative number of occurrences of a certain abnormal scenario in the process of the first language model executing the interactive task reaches a preset threshold, it is determined that a marketing abnormal event corresponding to the abnormal scenario has occurred.
[0164] In this implementation, step S501 includes: matching the latest round of human-computer interaction information with preset abnormal scenario description information, and generating second execution information when the latest round of human-computer interaction information matches the description information corresponding to the first abnormal scenario, wherein the abnormal scenario description information includes description information corresponding to each abnormal scenario type, and the second execution information is used to instruct the object receiving the second execution information to add one to the cumulative number of occurrences of the first abnormal scenario, and return the updated cumulative number of occurrences of the first abnormal scenario; obtain the updated cumulative number of occurrences of the first abnormal scenario; determine whether a marketing abnormal event corresponding to the first abnormal scenario occurs based on the updated cumulative number of occurrences of the first abnormal scenario and the preset number threshold corresponding to the first scenario, and generate marketing abnormal information when it is determined that the marketing abnormal event corresponding to the first abnormal scenario occurs, wherein the marketing abnormal information includes description information of the marketing abnormal event corresponding to the first abnormal scenario and the marketing abnormality handling strategy corresponding to the first abnormal scenario.
[0165] Specifically, the human-computer interaction system will match the latest round of human-computer interaction information with the preset abnormal scenario description information. If the latest round of human-computer interaction information matches the description information of the first abnormal scenario, it means that the current human-computer interaction is in this first abnormal scenario. The human-computer interaction system will generate a second execution information. The object that receives the second execution information will add one to the cumulative number of occurrences of the first abnormal scenario according to the instructions of the second execution information, and return the updated cumulative number. The human-computer interaction system then obtains the updated cumulative number, compares the updated cumulative number with the preset number threshold corresponding to the first abnormal scenario, and if the cumulative number reaches this threshold, the system determines that a marketing abnormal event corresponding to the first abnormal scenario has occurred. Once it is determined that a marketing abnormal event has occurred, the human-computer interaction system will generate marketing abnormal information. The marketing abnormal information includes two parts: the description information of the marketing abnormal event corresponding to the first abnormal scenario (i.e., what specific abnormality has occurred), and the marketing abnormality handling strategy preset for the first abnormal scenario (i.e., how to deal with the marketing abnormal event corresponding to the first abnormal scenario).
[0166] S502: Input the marketing abnormality information into the first large language model, so that the first large language model generates second output information according to the marketing abnormality information.
[0167] After the marketing exception information is generated, the generated marketing exception information is input into the first large language model, and the first large language model generates second output information according to the marketing exception information. The second output information is used to process the marketing exception event according to the marketing exception processing strategy corresponding to the marketing exception event.
[0168] Exemplary Devices
[0169] Corresponding to the above-mentioned human-computer interaction processing method, the embodiment of the present application also provides a human-computer interaction processing device. Figure 6 is a structural diagram of a human-computer interaction processing device provided in an embodiment of the present application, such as Figure 6 As shown, the human-computer interaction processing device provided in the embodiment of the present application includes:
[0170] The first unit 601 is used to obtain interaction task information, wherein the interaction task information includes interaction purpose, user information, product information and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of the target user;
[0171] The second unit 602 is used to generate a first task instruction according to the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user according to the interactive task information to promote the product corresponding to the product information to the target user;
[0172] The third unit 603 is used to input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
[0173] The human-computer interaction processing device provided in the embodiment of the present application can be applied to a human-computer interaction system. The device automatically generates a first task instruction according to the interaction task information including multiple main processes, and the first task instruction automatically triggers the first large language model to execute the interaction task. According to the interaction task information, human-computer interaction is performed with the target user, and the interaction process including multiple main processes is used to achieve global tracking and control of the human-computer interaction state, thereby realizing human-computer interaction that matches the interaction process, rather than relying on complex flowchart design, thereby reducing the configuration requirements of the human-computer interaction system for each process node, so that the human-computer interaction system can respond to user intentions more flexibly instead of strictly following preset process nodes.
[0174] On the other hand, relying on the excellent natural language processing capabilities of the large language model, applying the present application scheme to the human-computer interaction system can comprehensively and accurately understand the interactive task information and user input information, and adjust the system output information according to the real-time human-computer interaction information, rather than relying on preset response scripts, thereby increasing the naturalness and personalization of the conversation, improving the user experience, and thereby improving the conversion rate of marketing activities and user satisfaction.
[0175] Optionally, interacting with the target user according to the interactive task information includes:
[0176] Determine the current main process according to the interaction process and the human-computer interaction information, wherein the human-computer interaction information includes historical interaction content and current user input;
[0177] For the purpose of completing the current main process, first output information is generated according to the interactive task information and the human-computer interaction information, wherein the first output information includes reply information to the current user input and / or guidance information to the target user.
[0178] Optionally, the first language model may be specifically used for:
[0179] Determine a marketing strategy based on the human-computer interaction information;
[0180] For the purpose of completing the current main process, the first output information is generated according to the marketing strategy, the interactive task information and the human-computer interaction information.
[0181] Optionally, the first language model may be specifically used for:
[0182] Determine, according to the interactive task information and the human-computer interaction information, a first instruction corresponding to the current user input, wherein the first instruction is used to instruct the object receiving the first instruction to perform a first process corresponding to the current user input according to the first instruction, and return an execution result corresponding to the first instruction, wherein the first process includes at least one of memory information retrieval, knowledge query, and business handling;
[0183] Obtaining an execution result corresponding to the first instruction;
[0184] Determining a marketing strategy according to the execution result corresponding to the first instruction and the human-computer interaction information;
[0185] With the purpose of completing the current main process, the first output information is generated according to the execution result, the marketing strategy, the interactive task information and the human-computer interaction information.
[0186] Optionally, the device further comprises:
[0187] The fourth unit is used to monitor the interaction process between the first large language model and the target user for marketing anomalies according to a preset marketing anomaly identification strategy, and generate marketing anomaly information when a marketing anomaly event is identified in the interaction process, and input the marketing anomaly information into the first large language model so that the first large language model generates second output information according to the marketing anomaly information, wherein the marketing anomaly information includes a marketing anomaly handling strategy corresponding to the marketing anomaly event.
[0188] Optionally, the fourth unit may be specifically used for:
[0189] The pre-trained second language model is triggered to perform marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generates marketing anomaly information when a marketing anomaly event is identified from the interaction process.
[0190] Optionally, the marketing anomaly identification strategy includes:
[0191] When the cumulative number of occurrences of the same abnormal scenario reaches a preset number threshold corresponding to the abnormal scenario, it is determined that a marketing abnormal event corresponding to the abnormal scenario has occurred;
[0192] The fourth unit can be specifically used for:
[0193] Matching the latest round of human-computer interaction information with preset abnormal scenario description information, and generating second execution information when the latest round of human-computer interaction information matches the description information corresponding to the first abnormal scenario, wherein the abnormal scenario description information includes description information corresponding to each abnormal scenario type, and the second execution information is used to instruct the object receiving the second execution information to add one to the cumulative number of occurrences of the first abnormal scenario, and return the updated cumulative number of occurrences of the first abnormal scenario;
[0194] Obtaining the updated cumulative number of occurrences of the first abnormal scenario;
[0195] Based on the updated cumulative number of occurrences of the first abnormal scenario and the preset number threshold corresponding to the first scenario, determine whether a marketing abnormal event corresponding to the first abnormal scenario has occurred, and when it is determined that a marketing abnormal event corresponding to the first abnormal scenario has occurred, generate marketing abnormality information, wherein the marketing abnormality information includes description information of the marketing abnormal event corresponding to the first abnormal scenario and a marketing abnormality handling strategy corresponding to the first abnormal scenario.
[0196] The human-computer interaction processing device provided in this embodiment belongs to the same application concept as the human-computer interaction processing method provided in the above embodiments of this application, and can execute the human-computer interaction processing method provided in any of the above embodiments of this application, and has the corresponding functional modules and beneficial effects of executing the human-computer interaction processing method. For technical details not fully described in this embodiment, please refer to the specific processing content of the human-computer interaction processing method provided in the above embodiments of this application, and will not be repeated here.
[0197] Exemplary Electronic Devices
[0198] Another embodiment of the present application also provides an electronic device, see Figure 7 As shown, the device includes:
[0199] Memory 200 and processor 210;
[0200] The memory 200 is connected to the processor 210 and is used to store programs;
[0201] The processor 210 is used to implement the human-computer interaction processing method disclosed in any of the above embodiments by running the program stored in the memory 200.
[0202] Specifically, the electronic device may further include: a bus, a communication interface 220 , an input device 230 and an output device 240 .
[0203] The processor 210, the memory 200, the communication interface 220, the input device 230 and the output device 240 are connected to each other via a bus.
[0204] A bus may include a pathway that transfers information between components of a computer system.
[0205] The processor 210 may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the scheme of the present invention. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
[0206] The processor 210 may include a main processor, and may also include a baseband chip, a modem, and the like.
[0207] The memory 200 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes a computer operation instruction. More specifically, the memory 200 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.
[0208] The input device 230 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.
[0209] Output device 240 may include a device that allows information to be output to a user, such as a display screen, a printer, a speaker, etc.
[0210] The communication interface 220 may include any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.
[0211] The processor 210 executes the program stored in the memory 200 and calls other devices, which can be used to implement each step of any human-computer interaction processing method provided in the above embodiments of the present application.
[0212] Exemplary computer program products and storage media
[0213] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the human-computer interaction processing method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.
[0214] The computer program product may be written in any combination of one or more programming languages to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.
[0215] In addition, the embodiment of the present application may also be a storage medium on which a computer program is stored. The computer program is executed by a processor to execute the steps of the human-computer interaction processing method according to various embodiments of the present application described in the above “Exemplary Method” section of this specification, and specifically the following steps may be implemented:
[0216] S301. Acquire interaction task information, wherein the interaction task information includes interaction purpose, user information, product information, and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of a target user.
[0217] S302. Generate a first task instruction based on the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user based on the interactive task information to promote the product corresponding to the product information to the target user.
[0218] S303: Input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
[0219] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.
[0220] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0221] The steps in the methods of each embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.
[0222] The modules and sub-modules in the devices and terminals in the various embodiments of the present application can be combined, divided and deleted according to actual needs.
[0223] In the several embodiments provided in the present application, it should be understood that the disclosed terminals, devices and methods can be implemented in other ways. For example, the terminal embodiments described above are only schematic, for example, the division of modules or submodules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple submodules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.
[0224] The modules or submodules described as separate components may or may not be physically separated, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place, or they may be distributed on multiple network modules or submodules. Some or all of the modules or submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0225] In addition, each functional module or submodule in each embodiment of the present application may be integrated into one processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into one module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or in the form of software functional modules or submodules.
[0226] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0227] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, software units executed by a processor, or a combination of the two. The software units may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0228] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0229] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A human-computer interaction processing method, characterized in that: include: Acquire interaction task information, wherein the interaction task information includes interaction purpose, user information, product information, and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of the target user; Generate a first task instruction according to the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user according to the interactive task information to promote the product corresponding to the product information to the target user; The first task instruction is input into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
2. The method according to claim 1, characterized in that Interacting with the target user according to the interactive task information includes: Determine the current main process according to the interaction process and the human-computer interaction information, wherein the human-computer interaction information includes historical interaction content and current user input; For the purpose of completing the current main process, first output information is generated according to the interactive task information and the human-computer interaction information, wherein the first output information includes reply information to the current user input and / or guidance information to the target user.
3. The method according to claim 2, characterized in that The step of generating first output information according to the interactive task information and the human-computer interaction information for the purpose of completing the current main process includes: Determine a marketing strategy based on the human-computer interaction information; For the purpose of completing the current main process, the first output information is generated according to the marketing strategy, the interactive task information and the human-computer interaction information.
4. The method according to claim 3, characterized in that Before determining the marketing strategy according to the human-computer interaction information, the method further includes: Determine, according to the interactive task information and the human-computer interaction information, a first instruction corresponding to the current user input, wherein the first instruction is used to instruct the object receiving the first instruction to perform a first process corresponding to the current user input according to the first instruction, and return an execution result corresponding to the first instruction, wherein the first process includes at least one of memory information retrieval, knowledge query, and business handling; Obtaining an execution result corresponding to the first instruction; Determining a marketing strategy according to the human-computer interaction information includes: Determining a marketing strategy according to the execution result corresponding to the first instruction and the human-computer interaction information; The step of generating the first output information for the purpose of completing the current main process according to the marketing strategy, the interactive task information and the human-computer interaction information includes: With the purpose of completing the current main process, the first output information is generated according to the execution result, the marketing strategy, the interactive task information and the human-computer interaction information.
5. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Performing marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generating marketing anomaly information when a marketing anomaly event is identified during the interaction process, wherein the marketing anomaly information includes a marketing anomaly handling strategy corresponding to the marketing anomaly event; The marketing abnormality information is input into the first large language model, so that the first large language model generates second output information according to the marketing abnormality information.
6. The method according to claim 5, characterized in that The interaction process between the first language model and the target user is monitored for marketing anomalies according to a preset marketing anomaly identification strategy, and when a marketing anomaly event is identified in the interaction process, marketing anomaly information is generated, including: The pre-trained second language model is triggered to perform marketing anomaly monitoring on the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy, and generates marketing anomaly information when a marketing anomaly event is identified from the interaction process.
7. The method according to claim 5, characterized in that The marketing anomaly identification strategy includes: When the cumulative number of occurrences of the same abnormal scenario reaches a preset number threshold corresponding to the abnormal scenario, it is determined that a marketing abnormal event corresponding to the abnormal scenario has occurred; The step of monitoring the interaction process between the first language model and the target user according to a preset marketing anomaly identification strategy and generating marketing anomaly information when a marketing anomaly event is identified during the interaction process includes: Matching the latest round of human-computer interaction information with preset abnormal scenario description information, and generating second execution information when the latest round of human-computer interaction information matches the description information corresponding to the first abnormal scenario, wherein the abnormal scenario description information includes description information corresponding to each abnormal scenario type, and the second execution information is used to instruct the object receiving the second execution information to add one to the cumulative number of occurrences of the first abnormal scenario, and return the updated cumulative number of occurrences of the first abnormal scenario; Obtaining the updated cumulative number of occurrences of the first abnormal scenario; Based on the updated cumulative number of occurrences of the first abnormal scenario and the preset number threshold corresponding to the first scenario, determine whether a marketing abnormal event corresponding to the first abnormal scenario has occurred, and when it is determined that a marketing abnormal event corresponding to the first abnormal scenario has occurred, generate marketing abnormality information, wherein the marketing abnormality information includes description information of the marketing abnormal event corresponding to the first abnormal scenario and a marketing abnormality handling strategy corresponding to the first abnormal scenario.
8. A human-computer interaction processing device, characterized in that: include: A first unit is used to obtain interaction task information, wherein the interaction task information includes interaction purpose, user information, product information and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of a target user; A second unit is configured to generate a first task instruction according to the interactive task information, wherein the first task instruction is used to trigger the large language model to perform an interactive task, and the interactive task is used to interact with the target user according to the interactive task information to promote the product corresponding to the product information to the target user; The third unit is used to input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
9. An electronic device, characterized in that: including memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the human-computer interaction processing method as described in any one of claims 1 to 7 by running the program in the memory.
10. A computer program product, characterized in that The method comprises computer program instructions, which, when executed by a processor, enable the processor to implement the human-computer interaction processing method according to any one of claims 1 to 7.
11. A human-computer interaction system, characterized in that: include: An input-output module and an interaction module, wherein the input-output module is communicatively connected with the interaction module; The input-output module is used to receive input information from the user and output output information generated by the interaction module; The interaction module is used to obtain interaction task information, wherein the interaction task information includes interaction purpose, user information, product information and interaction process, the interaction process includes multiple main processes for achieving the interaction purpose, and the user information includes identification information of the target user; generate a first task instruction according to the interaction task information, wherein the first task instruction is used to trigger the large language model to perform the interaction task, and the interaction task is used to interact with the target user according to the interaction task information to promote the product corresponding to the product information to the target user; input the first task instruction into a pre-trained first large language model, so that the first large language model promotes the product to the target user by executing the first task instruction.
12. The system according to claim 11, characterized in that The system further comprises: An anomaly monitoring module is used to monitor the interaction process between the first large language model and the target user for marketing anomalies according to a preset marketing anomaly identification strategy, and generate marketing anomaly information when a marketing anomaly event is identified in the interaction process, and input the marketing anomaly information into the first large language model so that the first large language model generates second output information according to the marketing anomaly information, and the marketing anomaly information includes a marketing anomaly handling strategy corresponding to the marketing anomaly event.
Citation Information
Cited By
Robot exception handling method and device, computer equipment and storage medium
CN120715882A
Digital twinning-oriented multi-modal man-machine interaction method
CN120780142A
