Information processing method, electronic device, storage medium and program product
Through artificial intelligence agents, analyzing user intentions expressed in natural language and automatically completing e-commerce operations, solving the problem of user interaction operation complexity in e-commerce applications and improving user experience and operation flexibility.
Patent Information
- Application Number
- CN202411026609.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-29
AI Technical Summary
The user interaction links in e-commerce applications are complex and inflexible, which affects the user experience.
Provide an information processing method, using target agents based on artificial intelligence, analyze user intentions through service requirements expressed in natural language, and call corresponding service tools to replace users to complete operations, simplifying user operation process.
Reduce the complexity of the interactive link, improve operational flexibility, reduce user interaction costs, and improve user experience.
Smart Images

Figure CN119045697A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to an information processing method, electronic equipment, storage medium and program product. Background Art
[0002] With the development of Internet technology, users can purchase various products online through e-commerce applications. In order to improve the user's shopping experience, e-commerce applications provide a variety of services including product search, product browsing, ordering, collection, pre-sales consultation, return and refund. Among them, each service has its operation link. Users can enjoy the corresponding service by following the corresponding operation link in the e-commerce application. For example, the user enters the product name in the search box of the e-commerce application and clicks the search control next to the search box to see the product list page containing the product name. For another example, the user can enter the order details page and view the order details on the order details page. For another example, the user can enter the completed order page and initiate after-sales processes such as return or refund for the completed order on the completed order page.
[0003] Although e-commerce applications have brought great convenience to users, the interactive operations between users and e-commerce applications have complex and diverse links and inflexibility, which affects the user experience. Summary of the invention
[0004] Multiple aspects of the present application provide an information processing method, an electronic device, a storage medium, and a program product to reduce the complexity of interactive links, improve operational flexibility, reduce user interaction costs, and improve user experience.
[0005] An embodiment of the present application provides an information processing method, which is applied to a terminal device installed with a target application, and the target application is associated with a target agent based on artificial intelligence; the method includes: displaying a first page provided by the target application, the first page including a wake-up control of the target agent; in response to a wake-up operation on the wake-up control, displaying a second page for the user to interact with the target agent; in response to an input operation on the second page, obtaining at least one round of service demand description information expressed by the user based on natural language, at least one round of service demand description information belonging to the same user intention; calling the target agent to parse at least one user behavior operation corresponding to the user intention from at least one round of service demand description information, and calling at least one service tool to complete the at least one user behavior operation on behalf of the user; and displaying the execution results of at least part of the at least one service tool on the first page, the second page or the third page as service result information corresponding to the user intention.
[0006] An embodiment of the present application also provides a shopping information processing method, which is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with a target intelligent agent based on artificial intelligence; the method includes: displaying a first page provided by the e-commerce application, the first page including a wake-up control of the target intelligent agent; responding to a wake-up operation on the wake-up control, displaying a second page for the user to interact with the target intelligent agent; responding to an input operation on the second page, obtaining shopping description information expressed by the user based on natural language; calling the target intelligent agent to parse at least one user interaction operation among the search operation, the screening operation, the add-to-cart operation and the order operation from the shopping description information, and calling at least one e-commerce tool among the search tool, the screening tool, the add-to-cart tool and the order tool to complete the corresponding user interaction operation on behalf of the user; and displaying the execution result information of at least part of the at least one e-commerce tool on the first page, the second page or the third page as the purchase result information corresponding to the shopping description information.
[0007] An embodiment of the present application also provides a return information processing method, which is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with a target intelligent agent based on artificial intelligence; the method includes: displaying a first page provided by the target application, the first page including a wake-up control of the target intelligent agent; responding to a wake-up operation on the wake-up control, displaying a second page for the user to interact with the target intelligent agent; responding to an input operation on the second page, obtaining a return description information expressed by the user based on natural language; calling the target intelligent agent to parse at least one user interaction operation among order query operation, logistics information query operation, refund operation, and appointment pickup operation from the return description information, and calling at least one e-commerce tool among order query tool, logistics information query tool, refund tool, and appointment pickup tool to complete the corresponding user interaction operation on behalf of the user; and displaying the execution result information of at least part of the at least one e-commerce tool on the first page, the second page, or the third page as the return result information corresponding to the return description information.
[0008] An embodiment of the present application also provides an information processing method, which is applied to an artificial intelligence-based target agent associated with a target application, the method comprising: receiving at least one round of service demand description information expressed by a user based on natural language, wherein the at least one round of service demand description information belongs to the same user intent; parsing at least one user behavior operation corresponding to the user intent from the at least one round of service demand description information, and calling at least one service tool to complete the at least one user behavior operation on behalf of the user; and providing the execution results of at least part of the at least one service tool to the target application as service result information corresponding to the user intent, so that the target application can display it.
[0009] An embodiment of the present application also provides an electronic device, comprising: a memory and a processor; the memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program to implement each step in the information processing method, shopping information processing method and return information processing method provided in the embodiment of the present application.
[0010] The embodiments of the present application also provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor implements each step of the information processing method, shopping information processing method and return information processing method provided in the embodiments of the present application.
[0011] In the embodiment of the present application, a page corresponding to the target intelligent agent is provided in the target application, and the user can enter the service demand description information based on natural language expression on the page. The target intelligent agent can parse the user behavior operation corresponding to the user's intention from the service demand description information, and call the pre-defined service tool to complete at least one user behavior operation on behalf of the user; the execution result of the service tool is used as the service result information corresponding to the user's intention. During the whole process, at least part of the user behavior operation can be completed by the target intelligent agent instead of the user, which is conducive to simplifying the user's operation process, reducing the complexity of the interaction link, improving the operation flexibility, reducing the user interaction cost, and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0013] Figure 1 A flowchart of an information processing method provided for an exemplary embodiment of the present application;
[0014] Figure 2a A schematic diagram of a first page provided for an exemplary embodiment of the present application;
[0015] Figure 2b A schematic diagram of another first page provided for an exemplary embodiment of the present application;
[0016] Figure 2c A schematic diagram of a second page provided for an exemplary embodiment of the present application;
[0017] Figure 2d A schematic diagram of a second page or a third page provided for an exemplary embodiment of the present application;
[0018] Figure 2e A schematic diagram of another second page or a third page provided for an exemplary embodiment of the present application;
[0019] Figure 2f A schematic diagram of a task list page provided for an exemplary embodiment of the present application;
[0020] Figure 3a A reasoning framework diagram of a target intelligent agent provided for an exemplary embodiment of the present application;
[0021] Figure 3b A schematic diagram of a thinking-action-viewing process involved in a target intelligent agent provided for an exemplary embodiment of the present application;
[0022] Figure 4a A flowchart of an information processing process in which a target agent includes a main agent and multiple sub-agents is provided for an exemplary embodiment of the present application;
[0023] Figure 4b A schematic diagram of the structure of a target agent provided for an exemplary embodiment of the present application includes a main agent and multiple sub-agents;
[0024] Figure 5 A schematic structural diagram of an electronic device provided for an exemplary embodiment of the present application. DETAILED DESCRIPTION
[0025] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. 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 making creative work are within the scope of protection of the present application.
[0026] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.
[0027] In view of the above technical problems, in an embodiment of the present application, a page corresponding to the target intelligent agent is provided in the target application, on which the user can enter the service demand description information based on natural language expression, and the target intelligent agent can parse the user behavior operation corresponding to the user's intention from the service demand description information, and call the predefined service tool to complete at least one user behavior operation on behalf of the user; the execution result of the service tool is used as the service result information corresponding to the user's intention. In the whole process, at least part of the user behavior operation can be completed by the target intelligent agent instead of the user, which is conducive to simplifying the user's operation process, reducing the complexity of the interactive link, improving the flexibility of operation, reducing the user interaction cost, and improving the user experience.
[0028] A solution provided by an embodiment of the present application is described in detail below in conjunction with the accompanying drawings.
[0029] Figure 1 A flowchart of an information processing method provided for an exemplary embodiment of the present application is provided, wherein the method is applied to a terminal device having a target application installed thereon, wherein the target application is associated with a target agent based on artificial intelligence; Figure 1 As shown, the method includes:
[0030] 101a, displaying a first page provided by a target application, where the first page includes a wake-up control for a target agent;
[0031] 102a, in response to the awakening operation on the awakening control, displaying a second page for the user to interact with the target agent;
[0032] 103a. In response to an input operation on the second page, obtaining at least one round of service demand description information expressed by the user based on natural language, wherein the at least one round of service demand description information belongs to the same user intention;
[0033] 104a, calling a target agent to parse at least one user behavior operation corresponding to the user's intention from at least one round of service demand description information, and calling at least one service tool to complete the at least one user behavior operation on behalf of the user;
[0034] 105a. Display the execution results of at least part of the at least one service tool on the first page, the second page or the third page as service result information corresponding to the user intention.
[0035] In this embodiment, the target agent (AI Agent) based on artificial intelligence is an intelligent entity that can perceive the environment, make decisions and perform actions. Unlike traditional artificial intelligence, AI Agent has the ability to gradually complete a given goal by thinking independently and calling tools, without the need to manually specify each step of the operation. In other words, the target agent is used as a representative of a person or an organization to perform a certain behavior, reduce the complexity of a person or organization's work, and reduce workload and communication costs. Among them, the Large Language Model (LLM) can be used as the core computing engine of the target agent, providing the target agent with powerful text generation and comprehension capabilities. On this basis, the target agent can show a wider range of functions, such as dialogue, task completion, reasoning, and a certain degree of autonomous behavior.
[0036] In this embodiment, the implementation method of the target agent is not limited. For example, the target agent can be implemented as a system-level service, for example, the target agent is implemented as a voice assistant on a terminal device. For another example, the target agent can also be implemented as an application-level service, for example, the target agent is implemented as an intelligent assistant of a target application.
[0037] The target application may be any type of application, and any application that can process information based on the target agent is applicable to the embodiments of the present application. For example, the target application may include but is not limited to: e-commerce applications (such as shopping applications, travel applications, etc.), social applications, game applications, instant messaging applications, online education applications, etc.
[0038] In this embodiment, the target application is deployed on the terminal device, and accordingly, the target agent can be deployed on the server side or on the terminal device where the target application is located, without limitation. In either case, the target application can correspond to a first page corresponding to the target agent, and the first page includes a wake-up control for the target agent, which is used to activate the target agent.
[0039] For example, the first page may be the home page of the target application, and the wake-up control may be a button or icon set in a designated area of the home page row, and the shape of the button or icon may be circular, square, water drop or other shapes. Figure 2a In the example, the wake-up control is implemented as an AI assistant, and the wake-up control corresponds to a prompt message, such as "Xiaobao, your AI assistant". In addition, the wake-up control is located in the middle of the menu bar on the home page, but it is not limited thereto. Figure 2a The target application's home page also includes other information, such as a search box, a QR code, a message dialog box, and product information, which are only examples and are not limited to these.
[0040] For another example, the first page is a message page, the wake-up control is a specific reminder message displayed at a specified position on the message page, and the specific reminder message supports interactive operations. Figure 2b In the figure, the target intelligent agent is implemented as an AI assistant, and the wake-up control is implemented as a reminder message corresponding to the AI assistant. The user can click the reminder message to enter the interactive page in the dialog state, that is, the second page. Among them, the reminder message includes a wake-up icon and text information. The text information can be "Hello, I am your intelligent assistant", Figure 2b In addition to the reminder messages corresponding to the AI assistant, it also includes other reminder messages, for example, reminder messages sent by other merchants. This is only an example and is not limited to this.
[0041] For another example, the first page can also be implemented as other pages in the target application. For example, the first page can be implemented as a product list page or a product details page, and the wake-up control is located above the product list page, and the position can be moved.
[0042] In this embodiment, the user can initiate a wake-up operation on the wake-up control, and the terminal device responds to the wake-up operation on the wake-up control by displaying a second page for the user to interact with the target agent. Figure 2c As shown, an example of a second page is provided, but it is not limited to this. Among them, the second page includes a keyboard and a voice input control. Among them, the wake-up operation is not limited. For example, the wake-up operation can be a user operation initiated for the wake-up control, for example, a single-click operation, a double-click operation, a long press operation, a drag operation, etc. of the user on the first page for the wake-up control. For another example, the terminal device has a voice recognition function, and the wake-up operation can be implemented as a voice wake-up operation. The wake-up word can be pre-set, and when the user's wake-up word is detected, the second page for the user to interact with the target intelligent agent is displayed. For another example, the terminal device has a head recognition function, and the wake-up action is pre-set, for example, "nodding". When the terminal device displays the first page, if the "nodding" action is detected, it is determined to initiate a wake-up operation for the wake-up control, and the second page for the user to interact with the target intelligent agent is displayed. For another example, the terminal device has an eye recognition function. When it is detected that the user's eyes are looking at the wake-up control, it is determined to initiate a wake-up operation for the wake-up control, and the second page for the user to interact with the target intelligent agent is displayed.
[0043] Optionally, in response to the wake-up operation of the wake-up control, the second page can be displayed directly, or in response to the wake-up operation of the wake-up control, the task information can be displayed through a page, a pop-up window or a floating layer, etc.; in response to the user operation on the task information, the second page is displayed. Among them, the task information can also be implemented as verification information, game information, recharge information, advertising information, etc. Among them, the verification information includes a jigsaw task, and in response to the user sliding the missing image block to the missing position, the jigsaw task is completed to verify the user's real identity or adult identity, and the second page is displayed. Various games can be displayed on the game information, and in response to the user interaction operation initiated for the game, the second page is displayed when the user successfully completes the game. The recharge information includes user recharge membership information, recharge AI assistant information, etc., and in response to the user's recharge operation on the recharge information, the second page is displayed. The advertising information includes advertising images or advertising videos, etc., and in response to the operation of the advertising information being played, the second page is displayed.
[0044] It should be noted that, in the case of displaying task information, the above-mentioned operation of displaying task information can also be considered as a part of the wake-up operation, and there is no limitation on this.
[0045] In this embodiment, the user can initiate an input operation on the second page, and the input operation may include but is not limited to: voice input operation, keyboard input operation or handwriting input operation. The terminal device responds to the input operation on the second page and obtains at least one round of service demand description information expressed by the user based on natural language, and at least one round of service demand description information belongs to the same user intention. Service demand description information is a detailed elaboration and explanation of service requirements so that the target agent can clearly understand the nature, scope and requirements of the required service. The service demand description information can be the user's demand description information for the set object (such as a commodity), for example, "I want a thermos cup" or "I want to buy a jacket under 500 yuan", etc. The service demand description information can also be the user's status description information, for example, "It's so hot today", "I don't have a suit", "My shoulders are not very comfortable", etc. The user intention is the goal or demand that the user expects to achieve when using the target application. For example, if the user's intention is "buy a jacket under 500 yuan", at least one round of service demand description information can be at least one round of description information around the color, material, thickness, size, etc. of the jacket.
[0046] In this embodiment, the terminal device can call the target agent to parse at least one user behavior operation corresponding to the user intention from at least one round of service demand description information. Among them, the user behavior operation is a behavior operation that the user needs to perform in the target application to realize the user intention. For example, when the target application is implemented as an e-commerce application, the target agent can cover the entire service link that the e-commerce application can provide, for example, for pre-sales, in-sales and after-sales involved in the scene processing, pre-sales include but not limited to: search, screening and add purchase, etc.; in-sales include but not limited to: bargaining, price and order, etc.; after-sales include but not limited to: return, refund and door-to-door pickup, etc. For example, the user intention is "buy a red thin jacket below 500 yuan", the user behavior operation includes but is not limited to: search jacket, filter price, filter color, filter thickness, select size and place an order and other behavior operations. Among them, the target agent can parse one or more user behavior operations for each round of service demand information. In this embodiment, the target agent can maintain at least one service tool for the target application in advance, and each service tool can correspond to a user behavior operation. In other words, the browsing, decision-making, purchasing, and fulfillment processes of users in the target application can be transformed into at least one service tool corresponding to the target intelligent entity. The service tool can be a capability unit provided to the target intelligent entity for calling. For example, the service tool can be an application programming interface (API), such as an order search interface, a product query information interface, etc. The above API interface corresponds to a corresponding execution unit, which can complete the corresponding function.
[0047] For example, in the case where the target application is an e-commerce application, at least one service tool may include a basic service tool and a high-level functional service tool, wherein the basic service tool may include but not be limited to: a query service tool and an execution service tool. Among them, the query service tool is a tool that can obtain product information or order information through operations such as search or screening, and does not have a service process. For example, the query service may include but not be limited to: a search tool, a screening tool, a logistics query tool, etc. An execution service tool is a tool that can define a set of service processes in order to execute a certain user behavior operation, such as placing an order, adding a purchase, and refunding. For example, the ordering tool needs to define a set of processes including selecting a color, selecting a size, submitting an order, and paying. The execution service tool may include but not be limited to: an ordering tool, a tool for urging shipment, a return tool, an appointment for door-to-door pickup, a tool for initiating a return and refund, and a tool for asking merchants, etc. Among them, the high-level functional service tool is a tool that realizes a part of the application scenario through high-level functions, which can be defined and implemented as needed, but is not limited to this. For example, price tracking is a high-level functional service tool that can regularly record commodity prices and send notifications to users when prices drop. High-level functional service tools can include but are not limited to: helping with bargaining, helping with grabbing, etc.
[0048] In this embodiment, the terminal device calls at least one service tool through the target intelligent agent to complete at least one user behavior operation on behalf of the user, thereby reducing the user's operation links in e-commerce applications, increasing the flexibility of using e-commerce operations, and improving user experience.
[0049] Optionally, in addition to parsing at least one user behavior operation corresponding to the user intention from at least one round of service demand description information, the target agent can also parse out at least one information object associated with the user behavior operation. The information object is a type of input information required to implement at least one user behavior operation. For example, the information object may include but is not limited to: search keywords, queried order information, product type information, product name information, etc. Accordingly, in the process of the terminal device calling at least one service tool through the target agent to replace the user to complete at least one user behavior operation, the information object associated with at least one user behavior operation can be input into the at least one service tool to replace the user to complete at least one user behavior operation and obtain the execution result of the at least one service tool.
[0050] For example, the service tool is implemented as a search tool, the user behavior operation is to search for "jackets", the input of the search tool is "jackets", the target intelligent agent calls the search tool, and uses "jackets" as the input information of the search result. The execution result of the search tool is the search result information with "jackets" as the keyword, that is, the product list information of "jackets".
[0051] In this embodiment, the execution results of at least part of the at least one service tool can be displayed as the service result information corresponding to the user's intention. For example, from the type dimension of the service tool, some display service tools can output the execution results to the outside and display them, while other auxiliary service tools provide intermediate results for other service tools, and the execution results may not be displayed. For another example, the first category of service tools and the second category of service tools can be pre-set, and the execution results of the first category of service tools are displayed, while the execution results of the second category of service tools are not displayed. Among them, the implementation method of displaying the execution results of at least part of the service tools is not limited. An exemplary description is given below.
[0052] For example, the execution results of some service tools are displayed. For example, the execution result of the last service tool is displayed, or the execution result of the service tool in the middle of the execution order and the execution result of the last service tool are displayed. For example, in the process of user return, the user behavior operations parsed by the target agent include order query operation, return and refund operation, and appointment door-to-door pickup operation. Correspondingly, the corresponding three service tools include: order query tool, return and refund tool, and appointment door-to-door pickup tool, wherein the query result of the order query tool is the order information on which the user can perform the return operation, and the order information is displayed to the user so that the user can select the target order to be returned. The target agent calls the return and refund tool to return and refund the target order, and calls the appointment door-to-door pickup tool to make an appointment for the courier to pick up the item, and displays the pickup code to the user. In the above process, the execution results of the order query tool and the appointment door-to-door pickup tool are displayed, that is, the execution results of some service tools are displayed.
[0053] For another example, the execution results of at least one service tool are displayed without limitation. For example, the service demand description information input by the user is "It is very hot today", the terminal device calls the target agent, and the user behavior operation corresponding to the user intention is parsed from the service demand description information: search for product information matching "It is very hot" such as cooling or sun protection, the service tool is a search tool, the information object is "cooling" or "sun protection", and the search results of the search tool can include but not be limited to: "sun umbrella", "sunscreen", "popsicle", "cold drink", "air conditioner" and "fan" and other product information. Among them, the search results of the search tool can be used as the service result information corresponding to the user's intention.
[0054] In this embodiment, the execution results of at least part of the at least one service tool can be displayed on the second page, and the execution results of at least part of the at least one service tool can be displayed on the third page, depending on the implementation form of the target application or the type of execution results.
[0055] The third page may be a new page different from the second page. For example, when the user initiates an input operation on the second page, the terminal device jumps from the second page to the third page, and the execution result of at least part of the service tool is displayed on the third page. In an implementation scheme of the target application, the execution result of at least part of the service tool in at least one service tool may be displayed on the second page by default; in another implementation scheme of the target application, the execution result of at least part of the service tool in at least one service tool may be displayed on the third page by default; in another implementation scheme of the target application, the execution result of at least part of the service tool in at least one service tool may be displayed on the first page by default; in another implementation scheme of the target application, it may be determined whether the execution result is displayed on the first page, the second page, or the third page according to the type of the execution result or the type of the service tool. Specifically, before displaying the execution result of at least part of the service tool in at least one service tool, the type of the execution result to be displayed or the type of the service tool to be displayed may be determined. If it belongs to the first type, the execution result of the service tool is displayed on the first page; if it belongs to the second type, the execution result of the service tool is displayed on the second page; if it belongs to the third type, the execution result of the service tool is displayed on the third page.
[0056] It should be noted that the specific page on which the execution results of at least part of at least one service tool are displayed can be determined according to the implementation method or application scenario of the target application. For example, in some applications, it is displayed on the first page, such as social applications; in other applications, it may be chosen to be displayed on the second page, such as shopping applications. For another example, for the same application, in some application scenarios, it can be displayed on the first page or the second page, for example, the execution results are displayed as text, and in other scenarios, it can be displayed on a new page (such as the third page), for example, the execution results are video information.
[0057] It should be noted that the execution results of at least some of the at least one service tool are displayed on the first page, the second page or the third page in any manner. For example, the display method may include but is not limited to: pop-up window, dialog box, notification bar, floating prompt, drop-down list, sidebar, card, carousel, etc. The specific method may be determined according to the implementation method or application scenario of the target application.
[0058] It should be noted that the service demand description information can be one round or multiple rounds. In the case where the service demand description information is implemented as one round, one round of service demand description information can be the service demand description information input by the user at one time, and the one service demand description information can be located in the same voice, or in the same text, or it can be a chat record information with other users forwarded by the user, and the chat record information includes multiple texts. In the case where the service demand description information is implemented as multiple rounds, the user can input the service demand description round by round, and accordingly, the target intelligent agent can be called to parse the service demand description information round by round, and the service tool can be called round by round. The execution result of this round is not a one-time input of multiple rounds of service demand description information, not a one-time parsing of multiple rounds of service demand description information, nor a one-time call of the service tools required for multiple rounds of service demand description information. Among them, the parsing results of the previous round and the execution results of the service tool, etc., can be input to the target intelligent agent to participate in the parsing of the next round of service demand description information.
[0059] In the embodiment of the present application, a page corresponding to the target intelligent agent is provided in the target application, and the user can enter the service demand description information based on natural language expression on the page. The target intelligent agent can parse the user behavior operation corresponding to the user's intention from the service demand description information, and call the pre-defined service tool to complete at least one user behavior operation on behalf of the user; the execution result of the service tool is provided to the user as the service result information corresponding to the user's intention. During the whole process, at least part of the user behavior operation can be completed by the target intelligent agent instead of the user, which is conducive to simplifying the user's operation process, reducing the complexity of the interactive link, improving the flexibility of operation, reducing the user interaction cost, and improving the user experience.
[0060] In an optional embodiment, the service demand description information of the current round can be obtained while the target agent is called to parse the service demand description information. In the process of parsing the service demand description information of the current round, parsing anomalies may occur. For example, the parsing result indicates that the user demand is unclear. For example, in the case of multiple orders, the service demand description information of the user in the current round indicates that the order should be returned, and the parsing result of the target agent is unclear about which order to return. In this case, a parsing anomaly occurs. For another example, the service demand description information of the current round indicates that the target order should be returned, but the parsing result of the target agent indicates that the target order has been returned. In this case, a parsing anomaly also occurs.
[0061] In response to the above-mentioned situation where parsing anomalies occur, a specific implementation method of obtaining at least one round of service demand description information expressed by the user based on natural language in response to input operations on the second page includes: the terminal device responds to the input operations on the second page to obtain the current round of service demand description information expressed by the user based on natural language; if a parsing anomaly occurs in the process of calling the target intelligent agent to parse the user behavior operation from the current round of service demand description information, then output intention query information to guide the user to input the next round of service demand description information to improve the user's intention. Among them, the method of outputting the intention query information is not limited. For example, the intention query information can be directly output on the second page or the third page.
[0062] Optionally, if a parsing exception occurs in the process of calling the target intelligent agent to parse the user behavior operation from the current round of service demand description information, the implementation method of outputting the intention query information to guide the user to input the next round of service demand description information is not limited and is exemplified below.
[0063] Example A1: When the user behavior operation cannot be parsed from the current round of service demand description information, it means that the parsing cannot continue, and re-input guidance information is output to guide the user to re-enter the current round of service demand description information to improve the user's intention.
[0064] For example, when the current round of service demand description information indicates a return for the target order, the target agent's parsing result is that the target order has been returned and the target order cannot be returned. The user behavior operation is not parsed, and a parsing exception occurs. Then, the guidance information "The target order has been returned, do you need other services?" is output to the user to guide the user to re-enter the service demand description information.
[0065] Example A2: When the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service demand description information includes multiple candidate information, indicating that the parsing cannot continue, multiple candidate information is output for the user to select target information as the next round of service demand description information. It should be noted that if a user behavior operation is parsed from the current round of service demand description information, it can be considered that the execution result of the service tool corresponding to the user behavior operation includes multiple candidate information; if multiple user behavior operations are parsed from the current round of service demand description information, it can be considered that the execution result of the service tool corresponding to the last user behavior operation includes multiple candidate information. The user selects the target information from the multiple candidate information as the next round of service demand description information, and the target agent continues to parse it. For example, when the current round of service demand description information indicates "return", the target intelligent agent parses out multiple order information that can be returned. The target intelligent agent cannot clearly determine which order to return, and a parsing exception occurs. The target intelligent agent can then output the multiple order information that can be returned, so that the user can select the target order that he wants to return from the multiple order information that can be returned, and use the target order as the next round of service demand description information to complete the user's intention.
[0066] The implementation method of outputting multiple candidate information is not limited. For example, when asking a user which order among multiple orders to return, multiple orders are output to the user in the form of cards. The user can view multiple orders by sliding the card left and right and select the target order to return. For another example, the service demand description information entered by the user is "Next Tuesday I will go hiking with my friends at Qingliang Peak. Please buy me a music player". After parsing, the target agent calls the search tool and outputs multiple "music player" product information. The multiple product information can be slid left and right. Figure 2d In the example, the product F1 is currently displayed, and the performance and price of the product F1 are displayed only for example and are not limited thereto. Figure 2d In the example, in addition to outputting multiple product information, the target agent also outputs in the dialog box "It will rain in Lin'an next week. You need a speaker that is 'waterproof' and 'long battery life'. These models are also good looking." Among them, "waterproof" and "long battery life" are popped up when the user needs to provide additional information. The user can trigger "waterproof" and "long battery life". After the user triggers, the user-triggered information can be used as new service demand description information. Figure 2d The illustration is made by taking the second page or the third page as an example.
[0067] Among them, when the user inputs through voice, the terminal device can recognize and type in real time to display the service demand description information input by the user; when the user inputs the service demand description information through the keyboard, the service demand description information input by the user once can be directly displayed; the service demand description information currently sent can be displayed, and the service demand description information previously sent can be overwritten and not displayed, or the service demand description information previously sent can be reduced.
[0068] It should be noted that the multiple candidate information may be for the user to specify a specific order or product, or for the user to specify a specific time, etc., and there is no limitation on this.
[0069] Of course, in addition to outputting the intent query information in the form of cards, the intent query information can also be output in other ways, such as outputting the intent query information in the form of floating layers, pop-up windows, etc. The intent query information can be in the form of text, pictures, or a combination of pictures and text.
[0070] Example A3: When the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service requirement description information includes question information, indicating that the parsing cannot continue, the question information and response guidance information are output to guide the user to enter the answer information of the question information as the next round of service requirement description information.
[0071] For example, the user's current round of service demand description information is "want to buy something", and the user behavior operation parsed by the target agent from the current round of service demand description information includes "buy something". This user behavior operation corresponds to the screening tool, and the screening result of the screening tool includes question information, that is, "what kind of thing", then the target agent can output the question information and reply guidance information, for example, in Figure 2e In the example, the reply guidance information may include "I want something like what", where "what" and "what" are reply guidance information. The reply guidance information corresponds to an input box, which displays the reply guidance information. The user can enter the answer information in the input box, and the answer information can cover the reply guidance information. For example, if the answer information entered by the user is "sunshade" and "umbrella", "I want a sunshade umbrella" can be used as the next round of service demand description information. Figure 2e The illustration is made by taking the second page or the third page as an example.
[0072] In an optional embodiment, the implementation method of calling the target agent to parse at least one user behavior operation corresponding to the user intent from at least one round of service demand description information is not limited. The implementation method varies depending on the deployment location of the target agent, and an exemplary description is given below. Example B1, the target agent and the target application are both running locally on the terminal device, and at least one round of service demand description information is sequentially input into the target agent for intent parsing to obtain at least one user behavior operation corresponding to the user intent and its associated information object.
[0073] Example B2, the target agent runs on the server, the target application runs on the terminal device, and at least one round of service demand description information is sent to the server in sequence, so that the server can input at least one round of service demand description information into the target agent in sequence for intent analysis, so as to obtain at least one user behavior operation corresponding to the user intent and its associated information object. Wherein, in the case where at least one round of service demand description information is implemented as multiple rounds of service demand description information, each time the user inputs a round of service demand description information, the round of service demand description information can be provided to the server for analysis by the target agent on the server.
[0074] Optionally, no matter which of the above situations, at least one round of service demand description information is sequentially input into the target intelligent agent for intent parsing to obtain at least one user behavior operation corresponding to the user intention and its associated information object. The following is a specific implementation method: the target intelligent agent includes an intent parsing model, and the intent parsing model can perform intent parsing for at least one round of service demand description information to obtain at least one user behavior operation representing the user intention and its associated information object. Among them, the intent parsing model can include but is not limited to: a large language model (LLM). Among them, the ability of reasoning and acting can be combined in the intent parsing model to improve the performance of the model in language understanding and interactive decision-making tasks. By interleaving the generation of reasoning trajectories and generating actions (such as calling service tools for user behavior operations), greater synergy is generated between reasoning and acting. The reasoning trajectory helps the model induce, track and update the action plan, and handle abnormal situations, while the action allows it to connect with external resources (such as user input, execution interface of the service tool) and collect additional information. For example, the ability of reasoning and acting can be achieved through the ReAct (Reasoning and Acting) method, which is a method that combines reasoning and acting and aims to improve the performance and interpretability of large language models (e.g., LLMs) in solving complex tasks.
[0075] Specifically, for any round of service demand description information, any round of service demand description information is input into the target intelligent agent, and within the target intelligent agent, prompt words are generated according to the set prompt word template format based on any round of service demand description information, tool description information and context information; the prompt words are input into the intent parsing model corresponding to the target intelligent agent for intent parsing to obtain at least one user behavior operation and its associated information object.
[0076] Among them, the prompt word template may include: tools, tool names, chat history, input, and agent thinking process. The tools variable is a list containing the service tools corresponding to the target agent. Each element in the list contains the name of the tool and the description of the tool. The tool variable is a list of tool names, the chat history is the historical information of the conversation between the user and the target agent, the input is the service demand description information input by the user, and the agent thinking process is the previous thinking process. Furthermore, the prompt word template also includes the role information played by the target agent. For example, in an e-commerce application, the role played by the target agent is an AI smart assistant.
[0077] Further optionally, in each round of parsing, at least one user behavior operation and its associated information object corresponding speech information may be output, wherein the speech information is speech information based on natural language description, and the speech information may be understood as the process of the target intelligent agent thinking and taking action.
[0078] For example, the following is an exemplary description of at least one round of service demand description information and at least one user behavior operation output by the target agent and the corresponding speech information of the associated information object.
[0079] The first round of service demand description information is: "I want to buy a jacket of brand X, within 500 yuan";
[0080] Target Agent: "Show a list of products (category - jacket, brand - X, price - less than 500 yuan)"; The second round of service demand description information is: "I looked at it and found that it seems that they are all the kind of clothes for spring and autumn, not suitable for winter. I want the kind for winter, but these are too thin";
[0081] Target agent: "Show product list (category - jackets, winter warmth, brand - X, price - less than 500 yuan)"
[0082] The third round of service demand description information is: "After looking at it, it seems that there are fewer choices. Is it not so cheap? Then a more expensive one is acceptable";
[0083] Target Agent: “I don’t know how much more expensive it will be, it’s a bit vague, so I decided to ask the user how much more expensive it is for them to accept?”;
[0084] The fourth round of service demand description information is: "within 800";
[0085] Target Agent: "Show a list of products (category - jackets, winter warmth, brand - X, price - less than 800 yuan)"
[0086] The user views the product list and manually adds the desired items to the shopping cart.
[0087] In an optional embodiment, the first page, the second page, or the third page of the target application includes the task list entry information. Figure 2c As shown, the second page includes the task list entry information as an example for exemplary display. The method provided in the embodiment of the present application also includes: responding to the trigger operation of the task list entry information, displaying the task list, the task list includes at least one task and its execution status displayed in the form of a card; wherein the task refers to the task executed by the target agent based on the user's intention; the execution status of the task includes ended, canceled, executing or terminated. In addition, the tasks in the task list may also include product pictures, product details information, task names, task introductions, etc. As Figure 2f As shown, the four tasks included in the task list are displayed, and the four tasks are: Task C1, Task C2, Task C3 and Task C4. For example, for commodity G1, Task C1 is to help bargaining, the status of Task C1 is canceled, and the task description can be "You canceled the task"; for commodity G2, Task C2 is to urge delivery, the status of Task C2 is in execution (e.g., urging delivery), and the task description can be "The merchant is replenishing the stock and will deliver the goods in 3 days"; for commodity G3, Task C3 is to help with grabbing, which means to buy first when the set time arrives, or to buy immediately once the new product is available, the status of Task C3 is ended, and the task description can be "Timeout and failure to pay, the preferential price has expired"; for commodity G4, Task C4 is to help with bargaining, the status of Task C4 is ended, and the task description can be "No discount was negotiated". It should be noted that the task can be of order dimension, that is, a task is established for an order, or it can be of commodity dimension, that is, a task is established for a commodity, and there is no limitation on this.
[0088] It should be noted that the task list page can also be implemented on any page of the target application (such as the first page, the second page or the third page) through a pull-up list, a drop-down list or a sidebar. In response to a pull-up operation on any page of the target application, a task list page in the form of a pull-up list is displayed; or, in response to a pull-down operation on any page of the target application, a task list page in the form of a drop-down list is displayed; or, in response to a sliding operation from the left or right edge of the screen to the center on any page of the target application, a task list page in the form of a sidebar is displayed. Optionally, a function control is displayed in the associated area of at least one task, and tasks in different execution states correspond to different function controls, and different function operations on the task can be triggered by different function controls. For example, in Figure 2f In the figure, the status of task C1 is "cancelled", and the functional control corresponding to task C1 is the "re-initiate" control. The terminal device responds to the triggering operation of the functional control and re-initiates the task; the status of task C2 is "executing", and the functional control corresponding to task C2 is the "view" control. The terminal device responds to the triggering operation of the view control and displays the task details page of task C2. The task details page displays various thoughts and actions performed by the target intelligent agent to complete task C2; the status of task C3 is "ended", and the functional control corresponding to task C3 is "re-initiate", which is the same as or similar to the "re-initiate" control of the above-mentioned task C1; the functional control of task C4 is the "view" control. For details, please refer to the introduction of the "view" control in task C2, which will not be repeated here.
[0089] Optionally, the task list entry information includes a quantity identifier, which corresponds to the number of tasks included in the current task list. Figure 2c In the example, the task list entry information includes a quantity identifier, and the quantity identifier is 7, but it is not limited to this. The number of tasks in the task list is not limited. For example, a maximum of 30 tasks can be retained. If the number of tasks exceeds 30, the tasks in the non-executable state can be deleted according to the creation time. In addition, it also supports users to manually delete tasks in the task list.
[0090] In an optional embodiment, the execution results of at least part of at least one service tool are displayed on the first page, the second page or the third page. The implementation method of the service result information corresponding to the user intention is not limited. An example is provided below. According to the attribute information of the service tool, at least part of the service tools are determined from at least one service tool; for example, the attribute information of the service tool includes: type or execution order, etc. For example, the execution result of the set service tool (such as a query tool) is output; for another example, the execution result of the last service tool obtained by each round of parsing of the target intelligent agent is output; for another example, the execution result of the last service tool obtained by parsing is output, and this is not limited; the output templates corresponding to at least part of the service tools are loaded; wherein each service tool corresponds to its own output template, and the output template may include information such as text, layout style, etc.; the execution results of at least part of the service tools are added to the output templates corresponding to at least part of the service tools, respectively, to obtain the service result information corresponding to the user intention; the service result information is displayed on the first page, the second page or the third page.
[0091] Furthermore, if the target agent does not use the service tool but uses the intent parsing model to output text information, the set text component can be used to display the text information. For example, the service requirement description information entered by the user is "Help me turn on the fan". After parsing, the target agent finds that it cannot complete the user's service requirement, so it calls the text component to output "Sorry, this operation is not supported yet".
[0092] In an optional embodiment, in addition to autonomously parsing the user's intention for at least one round of service demand description information output by the user, the target agent can also provide preset services. The preset services do not need to autonomously plan user behavior operations, but specify service tools in advance, and implement the preset services by calling the specified service tools. Different preset services can correspond to different scenarios of users in the target application. For example, preset services may include but are not limited to: help select goods service, save money to help buy service, picture recognition help match service, refund difference and insurance service, etc. Among them, the help select goods service can call search tools and add purchase tools based on the service demand description information input by the user, select goods for the user, and add goods to the shopping cart according to user requirements. The save money to help buy service can use bargaining tools or price tracking tools for the goods specified by the user to purchase the goods specified by the user at a relative price. The picture recognition help match service can be a user input image, the image includes existing goods, and the goods in the image are matched with other goods through the set clothing matching tool. The price difference refund and price guarantee service can call the price monitoring tool to obtain the price information of the goods in the order within a set time range (such as one month) for the purchased order specified by the user. If the price drops, a price difference refund request will be initiated to the merchant to provide price guarantee service.
[0093] Based on this, the second page includes at least one service card, each card corresponds to a preset service, each service card expresses a user intention, and is associated with a designated service tool; in response to a trigger operation on any service card, the target agent is called, so that the target agent calls the designated service tool associated with any service card to perform the corresponding user behavior operation on behalf of the user; the execution result of the designated service tool is displayed on the second page; or, in response to a trigger operation on any service card, jump to the first page, and display the execution result of the designated service tool on the first page. In response to a trigger operation on any service card, it can also jump to the fourth page, and display the execution result of the designated service tool on the fourth page. In response to a trigger operation on any service card, the fourth page can be the same as the third page, or the fourth page can be different from the third page.
[0094] For example, Figure 2c A second page is displayed, the second page includes: a help product selection service card, a money saving help purchase service card, a picture recognition help matching service card, and a refund and price difference insurance service card. In response to a trigger operation on the help product selection service card, the user jumps from the second page to display a fourth page, Figure 2d is an exemplary display of the fourth page. Among them, the designated service tool corresponding to the help product selection service card can be a search tool, and the search result is multiple product information. Figure 2d In the example, multiple product information is displayed in the form of cards, and multiple product information can be displayed on the fourth page by sliding left and right. Of course, there is no need to jump from the second page to the fourth page, and the execution result of the specified service tool can be directly displayed on the second page.
[0095] In this embodiment, when the target application is implemented as an e-commerce application, an information processing method is also provided. The method is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with an artificial intelligence-based target agent. The method includes:
[0096] 101b, displaying a first page provided by the e-commerce application, where the first page includes a wake-up control for the target intelligent agent;
[0097] 102b, in response to the awakening operation on the awakening control, displaying a second page for the user to interact with the target agent;
[0098] 103b. In response to an input operation on the second page, obtaining shopping description information expressed by the user based on natural language;
[0099] 104b, calling the target agent to parse at least one user interaction operation among a search operation, a screening operation, an add-to-cart operation, and an order operation from the shopping description information, and calling at least one e-commerce tool among the search tool, the screening tool, the add-to-cart tool, and the order tool to complete the at least one user interaction operation on behalf of the user;
[0100] 105b. Displaying execution result information of at least part of the at least one e-commerce tool on the first page, the second page or the third page as purchase result information corresponding to the shopping description information.
[0101] Among them, for the detailed introduction of the above-mentioned information processing method, please refer to the aforementioned embodiment, which will not be repeated here.
[0102] The following is an exemplary description of a shopping scenario.
[0103] The first round of service demand description information is: "I want to buy a jacket of brand X, within 500 yuan";
[0104] Target Agent: "Show a list of products (category - jackets, brand - X, price - less than 500 yuan)"
[0105] User interaction operations include: search operation, search for (category - jacket, brand - X, price - less than 500 yuan) jacket;
[0106] Call the search tool to search and display the search results to the user;
[0107] The second round of service demand description information is: "I looked at it and found that they seem to be the kind of clothes for spring and autumn, not suitable for winter. I want the kind of clothes for winter, but these are too thin";
[0108] Target agent: "Show product list (category - jackets, winter warmth, brand - X, price - less than 500 yuan)"
[0109] The user interaction operation includes: filtering operation, filtering the jackets corresponding to "winter warmth" from the above search results;
[0110] Call the screening tool to perform screening and display the screening results to the user;
[0111] The third round of service demand description information is: "I took a look, please help me add the first item P to the shopping cart, and then check if there is any discount?";
[0112] Target Agent: "Add product P to the shopping cart and communicate with the merchant to determine whether there is any discount information for this product";
[0113] User interaction operations include: adding items to a cart and communicating with merchants;
[0114] Call the add-to-cart tool to obtain the user's size information, and add the first item P to the shopping cart based on the size information;
[0115] Call the communication tool to communicate with the merchant about the discount information, confirm that a 20 yuan discount is available for purchases at 8 pm, and display the discount information to the user;
[0116] The fourth round of service demand description information is: "Please place an order for me at 8 pm";
[0117] Target Agent: “Place an order for product P at 8pm”;
[0118] User interaction operations include: timing operations and order placement operations;
[0119] Call the timing tool and set it to call the order tool at 8 pm;
[0120] At 8 o'clock in the evening, the ordering tool is called to purchase item P in the shopping cart at a discounted price of 20 yuan, and the user is notified of the successful purchase.
[0121] In this embodiment, when the target application is implemented as an e-commerce application, a method for processing return information is also provided. The method is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with an artificial intelligence-based target agent; the method includes:
[0122] 101c, displaying a first page provided by the target application, where the first page includes a wake-up control for the target agent;
[0123] 102c, in response to the awakening operation on the awakening control, displaying a second page for the user to interact with the target agent;
[0124] 103c. In response to the input operation on the second page, obtaining the return description information expressed by the user based on natural language;
[0125] 104c. Invoke the target agent to parse at least one user interaction operation among order query operation, logistics information query operation, refund operation, and appointment pickup operation from the return description information, and invoke at least one e-commerce tool among order query tool, logistics information query tool, refund tool, and appointment pickup tool to complete at least one user interaction operation on behalf of the user;
[0126] 105c. Displaying the execution result information of at least part of the at least one e-commerce tool on the first page, the second page or the third page as the return result information corresponding to the return description information.
[0127] Among them, for the detailed introduction of the above-mentioned information processing method, please refer to the aforementioned embodiment, which will not be repeated here.
[0128] The following is an exemplary description of the return scenario.
[0129] The first round of service demand description information: "I want to return the product";
[0130] Target Agent: Query the orders that can be returned to see how many can be returned. If there are multiple, ask the user which one he wants to return?
[0131] User behavior operations include: order query operations;
[0132] Call the order query tool to obtain the items that can be returned. There are 3 items;
[0133] Call the order display tool to display three products to the user, and let the user choose;
[0134] The second round of service demand description information: "Select product M";
[0135] Target agent: Return product M and check the logistics status of product M to see whether it has been received.
[0136] User behavior operation: query the logistics of product M;
[0137] Call the logistics information query tool to query the logistics of product M and find that the user has received the goods;
[0138] Target Agent: Ask the user whether to get a refund or return the product.
[0139] Call the display tool to output inquiry information, which includes two options: refund only and return of goods;
[0140] The third round of service demand description information: "return and refund";
[0141] Target Agent: "Initiate a return and refund operation for product M"
[0142] User interaction behavior: refund operation;
[0143] Call the refund tool to initiate a return and refund operation for product M;
[0144] Target Agent: “Ask the user if they want to schedule a pickup”
[0145] Call the display tool to output the query information whether an appointment is needed for pickup;
[0146] The fourth round of service demand description information: "need";
[0147] Target Agent: "Schedule a pickup for the user"
[0148] User behavior operation: make an appointment to pick up the item;
[0149] Call the appointment pickup tool to obtain the merchant address corresponding to product M and the user's delivery address, make an appointment for the courier to pick up the item for the user, and output the pickup code to the user.
[0150] In the embodiment of the present application, a reasoning framework of a target agent is also provided, such as Figure 3a As shown in the figure, the framework is divided into four levels: service scenario layer, service capability layer, target agent layer and service tool layer.
[0151] The service scenario layer corresponds to the application layer. For example, the service scenario layer may include: intelligent assistants, shopping guide assistants, and writing assistants, etc. Different assistants may correspond to different pages, and different assistants may be deployed in one application or in different applications. There is no limitation on this. In addition, different assistants may correspond to different intelligent entities or the same intelligent entity. Furthermore, the service scenario layer also includes: a test and configuration platform, which may test the functions of the target intelligent entity and the service tool.
[0152] Among them, the service capability layer encapsulates the application layer configuration of the target intelligent agent and service tools. Among them, the service capability layer includes: tool selection and description module, intelligent agent configuration module and tool configuration module. The tool selection and description module mainly maintains the service tools that can be used by the intelligent assistant, and maintains the description information of the service tools. The intelligent agent configuration module maintains the correspondence between the intelligent assistant and the intelligent agent, and can further define the responsibilities, behaviors, inputs and outputs of the intelligent agent. The tool configuration module mainly maintains the basic configuration of the service tool, for example, defines the interface configuration of the service tool, the communication method between the intelligent agent and the service tool, etc.
[0153] Among them, the target agent layer completes the core reasoning and iteration. The target agent layer includes an agent reasoning service, which is mainly used to parse at least one user behavior operation corresponding to the user intent from at least one round of service demand description information. The agent reasoning service is implemented through a plan module, a tool management module, an agent management module, and a prompt word management module. Among them, the prompt word management module is mainly used to generate prompt words according to a preset template, and input the prompt words into the intent parsing model. The agent management module is mainly used to adapt different intent parsing models according to parsing requirements; the plan module is used to parse and obtain at least one user behavior operation; the tool management module is mainly used to determine which service tools implement at least one user behavior operation.
[0154] The service tool layer completes the creation and calling of service tools. The service tool layer includes a tool platform, which can realize service tool management, tool calling, and call monitoring. The tool platform includes tools corresponding to various assistants, such as shopping guide tools, transaction tools, and other service tools. In addition, the tool platform also includes a tool generalization module and a test module. Tool generalization usually refers to the design and development of general tools or components so that the tool can be reused in a variety of different scenarios. The test module is mainly used to simulate the application of tools in actual scenarios and test the performance of tools.
[0155] The above framework allows the algorithm to focus on the effect iteration of prompt words and intent parsing models. The framework can cover other tasks such as standardization, tool routing, input and output protocols, interrupt selectors, error handling, tool platforms, etc., and on this basis, it precipitates the standardized implementation capabilities of the intelligent assistant (Assistant Agent) based on tool planning and calling capabilities.
[0156] In the embodiment of the present application, the target agent can refer to the concept and idea of ReAct (reasoning and action) in the process of "parsing at least one user behavior operation corresponding to the user's intention from at least one round of service demand description information, and calling at least one service tool to complete at least one user behavior operation on behalf of the user", such as Figure 3b shown.
[0157] The target intelligent agent mainly involves multiple rounds of iterations from "parsing at least one user behavior operation corresponding to user intention from at least one round of service demand description information, and calling at least one service tool to complete at least one user behavior operation on behalf of the user". Each round of iteration includes: thinking module, action module and observation module.
[0158] Among them, the thinking module is mainly used to parse at least one user behavior operation corresponding to the user intention in each round of service demand description information; the action module is mainly used to call at least one service tool to complete at least one user behavior operation on behalf of the user; the observation module is mainly used to observe the execution results, detect whether there is new input or judge whether it is necessary to continue iteration, etc. The observation module can also provide the thinking process and action results, new user input, and observation results of the execution results to the thinking module, and the thinking module will continue the next round of thinking until there is no new action output, or the set number of iterations or the set iteration time is reached, the iteration ends, and the execution result of the action is output. The core decision logic of the target intelligent agent is to enable the intention parsing model to determine the new round of user action operations based on dynamically changing environmental information (such as new user input, the execution results of the previous round of iterations), or to make judgments on the execution results of the service tool, and further affect the environmental information of the next round of iterations, and repeat the above steps through multiple rounds of iterations until the user intention is completed. Among them, in Figure 3b In the figure, action 1 and action 2 are taken as examples for illustration, which are only examples and are not limited to them.
[0159] Among them, in the embodiment of the present application, the browsing, decision-making, purchasing, and fulfillment processes of the user in the target application are converted into service tools corresponding to the target intelligent agent, and the calling process of these service tools is learned through the intention analysis model, so that the user can control the target intelligent agent to complete the browsing, shopping, return and other user behavior operations in the target application in a simple and convenient way through language or text, thereby improving the user experience. Among them, the encapsulation of e-commerce scene applications and e-commerce tools can greatly improve the accuracy of the intelligent agent in the e-commerce scene.
[0160] The present application also provides an information processing method, which is applied to an artificial intelligence-based target agent associated with a target application, and includes:
[0161] 101d. Receive at least one round of service demand description information expressed by a user based on natural language, where at least one round of service demand description information belongs to the same user intention;
[0162] 102d. parsing at least one user behavior operation corresponding to the user's intention from at least one round of service demand description information, and calling at least one service tool to complete the at least one user behavior operation on behalf of the user; and
[0163] 103d. Provide the execution results of at least part of the at least one service tool to the target application as service result information corresponding to the user intention, so that the target application can display it.
[0164] Among them, if the target intelligent agent is deployed on the server side and the target application is deployed on the terminal device, the execution result of at least part of the service tools in at least one service tool can be output to the terminal device, and the terminal device displays the execution result as the service result information corresponding to the user's intention, for example, on the first page, the second page, or the third page. If both the target intelligent agent and the target application are deployed on the terminal device, the execution result can be directly displayed as the service result information corresponding to the user's intention, for example, on the first page, the second page, or the third page. For details, please refer to the above embodiment.
[0165] In this embodiment, the target agent may include one agent or multiple agents. Considering that the service demand description information input by the user is long, or the target agent is expected to have a short response time, in order to improve the processing speed and quality of the target agent, in an optional embodiment, the target agent includes a main agent and multiple sub-agents. The main agent and multiple sub-agents cooperate with each other to perform information processing tasks. The main agent and multiple sub-agents are like the human brain, with each area clearly divided and able to work together. Different sub-agents correspond to different application scenarios. For example, in the field of e-commerce, sub-agents can be determined according to the ideas of pre-sales, sales and after-sales. For example, sub-agents include but are not limited to: shopping guide sub-agents, merchant sub-agents, order sub-agents, payment sub-agents, etc. Each sub-agent focuses on work within its own scenario and maintains a set of independent service tools. When the sub-agents are working, the sub-agents work independently of each other, and the execution results within the sub-agents are visible to other sub-agents. Sub-agents can also use the output results of other sub-agents as input to facilitate the interaction between sub-agents. After the interaction is completed, the sub-agent calls specific e-commerce tools and execution processes to complete the corresponding sub-tasks, and finally reports the execution results to the main agent. The main agent summarizes the execution results of the sub-agents and outputs them to the user.
[0166] Accordingly, the implementation method of parsing at least one user behavior operation corresponding to user intention from at least one round of service demand description information is not limited. For example, any round of service demand description information is split into at least one sub-service demand description information through a main agent, and at least one sub-service demand description information corresponds to at least one sub-agent; at least one user behavior operation corresponding to any round of service demand description information is obtained according to the result of parsing at least one sub-service demand description information by at least one sub-agent. Accordingly, for at least one user behavior operation corresponding to any round of service demand description information, at least one service tool can be called to complete the at least one user behavior operation on behalf of the user; for any round of service demand description information, the execution result of at least part of the at least one service tool is output as the service result information of that round of service demand description information.
[0167] In the case where the target agent includes a main agent and multiple sub-agents, the overall flow of the information processing process is introduced, such as Figure 4a shown.
[0168] E1. The main agent obtains the service requirement description information input by the user.
[0169] E2. The main agent splits the service demand description information, for example, according to the scenarios corresponding to each sub-agent, to obtain sub-description information. For example, according to the scenarios of shopping guide, order, refund and after-sales, the service demand description information is split into multiple sub-description information, and the multiple sub-description information is assigned to the sub-agents of the corresponding scenarios;
[0170] E3. For each sub-agent, the sub-agent analyzes at least one user behavior operation corresponding to the user's intention through the thinking module;
[0171] E4. During the parsing process, determine whether it is necessary to rely on the output results of other sub-agents. If it is necessary to rely on the output results of other sub-agents, wait for the output results of other sub-agents;
[0172] E5. Each sub-agent calls at least one service tool through the action module to complete at least one user behavior operation on behalf of the user;
[0173] E6. Each sub-agent outputs the execution result of at least one service tool for use by other sub-agents;
[0174] E7, each sub-agent combines the execution results of other sub-agents, observes the execution results through the observation module of the sub-agent, and determines whether the above iteration process needs to be repeated;
[0175] E8. If necessary, continue to iterate and provide the execution results to the sub-agent for the next round of iteration;
[0176] E9. If not necessary, the execution result is output to the main agent;
[0177] E10. The main agent receives the output results of each sub-agent, summarizes them, and outputs the summary results to the user.
[0178] like Figure 4b As shown, the user provides at least one round of service demand description information to the main agent, and the main agent splits the at least one round of service demand description information into at least one sub-service demand description information and assigns the at least one sub-service demand description information to at least one sub-agent. Figure 3b In the figure, three sub-agents are used as an example, but the illustration is not limited to this. The 3b sub-agents are: shopping guide sub-agent, text creation sub-agent and order sub-agent. Among them, different sub-agents can correspond to the same type of intent parsing model, or they can correspond to different intent parsing models. Figure 3b In the figure, the shopping guide sub-agent, text creation sub-agent and order sub-agent correspond to different intent parsing models D1, D2 and D3 respectively. Figure 4b The "environment" in the system is a resource pool jointly maintained by the main agent and multiple sub-agents. Independent memory is the information within a sub-agent, and shared memory is the information shared by multiple sub-agents. "Communication" is the process of information exchange between sub-agents, and "reflection" is the process of "observation" within a sub-agent.
[0179] The detailed implementation methods and beneficial effects of each step in the above method provided in the embodiments of the present application have been described in detail in the aforementioned embodiments and will not be elaborated here.
[0180] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 101a to 103a can be device A; for another example, the execution subject of steps 101a and 502a can be device A, and the execution subject of step 103a can be device B; and so on.
[0181] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel, and the sequence numbers of the operations, such as 101a, 102a, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.
[0182] The exemplary embodiment of the present application provides a structural schematic diagram of an information processing device, which is applied to a terminal device installed with a target application, and the target application is associated with a target agent based on artificial intelligence; the device includes:
[0183] A display module, used to display a first page provided by a target application, wherein the first page includes a wake-up control of a target intelligent agent;
[0184] The display module is further used to respond to the wake-up operation of the wake-up control and display a second page for the user to interact with the target intelligent agent;
[0185] An acquisition module, configured to respond to an input operation on the second page and acquire at least one round of service demand description information expressed by the user based on natural language, wherein the at least one round of service demand description information belongs to the same user intention;
[0186] A calling module, configured to call a target agent to parse at least one user behavior operation corresponding to a user's intention from at least one round of service demand description information, and call at least one service tool to complete at least one user behavior operation on behalf of the user; and
[0187] The processing module is used to display the execution results of at least part of the at least one service tool on the first page, the second page or the third page as the service result information corresponding to the user intention.
[0188] In an optional embodiment, the first page is the home page of the target application, and the wake-up control is a button set in a specified area of the home page row; or the first page is a message page, and the wake-up control is a specific reminder message displayed at a specified position on the message page, and the specific reminder message supports interactive operations.
[0189] In an optional embodiment, the acquisition module is specifically used to: in response to the input operation on the second page, obtain the current round of service demand description information expressed by the user based on natural language; if a parsing exception occurs in the process of calling the target intelligent agent to parse the user behavior operation from the current round of service demand description information, output intention inquiry information to guide the user to input the next round of service demand description information to improve the user's intention.
[0190] Optionally, the acquisition module is specifically used to: when the user behavior operation cannot be parsed from the current round of service demand description information, output re-input guidance information to guide the user to re-enter the current round of service demand description information to improve the user's intention; when the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service demand description information includes multiple candidate information, output multiple candidate information for the user to select target information as the next round of service demand description information; when the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service demand description information includes question information, output question information and reply guidance information to guide the user to input answer information of the question information as the next round of service demand description information.
[0191] In an optional embodiment, the calling module is specifically used to: when the target intelligent agent is running locally on the terminal device, input at least one round of service demand description information into the target intelligent agent in sequence for intent analysis to obtain at least one user behavior operation corresponding to the user intent and its associated information object; when the target intelligent agent is running on the server, send at least one round of service demand description information to the server in sequence, so that the server can input at least one round of service demand description information into the target intelligent agent in sequence for intent analysis to obtain at least one user behavior operation corresponding to the user intent and its associated information object.
[0192] Optionally, the calling module is specifically used to: for any round of service demand description information, input any round of service demand description information into the target intelligent agent, and within the target intelligent agent, generate prompt words according to the set prompt word template format based on any round of service demand description information, tool description information and context information; input the prompt words into the intention parsing model corresponding to the target intelligent agent for intent parsing to obtain at least one user behavior operation and its associated information object.
[0193] Further optionally, the device also includes an output module; the output module is used to output the speech information corresponding to at least one user behavior operation and its associated information object during each round of parsing.
[0194] In an optional embodiment, the first page, the second page or the third page includes task list entry information, and the display module is also used to: respond to a trigger operation on the task list entry information, display a task list, the task list includes at least one task and its execution status displayed in a card format; wherein a task refers to a task that is executed by a target agent based on user intention; the execution status of the task includes ended, canceled, in progress or terminated.
[0195] Optionally, a function control is displayed in an associated area of at least one task, and tasks in different execution states correspond to different function controls, and different function operations can be triggered to be performed on the task through different function controls.
[0196] Optionally, the task list entry information includes a quantity identifier, and the quantity identifier corresponds to the number of tasks included in the current task list.
[0197] In an optional embodiment, the processing module is specifically used to: determine at least some service tools from at least one service tool based on attribute information of the service tools; load output templates corresponding to at least some of the service tools; add execution results of at least some of the service tools to the output templates corresponding to at least some of the service tools to obtain service result information corresponding to the user's intention; and display the service result information on the first page, the second page or the third page.
[0198] In an optional embodiment, the second page includes at least one service card, at least one service card expresses at least one user intention and is associated with a designated service tool; the calling module is also used to: respond to a trigger operation on any service card, call the target intelligent entity, so that the target intelligent entity calls the designated service tool associated with any service card to perform the corresponding user behavior operation on behalf of the user; the display module is also used to: display the execution result of the designated service tool on the first page, the second page or the fourth page.
[0199] The embodiment of the present application further provides an information processing device, which is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with a target intelligent agent based on artificial intelligence; the device includes:
[0200] A display module, used to display a first page provided by the e-commerce application, wherein the first page includes a wake-up control of the target intelligent agent;
[0201] The display module is further used to respond to the wake-up operation of the wake-up control and display a second page for the user to interact with the target intelligent agent;
[0202] An acquisition module, used to respond to an input operation on the second page and acquire shopping description information expressed by the user based on natural language;
[0203] A calling module, used to call the target agent to parse at least one user interaction operation among a search operation, a screening operation, an add-to-cart operation and an order operation from the shopping description information, and call at least one e-commerce tool among the search tool, the screening tool, the add-to-cart tool and the order tool to complete the corresponding user interaction operation on behalf of the user; and
[0204] The display module is also used to display the execution result information of at least part of the at least one e-commerce tool on the first page, the second page or the third page as the purchase result information corresponding to the shopping description information.
[0205] The embodiment of the present application further provides an information processing device, which is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with a target intelligent agent based on artificial intelligence; the device includes:
[0206] A display module, used to display a first page provided by a target application, wherein the first page includes a wake-up control of a target intelligent agent;
[0207] The display module is further used to respond to the wake-up operation of the wake-up control and display a second page for the user to interact with the target intelligent agent;
[0208] An acquisition module, used to respond to the input operation on the second page and acquire the return description information expressed by the user based on natural language;
[0209] A calling module is used to call the target agent to parse at least one user interaction operation among order query operation, logistics information query operation, refund operation, and appointment pickup operation from the return description information, and call at least one e-commerce tool among order query tool, logistics information query tool, refund tool, and appointment pickup tool to complete the corresponding user interaction operation on behalf of the user; and
[0210] The display module is also used to display the execution result information of at least part of the at least one e-commerce tool on the first page, the second page or the third page as the return result information corresponding to the return description information.
[0211] The embodiment of the present application further provides an information processing device, which is applied to a target agent based on artificial intelligence associated with a target application, and the device includes:
[0212] A receiving module, configured to receive at least one round of service demand description information expressed by a user based on natural language, wherein at least one round of service demand description information belongs to the same user intention;
[0213] A parsing module, used for parsing at least one user behavior operation corresponding to the user intention from at least one round of service demand description information;
[0214] A calling module, used to call at least one service tool to complete at least one user behavior operation on behalf of the user;
[0215] The sending module is used to provide the execution results of at least part of the at least one service tool as service result information corresponding to the user intention to the target application for display by the target application.
[0216] Optionally, the target agent includes a main agent and multiple sub-agents, and different sub-agents correspond to different application scenarios; the parsing module is specifically used to: split at least one round of service demand description information into at least one sub-service demand description information through the main agent, and at least one sub-service demand description information corresponds to at least one agent; according to the results of parsing at least one sub-service demand description information by at least one sub-agent, at least one user behavior operation is obtained, wherein the input of at least some sub-agents depends on the output results of other sub-agents.
[0217] The detailed implementation and beneficial effects of each step in the above-mentioned device provided in the embodiments of the present application have been described in detail in the aforementioned embodiments and will not be elaborated here.
[0218] Figure 5 A schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present application, the device is applied to a terminal device installed with a target application, the target application is associated with a target agent based on artificial intelligence, such as Figure 5 As shown, the device includes: a memory 54 and a processor 55.
[0219] The memory 54 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method used to operate on the electronic device.
[0220] The processor 55 is coupled to the memory 54 and is used to execute the computer program in the memory 54 to: display a first page provided by the target application, the first page including a wake-up control of the target intelligent agent; in response to a wake-up operation on the wake-up control, display a second page for the user to interact with the target intelligent agent; in response to an input operation on the second page, obtain at least one round of service demand description information expressed by the user based on natural language, at least one round of service demand description information belongs to the same user intention; call the target intelligent agent to parse at least one user behavior operation corresponding to the user intention from at least one round of service demand description information, and call at least one service tool to complete at least one user behavior operation on behalf of the user; and display the execution results of at least part of the at least one service tool on the first page, the second page or the third page as service result information corresponding to the user intention.
[0221] In an optional embodiment, the first page is the home page of the target application, and the wake-up control is a button set in a specified area of the home page row; or the first page is a message page, and the wake-up control is a specific reminder message displayed at a specified position on the message page, and the specific reminder message supports interactive operations.
[0222] In an optional embodiment, when the processor 55 responds to the input operation on the second page and obtains at least one round of service demand description information expressed by the user based on natural language, it is specifically used to: respond to the input operation on the second page, obtain the current round of service demand description information expressed by the user based on natural language; if a parsing exception occurs in the process of calling the target intelligent agent to parse the user behavior operation from the current round of service demand description information, output intention inquiry information to guide the user to input the next round of service demand description information to improve the user's intention.
[0223] Optionally, if a parsing exception occurs during the process of calling the target intelligent agent to parse the user behavior operation from the current round of service demand description information, the processor 55 outputs intention inquiry information to guide the user to input the next round of service demand description information. Specifically, it is used to: when the user behavior operation cannot be parsed from the current round of service demand description information, output re-input guidance information to guide the user to re-enter the current round of service demand description information to improve the user's intention; when the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service demand description information includes multiple candidate information, output multiple candidate information for the user to select target information as the next round of service demand description information; when the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service demand description information includes question information, output question information and reply guidance information to guide the user to input answer information of the question information as the next round of service demand description information.
[0224] In an optional embodiment, when the processor 55 calls the target agent to parse at least one user behavior operation corresponding to the user intention from at least one round of service requirement description information, it is specifically used to: when the target agent runs locally on the terminal device, input at least one round of service requirement description information into the target agent in sequence for intent parsing to obtain at least one user behavior operation corresponding to the user intention and its associated information object; when the target agent runs on the server, send at least one round of service requirement description information to the server in sequence, so that the server can input at least one round of service requirement description information into the target agent in sequence for intent parsing to obtain at least one user behavior operation corresponding to the user intention and its associated information object.
[0225] Optionally, when the processor 55 sequentially inputs at least one round of service demand description information into the target intelligent agent for intent analysis to obtain at least one user behavior operation corresponding to the user intent and its associated information object, it is specifically used to: for any round of service demand description information, input any round of service demand description information into the target intelligent agent, and within the target intelligent agent, generate prompt words according to a set prompt word template format based on any round of service demand description information, tool description information and context information; input the prompt words into the intent analysis model corresponding to the target intelligent agent for intent analysis to obtain at least one user behavior operation and its associated information object.
[0226] Optionally, the processor 55 is further used to: during each round of parsing, output the speech information corresponding to at least one user behavior operation and its associated information object.
[0227] In an optional embodiment, the first page, the second page or the third page includes task list entry information, and the processor 55 is further used to: respond to a trigger operation on the task list entry information, display a task list, the task list includes at least one task and its execution status displayed in a card format; wherein a task refers to a task that is executed by a target agent based on user intention; the execution status of the task includes ended, canceled, in progress or terminated.
[0228] Optionally, a function control is displayed in an associated area of at least one task, and tasks in different execution states correspond to different function controls, and different function operations can be triggered to be performed on the task through different function controls.
[0229] Optionally, the task list entry information includes a quantity identifier, and the quantity identifier corresponds to the number of tasks included in the current task list.
[0230] In an optional embodiment, when the processor 55 displays the execution results of at least part of the at least one service tool as the service result information corresponding to the user intention, it is specifically used to: determine at least part of the service tools from the at least one service tool based on the attribute information of the service tools; load the output templates corresponding to at least part of the service tools; add the execution results of at least part of the service tools to the output templates corresponding to at least part of the service tools, respectively, to obtain the service result information corresponding to the user intention; and display the service result information on the first page, the second page or the third page.
[0231] In an optional embodiment, the second page includes at least one service card, which expresses at least one user intention and is associated with a designated service tool; the processor 55 is also used to: respond to a trigger operation on any service card, call the target intelligent entity, so that the target intelligent entity calls the designated service tool associated with any service card to perform the corresponding user behavior operation on behalf of the user; and display the execution result of the designated service tool on the first page, the second page or the fourth page.
[0232] Regarding the embodiments of the present application Figure 5 The detailed implementation and beneficial effects of each step in the illustrated device have been described in detail in the aforementioned embodiments and will not be elaborated on here.
[0233] Further, if Figure 5 As shown, the electronic device also includes: a communication component 56, a display 57, a power component 58, an audio component 59 and other components. Figure 5 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 5 In addition, Figure 5 The components in the dashed box are optional components, not mandatory components, and the specific components depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, or an IOT device, or a server device such as a conventional server, a cloud server, or a server array. If the electronic device of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, etc., it can include Figure 5 If the electronic device of this embodiment is implemented as a server device such as a conventional server, a cloud server or a server array, it may not include Figure 5 Components within the dashed box.
[0234] The embodiment of the present application also provides an electronic device, which is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with a target intelligent agent based on artificial intelligence; the implementation structure of the electronic device is similar to Figure 5 The implementation structure of the electronic device shown is the same or similar, and can refer to Figure 5 The electronic device provided in this embodiment is implemented as follows: Figure 5The difference between the electronic devices in the illustrated embodiments mainly lies in that the functions implemented by the processor executing the computer program stored in the memory are different. For the electronic device provided in this embodiment, its processor executes the computer program stored in the memory, which can be used to: display a first page provided by the e-commerce application, the first page includes a wake-up control of the target intelligent agent; respond to the wake-up operation of the wake-up control, display a second page for the user to interact with the target intelligent agent; respond to the input operation on the second page, obtain the shopping description information expressed by the user based on natural language; call the target intelligent agent to parse at least one user interaction operation among the search operation, the screening operation, the add-to-cart operation and the order operation from the shopping description information, and call at least one e-commerce tool among the search tool, the screening tool, the add-to-cart tool and the order tool to complete the corresponding user interaction operation on behalf of the user; and display the execution result information of at least part of the at least one e-commerce tool on the first page, the second page or the third page as the purchase result information corresponding to the shopping description information.
[0235] The embodiment of the present application also provides an electronic device, which is applied to a terminal device installed with an e-commerce application, and the e-commerce application is associated with a target intelligent agent based on artificial intelligence; the implementation structure of the electronic device is similar to Figure 5 The implementation structure of the electronic device shown is the same or similar, and can refer to Figure 5 The electronic device provided in this embodiment is implemented as follows: Figure 5 The difference between the electronic devices in the illustrated embodiments mainly lies in that the functions implemented by the processor executing the computer program stored in the memory are different. For the electronic device provided in this embodiment, its processor executes the computer program stored in the memory, which can be used to: display the first page provided by the target application, the first page includes the wake-up control of the target intelligent body; respond to the wake-up operation of the wake-up control, display the second page for the user to interact with the target intelligent body; respond to the input operation on the second page, and obtain the return description information expressed by the user based on natural language; call the target intelligent body to parse at least one user interaction operation among the order query operation, logistics information query operation, refund operation, and appointment pickup operation from the return description information, and call at least one e-commerce tool among the order query tool, logistics information query tool, refund tool, and appointment pickup tool to replace the user to complete the corresponding user interaction operation; and display the execution result information of at least part of the at least one e-commerce tool on the first page, the second page, or the third page as the return result information corresponding to the return description information.
[0236] The embodiment of the present application also provides an electronic device, which is applied to a target agent based on artificial intelligence associated with a target application; the implementation structure of the electronic device is similar to Figure 5 The implementation structure of the electronic device shown is the same or similar, and can refer to Figure 5 The electronic device provided in this embodiment is implemented as follows: Figure 5 The difference between the electronic devices in the illustrated embodiments mainly lies in that the functions implemented by the processor executing the computer program stored in the memory are different. For the electronic device provided in this embodiment, its processor executes the computer program stored in the memory, which can be used to: receive at least one round of service demand description information expressed by the user based on natural language, and at least one round of service demand description information belongs to the same user intention; parse at least one user behavior operation corresponding to the user intention from at least one round of service demand description information, and call at least one service tool to complete at least one user behavior operation on behalf of the user; and provide the execution results of at least part of the at least one service tool as service result information corresponding to the user intention to the target application for display by the target application.
[0237] In an optional embodiment, the target agent includes a main agent and multiple sub-agents, and different sub-agents correspond to different application scenarios; when the processor parses at least one user behavior operation corresponding to the user intention from at least one round of service demand description information, it is specifically used to: split at least one round of service demand description information into at least one sub-service demand description information through the main agent, and at least one sub-service demand description information corresponds to at least one agent; according to the results of parsing at least one sub-service demand description information by at least one sub-agent, at least one user behavior operation is obtained, wherein the input of at least some sub-agents depends on the output results of other sub-agents.
[0238] The detailed implementation and beneficial effects of the electronic device provided in the embodiments of the present application have been described in detail in the aforementioned embodiments and will not be elaborated here.
[0239] Accordingly, an embodiment of the present application also provides a computer program product, including a computer program / instruction. When the computer program / instruction is executed by a processor, the processor is caused to implement each step in the above method embodiment that can be executed by an electronic device.
[0240] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, enables a processor to implement each step that can be executed by an electronic device in the above method embodiment.
[0241] The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.
[0242] The above-mentioned communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (Ultra Wide Band, UWB) technology, Bluetooth (BlueTooth, BT) technology and other technologies.
[0243] The above-mentioned display includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.
[0244] The power supply assembly provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply assembly is located.
[0245] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (Microphone, MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.
[0246] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, compact disk read-only memory (Compact Disk Read-Only Memory, CD-ROM), optical storage, etc.) that contain computer-usable program code.
[0247] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0248] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0249] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0250] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input / output interface, network interface and memory.
[0251] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.
[0252] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (Phase-change Random Access Memory, PRAM), static random access memory (SRAM), dynamic random access memory (Dynamic Random Access Memory, DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory, etc.
[0253] (EEPROM), flash memory or other memory technology, CD-ROM, Digital Versatile Disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.
[0254] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0255] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.
Claims
1. An information processing method, characterized in that: The method is applied to a terminal device installed with a target application, wherein the target application is associated with a target agent based on artificial intelligence; the method comprises: Displaying a first page provided by the target application, wherein the first page includes a wake-up control of the target intelligent agent; In response to the wake-up operation on the wake-up control, displaying a second page for the user to interact with the target agent; In response to an input operation on the second page, obtaining at least one round of service demand description information expressed by the user based on natural language, wherein the at least one round of service demand description information belongs to the same user intention; Invoking the target agent to parse at least one user behavior operation corresponding to the user intention from the at least one round of service demand description information, and invoking at least one service tool to complete the at least one user behavior operation on behalf of the user; and The execution results of at least part of the at least one service tool are displayed on the first page, the second page or the third page as service result information corresponding to the user intention.
2. The method according to claim 1, characterized in that The first page is the home page of the target application, and the wake-up control is a button set in a designated area of the home page; or The first page is a message page, the wake-up control is a specific reminder message displayed at a specified position on the message page, and the specific reminder message supports interactive operations.
3. The method according to claim 1, characterized in that: In response to the input operation on the second page, obtaining at least one round of service demand description information expressed by the user based on natural language, including: In response to an input operation on the second page, obtaining description information of a current round of service requirements expressed by the user based on natural language; If a parsing anomaly occurs in the process of calling the target agent to parse the user behavior operation from the current round of service demand description information, output intention query information to guide the user to input the next round of service demand description information to improve the user intention.
4. The method according to claim 3, characterized in that If a parsing exception occurs in the process of calling the target agent to parse the user behavior operation from the current round of service demand description information, outputting intention query information to guide the user to input the next round of service demand description information, including: In the case that the user behavior operation cannot be parsed from the current round of service demand description information, outputting re-input guidance information to guide the user to re-input the current round of service demand description information to improve the user intention; When the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service demand description information includes multiple candidate information, output the multiple candidate information for the user to select target information as the next round of service demand description information; When the execution result of the service tool corresponding to the user behavior operation parsed from the current round of service requirement description information includes question information, the question information and reply guidance information are output to guide the user to input the answer information of the question information as the next round of service requirement description information.
5. The method according to claim 1, characterized in that: Calling the target agent to parse at least one user behavior operation corresponding to the user intention from the at least one round of service requirement description information includes: When the target agent is running locally on the terminal device, the at least one round of service requirement description information is sequentially input into the target agent for intent parsing to obtain at least one user behavior operation corresponding to the user intent and its associated information object; When the target intelligent agent is running on the server, the at least one round of service demand description information is sent to the server in sequence, so that the server can input the at least one round of service demand description information into the target intelligent agent in sequence for intent analysis, so as to obtain at least one user behavior operation corresponding to the user intent and its associated information object.
6. The method according to claim 5, characterized in that The at least one round of service requirement description information is sequentially input into the target agent for intent analysis to obtain at least one user behavior operation corresponding to the user intent and its associated information object, including: For any round of service demand description information, the said round of service demand description information is input into the said target intelligent body, and within the said target intelligent body, prompt words are generated according to the set prompt word template format based on the said round of service demand description information, the description information of the tool and the context information; the prompt words are input into the intention parsing model corresponding to the said target intelligent body for intention parsing, so as to obtain at least one user behavior operation and its associated information object.
7. The method according to claim 6, characterized in that Also includes: In each round of parsing, the speech information corresponding to the at least one user behavior operation and its associated information object is output.
8. The method according to claim 1, characterized in that The first page, the second page or the third page includes task list entry information, and the method further includes: In response to a triggering operation on the task list entry information, displaying a task list, wherein the task list includes at least one task and its execution status displayed in a card manner; The task refers to a task that is executed by calling the target agent based on the user's intention; the execution status of the task includes ended, canceled, executing or terminated.
9. The method according to claim 8, characterized in that A function control is displayed in the associated area of the at least one task, and tasks in different execution states correspond to different function controls, and different function operations can be triggered to be performed on the task through different function controls.
10. The method according to claim 1, characterized in that Displaying the execution results of at least part of the at least one service tool on the first page, the second page, or the third page as service result information corresponding to the user intention includes: determining at least some of the service tools from the at least one service tool according to the attribute information of the service tool; Loading output templates corresponding to at least some of the service tools; Adding the execution results of at least part of the service tools to their corresponding output templates respectively, to obtain service result information corresponding to the user intention; The service result information is displayed on the first page, the second page or the third page.
11. The method according to claim 1, characterized in that: The second page includes at least one service card, wherein the at least one service card expresses at least one user intention and is associated with a specified service tool; The method further includes: in response to a triggering operation on any service card, calling the target agent, so that the target agent calls a designated service tool associated with any service card to perform a corresponding user behavior operation on behalf of the user; The execution result of the designated service tool is displayed on the first page, the second page or the fourth page.
12. An information processing method, characterized in that: Applied to a terminal device installed with an e-commerce application, wherein the e-commerce application is associated with an artificial intelligence-based target agent; the method comprises: Displaying a first page provided by the e-commerce application, wherein the first page includes a wake-up control for the target intelligent entity; In response to the wake-up operation on the wake-up control, displaying a second page for the user to interact with the target agent; In response to an input operation on the second page, obtaining shopping description information expressed by the user based on natural language; Calling the target agent to parse at least one user interaction operation among a search operation, a screening operation, an add-to-cart operation, and an order operation from the shopping description information, and calling at least one e-commerce tool among the search tool, the screening tool, the add-to-cart tool, and the order tool to complete the corresponding user interaction operation on behalf of the user; and The execution result information of at least part of the at least one e-commerce tool is displayed on the first page, the second page or the third page as the purchase result information corresponding to the shopping description information.
13. An information processing method, characterized in that: Applied to a terminal device installed with an e-commerce application, wherein the e-commerce application is associated with an artificial intelligence-based target agent; the method comprises: Displaying a first page provided by the target application, wherein the first page includes a wake-up control of the target intelligent agent; In response to the wake-up operation on the wake-up control, displaying a second page for the user to interact with the target agent; In response to an input operation on the second page, obtaining return description information expressed by the user based on natural language; Calling the target agent to parse at least one user interaction operation among an order query operation, a logistics information query operation, a refund operation, and a scheduled pickup operation from the return description information, and calling at least one e-commerce tool among an order query tool, a logistics information query tool, a refund tool, and a scheduled pickup tool to complete the corresponding user interaction operation on behalf of the user; and The execution result information of at least part of the at least one e-commerce tool is displayed on the first page, the second page or the third page as the return result information corresponding to the return description information.
14. An information processing method, characterized in that: An artificial intelligence-based target agent applied to a target application association, the method comprising: receiving at least one round of service demand description information expressed by a user based on natural language, wherein the at least one round of service demand description information belongs to the same user intention; parsing at least one user behavior operation corresponding to the user intention from the at least one round of service demand description information, and calling at least one service tool to complete the at least one user behavior operation on behalf of the user; and The execution results of at least part of the at least one service tool are provided to the target application as service result information corresponding to the user intention, so as to be displayed by the target application.
15. The method according to claim 14, characterized in that The target agent includes a main agent and a plurality of sub-agents; Parsing at least one user behavior operation corresponding to the user intention from the at least one round of service requirement description information includes: Splitting any round of service demand description information into at least one sub-service demand description information through the main agent, wherein the at least one sub-service demand description information corresponds to at least one sub-agent; According to the results of the at least one sub-agent respectively parsing the at least one sub-service demand description information, at least one user behavior operation corresponding to any round of service demand description information is obtained, wherein the input of at least some sub-agents depends on the output results of other sub-agents.
16. An electronic device, characterized in that: include: Memory and processor; The memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program to implement the steps in the method of any one of claims 1-11, claims 12-13 and claims 14-15.
17. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to implement the steps of the method according to any one of claims 1 to 11, claims 12 to 13, and claims 14 to 15.
18. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the processor is caused to implement the steps in the method of any one of claims 1-11, claims 12-13, and claims 14-15.
Citation Information
Cited By
Bill service processing method and device based on intelligent agent
CN119917296A
Agent-based bill service processing method and device
CN119917296B
Data processing method, system and device, electronic equipment and computer storage medium
CN120711067A
Artificial intelligence-based itinerary planning method and computer program
CN121116447A
Information interaction method, electronic equipment and system
CN122022957A
