Interactive response method, electronic device and computer program product

Abstract

The invention provides an interactive response method, electronic equipment and a computer program product.The method comprises the steps that intention recognition is conducted on user input of a user side, and a control used for triggering display of a dialog box receiving the user input in the user side is any one of multiple specified controls preset in multiple service pages; a target agent suitable for the identified intention is determined from the multiple selectable agents, the different selectable agents are configured at different service ends, the different service ends are used for executing different types of service functions, and the different types of service interfaces correspond to the different types of service ends; reply content corresponding to user input is generated through the target agent; and generating reply data displayed in the dialog box based on the reply content.

Description

Technical Field

[0001] The present disclosure relates to technical fields such as communications, and more particularly to an interactive response method, an electronic device, and a computer program product. Background Art

[0002] As business needs continue to diversify, traditional single agents like simple chatbots or recommendation engines are no longer able to cope with more complex scenarios. For example, when an enterprise provides multiple business capabilities to users, different business units are required to independently develop and implement the same business functions within their respective sections or pages. This leads to duplicated functionality, resulting in low development efficiency and high maintenance costs. Furthermore, the lack of a unified implementation approach across different business units makes unified management difficult. Summary of the Invention

[0003] The present disclosure provides an interactive response method, an electronic device, a readable storage medium, and a computer program product.

[0004] The first aspect of the present disclosure proposes an interactive response method, including: identifying the intention of user input on a user terminal, wherein a control in the user terminal for triggering the display of a dialog box for receiving the user input is any one of a plurality of designated controls preset on a plurality of business pages; determining a target intelligent agent to which the identified intention is applicable from a plurality of optional intelligent agents, different optional intelligent agents are configured at different business terminals, different business terminals are used to perform different types of business functions, and different types of business interfaces correspond to different types of business terminals; generating reply content corresponding to the user input through the target intelligent agent; and generating reply data displayed in the dialog box based on the reply content.

[0005] According to some embodiments of the present disclosure, the method further includes: obtaining the ability to use the optional agent based on the agent configuration information sent by the service end.

[0006] According to some embodiments of the present disclosure, the ability to use an optional intelligent agent is obtained based on the intelligent agent configuration information sent by the business end, including: registering the intelligent agent interface according to the intelligent agent configuration information sent by the business end, and obtaining the ability to use the registered intelligent agent, and the intelligent agent becomes an optional intelligent agent after registration.

[0007] According to some embodiments of the present disclosure, performing intent recognition on user input at a user end includes: inputting the user input at the user end and intent information of predefined available intents into a language model, and obtaining the recognized intent output by the language model.

[0008] According to some embodiments of the present disclosure, the intent information includes an intent name and an intent description.

[0009] According to some embodiments of the present disclosure, intention recognition is performed on user input at a user terminal, including: content review of the user input at the user terminal; and if the content review passes, intention recognition is performed on the user input, otherwise specified content indicating that the answer cannot be given is directly replied to the user terminal.

[0010] According to some embodiments of the present disclosure, after obtaining the identified intent output by the language model, the method further includes: establishing a mapping relationship between the identified intent and the corresponding user input and caching it locally.

[0011] According to some embodiments of the present disclosure, user input on the user side and intent information of predefined available intents are input into a language model, including: performing a similarity comparison between the user input in the mapping relationship cached locally and the user input of the intent to be identified; and if there is a target similarity that meets the similarity requirements in the similarities obtained through the comparison, then the intent in the mapping relationship corresponding to the target similarity is used as the corresponding identified intent, otherwise the user input on the user side and intent information of predefined available intents are input into the language model.

[0012] According to some embodiments of the present disclosure, the method further includes: creating an available intent according to intent configuration information sent from the business end.

[0013] According to some embodiments of the present disclosure, the intent configuration information includes one or more information of the intent name, intent description, and information of the intelligent agent to which the intent applies.

[0014] According to some embodiments of the present disclosure, a target agent to which the identified intention is applicable is determined from among optional agents, specifically by taking the optional agent to which the identified intention is applicable as the target agent.

[0015] According to some embodiments of the present disclosure, the target intelligent agent generates reply content corresponding to the user input, including: sending relevant information of the user corresponding to the user end to the business end according to the request sent by the business end, the relevant information including one or more of the user location, historical conversations with the user, and historical consultation information of the user; and receiving return data from the business end, generating reply content corresponding to the user input based on the return data, the return data being obtained by the target intelligent agent performing natural language processing on the relevant information and the user input.

[0016] According to some embodiments of the present disclosure, the return data is a model output obtained by the target agent inputting the relevant information, the user input and the relevant business data into a language model.

[0017] According to some embodiments of the present disclosure, generating reply content corresponding to the user input based on the return data includes: encapsulating the return data of the business end into a specified format to obtain reply content corresponding to the user input, and the return data includes one or more of text, image, voice, and video.

[0018] According to some embodiments of the present disclosure, reply data displayed in the dialog box is generated based on the reply content, including: performing a content review on the reply content; and if the content review passes, generating reply data displayed in the dialog box, otherwise directly displaying the specified content indicating that the answer cannot be given in the dialog box.

[0019] According to some embodiments of the present disclosure, the method further includes: recording the conversation between the user terminal and the central control platform; and testing the updated intelligent agent based on the conversation.

[0020] A second aspect of the present disclosure proposes an electronic device, comprising: a memory storing execution instructions; and a processor executing the execution instructions stored in the memory, so that the processor executes the method described in any one of the above embodiments.

[0021] A third aspect of the present disclosure provides a readable storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor, it is used to implement the method described in any of the above embodiments.

[0022] A fourth aspect of the present disclosure provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it is used to implement the method described in any of the above embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings illustrate exemplary embodiments of the present disclosure and together with the description serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0024] Figure 1 A schematic diagram of an application scenario of the interactive response method according to some embodiments of the present disclosure is shown.

[0025] Figure 2-Figure 8 A schematic diagram of the overall flow of the interactive response method M100 according to some embodiments of the present disclosure is shown.

[0026] Figure 9 An interactive schematic diagram of an interactive response method according to some embodiments of the present disclosure is shown.

[0027] Figure 10A schematic diagram of a system architecture of some embodiments of the present disclosure is shown.

[0028] Figure 11 It is a schematic block diagram of the structure of an interactive response device according to an embodiment of the present disclosure.

[0029] Figure 12 1 is a schematic block diagram of the structure of an electronic device 1000 according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] The present disclosure will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to explain the relevant content and are not intended to limit the present disclosure. It should also be noted that, for ease of description, only the portions relevant to the present disclosure are shown in the accompanying drawings.

[0031] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present disclosure can be combined with each other. The technical solution of the present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0032] Unless otherwise stated, the exemplary embodiments / examples shown are to be understood as providing exemplary features of various details of some ways in which the technical concepts of the present disclosure can be implemented in practice. Therefore, unless otherwise stated, the features of the various embodiments / examples may be further combined, separated, interchanged, and / or rearranged without departing from the technical concepts of the present disclosure.

[0033] The terms used herein are for the purpose of describing specific embodiments and are not restrictive. As used herein, unless the context clearly indicates otherwise, the singular forms "a (kind, one)" and "the (the)" are also intended to include plural forms. In addition, when the terms "comprise" and / or "include" and their variations are used in this specification, the description indicates the presence of the stated features, wholes, steps, operations, parts, components and / or their groups, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, parts, components and / or their groups. It should also be noted that, as used herein, the terms "substantially", "approximately" and other similar terms are used as approximate terms and not as degree terms, so that they are used to explain the inherent deviations of the measured values, calculated values ​​and / or provided values ​​that will be recognized by those of ordinary skill in the art.

[0034] Taking the house-hunting APP (application) as an example, the APP includes different pages, such as the homepage, various channel pages and sub-pages under the channel pages. These pages may be set up with AI customer service (artificial intelligence customer service, hereinafter referred to as customer service). Users can enter the dialog box with customer service by clicking buttons or other methods on different pages, so that they can communicate with customer service, for example, they can ask customer service questions about housing availability, housing purchase policies, etc.

[0035] Because different sections and channels belong to different businesses, they are operated and maintained by different parties. For example, the homepage, new homes channel, pre-owned homes channel, and rental channel are each managed by different parties. When a user browses listings in a particular channel, they may inquire about information in other channels in the dialog box associated with that channel.

[0036] For example, when a user opens a dialog box on the homepage, they might ask a question about any section or channel. Therefore, the business unit responsible for maintaining the homepage needs to develop and configure the business logic for all channels to handle various user inquiries. Another example is when a user browses a subpage in the new home channel and opens a dialog box to ask about pre-owned home prices.

[0037] To meet cross-channel information provision requirements, each business entity, in addition to configuring business logic for its own business functions, must also independently develop business logic for other business functions, enabling it to answer different business questions posed by users in its own dialog box. The business logic between different business entities is maintained and managed independently by each business entity, making it non-universal. This requires each business entity to independently develop and configure multiple business functions, leading to duplicate development. This results in low development efficiency, a high development workload, and high maintenance costs for business functions. The multiple business functions implemented by different business entities lack a unified implementation approach, making unified management difficult.

[0038] To this end, the present disclosure proposes an interactive response method.

[0039] Figure 1 The following is a schematic diagram showing an application scenario of the interactive response method of some embodiments of the present disclosure. In this application scenario, a user terminal 10, a central control platform 20 and multiple service terminals may be included. Figure 1 Two business terminals, namely, the business terminal 31 and the business terminal 32, are shown in the figure. These two business terminals are only examples, and the present disclosure does not limit the number of business terminals.

[0040] The user terminal 10 can communicate with each service terminal through the central control platform 20 to send and receive data or instructions. In the present disclosure, the user terminal 10, the central control platform 20 and each service terminal respectively include at least one processor and at least one memory.

[0041] For example, the user terminal 10 can be a terminal device held by the user, and the central control platform 20 and various business terminals can be electronic devices of the enterprise. The central control platform 20 can identify the intent of the question entered by the user in the dialog box. Then, from the available agents provided by each business terminal, it identifies the agent corresponding to the identified intent. The identified agent searches and processes the user's question, generating a reply to the user's question. After packaging, it sends it to the central control platform 20, which then responds to the user in the dialog box with the packaged message.

[0042] The intelligent agents in this disclosure can be business components driven by large language models (e.g., various artificial intelligence models). These agents can perceive user input and generate responses tailored to specific business functions. These agents can be centrally managed and scheduled through a central control platform. For example, these agents could be new home consultation agents, pre-owned home consultation agents, rental recommendation agents, and renovation consultation agents.

[0043] exist Figure 1 The shapes and structures of the user terminal 10, the central control platform 20, and the service terminal shown in the figure should not be understood as limiting the scope of protection of this disclosure. In this disclosure, "terminal devices" can be different types of electronic devices, for example, terminal devices can be mobile phones, tablet computers, laptop computers, desktop computers, etc. In addition, the central control platform 20 and each service terminal can each be configured with a server, which can be a physical server or a cloud server. This disclosure does not limit the type of server.

[0044] Figure 2 FIG. 1 shows a schematic diagram of the overall flow of the interactive response method M100 according to some embodiments of the present disclosure. Figure 2 The method shown includes steps S110, S120, S130, and S140. The method can be executed by an electronic device such as a computer. The method is applied to a central control platform of a business party.

[0045] S110: Recognize the intent of a user input (e.g., a question about a second-hand house) at the user end. The control on the user end used to trigger the display of a dialog box for receiving the user input is any one of a plurality of designated controls pre-set on various business pages.

[0046] When users browse an app (application), they can open dialog boxes through portals set on multiple pages of the app, allowing them to chat with an AI customer service agent (AI assistant) and ask questions or inquire about content they want to learn about. The portal can be a designated button or other type of designated control, and the dialog box is opened by clicking the button or other operation.

[0047] The APP is equipped with multiple business pages. For example, for a house-hunting APP, the homepage can have entrances to multiple sections (channels) such as second-hand houses, new houses, rentals, and decoration, which are used to enter the main page of the section. The homepage and the main page of each section can each serve as a business page.

[0048] In some implementations, the main page of each section also has multiple levels of sub-pages. These sub-pages can be opened by operating the main page of the section. These sub-pages are also divided into categories, for example, they can include map query pages, property details pages, etc. The home page, the main page of each section, and each type of page under each section can also each be used as a business page.

[0049] Each of the above business pages may be provided with a designated control for opening a dialog box, for example, each different page may be provided with a designated control; or some pages may be provided with a designated control, but the above some pages may include multiple business pages.

[0050] When users browse different pages, they can click on the preset designated control on any page with a designated control to open a dialog box, then enter the question they want to ask, and click Send after completing the input. The APP will then send the content in the dialog box as user input to the central control platform.

[0051] All user inputs in the dialog boxes opened by the specified controls will be sent to the central control platform for unified intent recognition. For example, at time T1, user U opens a dialog box on a page under the new house section and enters question q1. The user end will use question q1 as user input and send it to the central control platform for intent recognition, and then respond to q1 through subsequent steps. Later, at time T2, user U opens a dialog box on a page under the second-hand house section and enters question q2. The user end will use question q2 as user input and send it to the central control platform for intent recognition, and then respond to q2 through subsequent steps. The specific method of intent recognition can be semantic recognition of user input, and the recognized intent can be in text form.

[0052] S120, determining a target agent to which the identified intention is applicable from a plurality of optional agents, wherein different optional agents are configured at different business ends, and different business ends are used to execute different types of business functions.

[0053] Different business ends can correspond to different types of business functions. For example, business ends can be divided into second-hand housing business end, new housing business end, rental business end, decoration business end, etc., or they can be classified according to other classification methods.

[0054] In some embodiments of the present disclosure, different types of business interfaces correspond to different types of business terminals. For example, the pre-owned housing business interface of a house-hunting app corresponds to the pre-owned housing business terminal, meaning that when a user operates on the pre-owned housing business interface, the pre-owned housing business terminal responds; and the new housing business interface corresponds to the new housing business terminal, meaning that when a user operates on the new housing business interface, the new housing business terminal responds.

[0055] The business end can be a software with business processing functions, or a software module with business processing functions.

[0056] For example, the new home business end and the second-hand home business end can be a software (new home business software, second-hand home business software) respectively, or they can be two software modules of the same real estate business software (new home business module, second-hand home business module).

[0057] Exemplarily, the new house consultation agent is configured at the new house business end, the second-hand house consultation agent is configured at the second-hand house business end, the rental recommendation agent is configured at the rental business end, and the decoration consultation agent is configured at the decoration business end.

[0058] Agents are used to perform business logic processing on user input and generate responses. The selectable agents are those currently available on the central control platform. Different user intents correspond to different business needs, requiring responses from the corresponding business-side agents.

[0059] For example, the new house business end is configured with an intelligent agent whose business function is to inquire about new house market conditions. If the intention y1 of question q1 is identified as "new house market conditions", then question q1 needs to be answered by the intelligent agent configured on the new house business end, and specifically by the intelligent agent z1 whose business function is to inquire about new house market conditions. That is, at this time, z1 is the target intelligent agent applicable to q1.

[0060] Since the intelligent agents configured at each business end and their corresponding business functions are known, the text similarity between the text corresponding to the identified intent and the introduction text of the business functions of each optional intelligent agent can be determined through similarity calculation and other methods, and the intelligent agent corresponding to the introduction text with the highest similarity to the intent text will be used as the target intelligent agent.

[0061] Therefore, each business end can only develop and configure intelligent agents related to its own business functions, without having to develop and configure intelligent agents used to implement the business functions of other business ends. The intelligent agents developed by each business end are optional intelligent agents, and the central control platform uniformly determines the target intelligent agent from them.

[0062] In a preferred embodiment of the present disclosure, the user terminal can obtain, based on a dialog box of any business page of the user terminal, a reply content provided by a target agent applicable to the intention input by the user via the dialog box.

[0063] For example, if a user enters a second-hand housing consultation question (i.e., user input) in the dialog box of the second-hand housing business interface on the user side, the reply content provided by the second-hand housing consultation intelligent agent can be obtained based on the dialog box of the second-hand housing business interface; if a user enters a new housing consultation question (i.e., user input) in the dialog box of the second-hand housing business interface on the user side, the reply content provided by the new housing consultation intelligent agent can be obtained based on the dialog box of the second-hand housing business interface.

[0064] In some embodiments of the present disclosure, different business terminals are configured on different servers, for example, a new house business terminal and a second-hand house business terminal are configured on two different servers.

[0065] In other embodiments of the present disclosure, different business terminals are configured on the same server, for example, a new house business terminal and a second-hand house business terminal are configured on the same server.

[0066] In light of the technical solution disclosed herein, those skilled in the art may select or adjust the configuration methods of different service terminals, which all fall within the scope of protection of the present disclosure.

[0067] S130, generating a response content corresponding to the user input through the target agent.

[0068] An agent is an agent that can perceive its environment and take actions to achieve specific goals. The agent in this disclosure may include business components or business function packages driven by a large language model (LLM). The large language model can be a generative pre-trained model (GPT model) or another type of large language model. For example, for question q1 (in text form), since agent z1 is suitable for providing new home market inquiries, agent z1 can generate a text response to question q1 based on the new home business data stored at the new home business end.

[0069] S140, generating reply data displayed in a dialog box based on the reply content.

[0070] After agent z1 generates a textual response, the new home service provider converts the textual response into a format suitable for processing and display on the central control platform and then sends it to the central control platform. After receiving the data from the new home service provider, the central control platform displays the textual response in a dialog box with user U, thus answering question q1.

[0071] According to the interactive response method proposed in the embodiment of the present invention, a unified central control platform is used to uniformly manage and schedule the intelligent agents of all business parties, realizing standardized intelligent agent scheduling and management, thereby improving overall efficiency and service quality; the central control platform uniformly responds to user input in the dialog box opened on any page in the application. No matter which page or entrance the user opens the dialog box to ask questions, the central control platform uniformly identifies the intention of the question and assigns the user input to the intelligent agent suitable for answering the question according to the user's intention to generate the reply content. This eliminates the need for business parties to repeatedly develop multiple business functions for each scenario, but only needs to develop their own business functions. User questions related to a certain business function in the dialog boxes opened on different business pages will be assigned to the intelligent agent of the business function by the central control platform, realizing the universality of the intelligent agent when facing questions from different business pages, reducing the repeated development and maintenance work of each business party, and reducing development and maintenance costs.

[0072] At the same time, this embodiment can lower the development threshold by providing business parties with standardized APIs (interfaces) and access methods, allowing business parties to use intelligent agents more conveniently, accelerating business innovation and the speed of function launch. In addition, this embodiment can intelligently schedule various intelligent agents according to different business needs, improve task processing efficiency, achieve efficient collaboration between business scenarios, better cope with complex business scenarios, and improve the response speed and accuracy of the system. In addition, this embodiment has strong flexibility and scalability, supports the flexible combination and collaboration of multiple intelligent agents, can adjust the configuration of intelligent agents at any time according to business changes, maintain business agility, and provide a standardized and modular solution for the application of AI technology, greatly improving the application of intelligent agents in various business scenarios.

[0073] Figure 3 FIG1 shows an overall flow chart of the interactive response method M100 according to other embodiments of the present disclosure. Figure 3 The interactive response method M100 may further include step S101. Step S101 may be performed before step S110.

[0074] S101, based on the agent configuration information sent from the business end, obtain the ability to use the optional agent.

[0075] Agents developed by different business entities must be configured on the central control platform before they can be used. The configuration process involves the business entity developing the agent and then sending its configuration information to the business entity. The central control platform then uses this information to obtain the agent's interface data, thereby gaining the ability to use the agent.

[0076] For example, step S101 may specifically include registering the agent with an interface according to the agent configuration information sent by the service end, and obtaining the ability to use the registered agent. After registration, the agent becomes a selectable agent. In other words, the ability to call the agent is obtained through interface registration. The agent configuration information may include information such as the agent name, agent function description, and interface address.

[0077] Figure 4 FIG2 shows a schematic diagram of the overall process of the interactive response method M100 according to some embodiments of the present disclosure. Figure 4 The method of performing intent recognition on the user input at the user end (step S110) may specifically include: inputting the user input at the user end and the intent information of the predefined available intent into the language model, and obtaining the recognized intent output by the language model.

[0078] Available intents refer to the intents that can be identified at the current entry point. Businesses can define some possible intents on the central control platform. For example, the new home business side can define intents such as "new home market inquiry," and the pre-owned home business side can define intents such as "pre-owned home market inquiry."

[0079] When a user clicks a designated control on a business page to open a dialog box, the central control platform can use the dialog box ID or other information to determine the predefined identifiable intents, that is, the available intents, and then determine the intent information of the available intents. The intent information is also predefined and can include the intent name and intent description.

[0080] The user input question q1 and the intent information of the predefined available intents for the dialog box or the designated control are input into the large language model LLM. The LLM is instructed to determine the intent that is most similar to the intent of question q1 from the input available intents, thereby identifying the intent of question q1. In other words, the intent of question q1 is one of the predefined available intents for the dialog box or the designated control.

[0081] Continue reading Figure 4 The method of performing intent recognition on the user input of the user terminal (step S110) may also include: performing content review on the user input of the user terminal; if the content review passes, performing intent recognition on the user input of the user terminal, otherwise directly replying to the user terminal with specified content indicating that the answer cannot be given.

[0082] In order to meet risk control requirements during user interaction, the user input can be reviewed before intent recognition to ensure that the user input does not contain sensitive words, uncivilized language, or other prohibited expressions. If the content review passes, the user input on the user side and the intent information of the predefined available intents are input into the language model to obtain the identified intent output by the language model. If the content review fails, the predefined specified content can be used to reply to the user directly. The specified content can be: "Please use civilized language" or other similar expressions.

[0083] After obtaining the identified intent output by the language model in step S110, a mapping relationship between the identified intent and the corresponding user input can be established and cached locally. For example, a feature vector (e.g., a text vector) of the user input can be calculated first, and then a key-value pair can be formed between the feature vector and the ID of the identified intent to store the mapping relationship in the form of a key-value pair. It is understood that other forms of storage of the mapping relationship can also be used.

[0084] Accordingly, in step S110, the method of inputting the user input on the user side and the intent information of the predefined available intents into the language model may specifically include the following steps: performing a similarity comparison between the user input in the locally cached mapping relationship and the user input of the intent to be identified; and if there is a target similarity that meets the similarity requirements in the similarity obtained through the comparison, then the intent in the mapping relationship corresponding to the target similarity is used as the corresponding identified intent, otherwise the user input on the user side and the intent information of the predefined available intent are input into the language model.

[0085] Whenever the central control platform starts to identify the intent of the received user input, it will first use the user input received this time as the user input q3 of the intention to be identified this time, and compare the similarity of the user input q3 with the user inputs in each cached mapping relationship. For example, a text vector of the user input q3 can be generated first, and then similarity calculations can be performed with the text vectors in each key-value pair in turn through cosine similarity calculation or other calculation methods to obtain multiple similarity values.

[0086] The similarity requirement can be expressed using a similarity threshold. If a similarity value exceeds the preset similarity threshold, it indicates that the target similarity meets the similarity requirement, indicating that user input q3 has previously been identified and cached. In this case, there is no need to perform repeated intent recognition. Instead, the cached mapping relationship is directly used to obtain the intent ID of user input q3, thereby determining q3's intent.

[0087] If no similarity value exceeds the preset similarity threshold, it means that no similarity value meets the similarity requirement, which indicates that the user input q3 has not been recognized. In this case, the user input q3 and the predefined intent information of the available intents need to be input into the language model to obtain the recognized intent output by the language model.

[0088] It's understandable that because the similarity comparison operation is faster than LLM's intent recognition speed, using caching for intent recognition can speed up the central control platform's response to user questions. Furthermore, if the user's input intent is recognized through a mapping relationship, there's no need to create a new mapping relationship, and thus no need to cache, because the user input has already been cached.

[0089] Figure 5 FIG1 shows an overall flow chart of the interactive response method M100 according to other embodiments of the present disclosure. Figure 5 The interactive response method M100 may further include step S102. Step S102 may be performed after step S101.

[0090] S102: Create an available intent according to the intent configuration information sent by the business end. The intent configuration information may include one or more of the following information: the intent name, the intent description, and the information of the agent to which the intent applies.

[0091] Step S102 is used to define the intent, and the defined intent is used to form the intent recognition scope for subsequent intent recognition, that is, the recognized intent is one of the defined intents. When defining an intent, you can define which agents among the registered agents the intent applies to. For example, when defining the intent of "new house market consultation", the agents to which the intent applies can be defined to include only registered new house market agents. It is understandable that an intent can be defined to have only one agent or multiple agents.

[0092] During page development, the business side can configure a designated control on the page to open a dialog box during user use. When configuring a designated control or dialog box, the intent that the designated control or dialog box can recognize can be edited. That is, some or all intents from the defined intents can be selected as the intent recognition range of the designated control. The intent that is finally recognized is one of the intents included in the intent recognition range. It is understandable that each business end can use the intents predefined by all business ends, including itself.

[0093] Accordingly, determining the target agent to which the identified intention is applicable from the optional agents (step S120 ) may specifically include: using the optional agent to which the identified intention is applicable as the target agent.

[0094] Assuming that the identified intent y1 only applies to agent z1, we directly use agent z1 as the target agent. If intent y1 applies to multiple agents, we can determine a target agent from among these multiple agents based on pre-defined agent priorities or some pre-defined rules.

[0095] It is understandable that as the business evolves, the intents and agent parameters configured on the central control platform can be added, modified, and deleted. Businesses can develop new pages, create new portals for these pages, define new intents, register new agents, and configure the portals to recognize the new intents and use the new agents to answer questions related to the new intents. Furthermore, new portals can reuse existing intents and agents during configuration.

[0096] Figure 6 FIG1 shows an overall flow chart of the interactive response method M100 according to other embodiments of the present disclosure. Figure 6 , generating reply content corresponding to the user input through the target agent (step S130) may include the following steps S131 and S132.

[0097] S131: Send relevant information of the user corresponding to the user terminal to the service terminal according to the request from the service terminal. The relevant information may include one or more of the user location, historical conversations with the user, and historical consultation information of the user.

[0098] When generating responses based on user input, the agent can leverage relevant information (context) to generate higher-quality responses. Specifically, upon receiving a request from the central control platform to invoke the agent for a response, the business end can first send a request to the central control platform to obtain information such as the user's city, the user's previous input in the dialog box, and the user's previous inquiries and recorded content at offline stores.

[0099] S132, receiving the return data from the service end, and generating a reply content corresponding to the user input based on the return data. The return data can be obtained by the target agent performing natural language processing on the relevant information and the user input.

[0100] After receiving a request from the business end, the central control platform sends the relevant information to the business end. Alternatively, when invoking the agent, the central control platform can proactively send the relevant information to the business end, eliminating the need for the business end to send a request. After receiving the relevant information, the business end uses the agent to perform natural language processing (NLP) on the user input and related information, such as semantic analysis, to generate return data and send it to the central control platform. The central control platform then performs simple processing such as packaging the returned data to obtain the response content.

[0101] The return data can be the model output obtained by the target agent inputting relevant information, user input, and relevant business data into the language model. In other words, the agent can input user input, relevant information, and business data pre-stored by the business end into the large language model (LLM) to obtain the return data. Business data can be property data, including parameters such as price, location, and area. When the business end registers the agent on the central control platform, it can also configure the agent's prompt word. The prompt word is the prompt word used by the agent when calling the large language model to generate the return data.

[0102] Accordingly, in step S132, the method for generating the reply content corresponding to the user input based on the returned data may specifically include encapsulating the returned data from the service end into a specified format to obtain the reply content corresponding to the user input. The returned data may include one or more of text, images, voice, and video.

[0103] The data returned by the business end to the central control platform is unformatted and unencapsulated. Therefore, the returned data can be assembled (encapsulated) according to a specified message protocol, converted into the required format, such as a property card or other format, and then sent to the user through a dialog box for viewing. The text in the returned data can include property parameter information, images can include images of the property's interior and surroundings, voice can be used to guide users through the response, and videos can include indoor footage of the property.

[0104] Figure 7 FIG1 shows an overall flow chart of the interactive response method M100 according to other embodiments of the present disclosure. Figure 7 , step S140 may include the following steps S141 and S142.

[0105] S141, conduct content review on the reply content.

[0106] S142: If the content is approved, the reply data is generated and displayed in the dialog box; otherwise, the designated content indicating that the answer cannot be given is directly displayed in the dialog box.

[0107] The agent's feedback can also be reviewed to avoid any non-compliant expressions or terms in the answers provided to the user. If the review passes, a response is provided in the dialog box. Otherwise, the user can be directly replied with predefined content, such as "This question cannot be answered" or other similar expressions.

[0108] Figure 8 FIG1 shows an overall flow chart of the interactive response method M100 according to other embodiments of the present disclosure. Figure 8 The interactive response method M100 may further include step S150 and step S160.

[0109] S150, recording the conversation between the user terminal and the central control platform.

[0110] S160, testing the updated intelligent agent based on the above dialogue.

[0111] The central control platform records all conversations between different users, across different business pages, and on different client platforms. When new services are iterated, such as when new agents are added, these real conversations can be used as test samples in a test environment. The test results can be fed into a large language model to determine if the test passes. If the test passes, the new agent is ready for deployment.

[0112] Figure 9 The interactive diagram of the interactive response method of some embodiments of the present disclosure is shown. Figure 9 , business party B represents the business end, Figure 9 Only one business party is shown and described using this business party as an example.

[0113] S401, business party B communicates with the central control platform P1 and performs business access, and sends data for registering the intelligent agent z4, data for defining the intention y4, etc. to the central control platform P1.

[0114] S402, the central control platform P1 completes the registration of intelligent agent z4 based on the data sent by the business party B.

[0115] In S403, the central control platform P1 completes the definition of intent y4 based on the data sent by the business party B and determines the intelligent agents applicable to different intents. At this point, the central control platform has the ability to uniformly manage all intelligent agents and intents.

[0116] S404, the user operates the user terminal C to browse the business page in the APP, opens the dialog box K in a certain page and enters the question text q4, and after clicking Send, the user terminal C sends the question text to the central control platform P1.

[0117] S405, the central control platform P1 receives the question text q4 sent by the dialog box K, and first conducts content review on the text q4 through the risk control platform P2.

[0118] In step S406, the risk control platform P2 feeds back the audit result of q4 to the central control platform P1. If the audit is passed, step S407 is executed.

[0119] S407, the central control platform P1 inputs the text q4 and the optional intentions and their intention descriptions contained in the pre-configured intention recognition range of the dialog box K into the large language model LLM, so that the LLM determines the intention closest to the text q4 from the corresponding optional intentions and obtains the recognized intention y4.

[0120] S408, feedback the intention y4 to the central control platform P1.

[0121] S409, the central control platform P1 caches the mapping relationship between the intention y4 and the text q4 for subsequent use when the user or other users ask the same question.

[0122] S410, based on the applicable agent pre-configured when the intention y4 is defined, the target agent z4 is determined, and the target agent z4 of the business party B is called to start generating the reply content.

[0123] S411, business party B requests context information from the central control platform P1, including historical conversation information, user location information, etc.

[0124] S412, the central control platform P1 feeds back the context information to the business party B.

[0125] S413 , business party B inputs the context information, text q4 and some business data into the large language model LLM, and instructs the LLM to answer the question in the text q4 .

[0126] S414, LLM generates return data for answering and feeds it back to business party B.

[0127] S415, business party B feeds back the returned data as reply content to the central control platform P1 as the result of the central control platform P1 calling the intelligent agent z4.

[0128] S416: The central control platform P1 assembles (formats) the returned data to obtain a response in the required format.

[0129] S417: The assembled data is audited through the risk control platform P2.

[0130] In step S418, the risk control platform P2 feeds back the audit results of the assembled data to the central control platform P1. If the audit is passed, step S419 is executed.

[0131] S419, the central control platform P1 sends the assembled data to the user terminal C, and the user terminal C displays the data in the dialog box K, thereby completing the answer to the user's question.

[0132] Figure 10 Schematic diagram of the system architecture of some embodiments of the present disclosure is shown. Figure 10 The front-end display layer includes PC (personal computer), APP (application) and mini-programs (such as mini-programs in social software). Users can open dialog boxes and interact with AI customer service through PC browsers, mobile APPs, mini-programs, etc.

[0133] The central control layer includes the agent central control platform, which centrally manages all agents. Risk control refers to the interaction between the central control platform and the risk control platform. Conversation management primarily operates on dialog boxes, enabling multimodal conversations with users, guiding them, sending them diverse messages, and receiving user feedback to optimize conversations. During a conversation, the central control platform identifies intent and schedules agents based on user conversation messages. Agents can generate responses based on scenario-based dialogue and FAQ information in the knowledge base, as well as relevant business data stored in the customer-side business layer. After a dialog box conversation is completed, it is stored as a historical conversation on the central control platform and can be used to form a user profile. The agent layer lists the various agents managed by the central control platform.

[0134] Based on any of the above embodiments, the present disclosure also provides an interactive response device. Figure 11 This is a schematic block diagram of the structure of an interactive response device according to an embodiment of the present disclosure. Figure 11 As shown, the interactive response device includes: an intention recognition module 110, an intelligent agent determination module 120, a reply content generation module 130 and a reply module 140.

[0135] The intention recognition module 110 is used to recognize the intention of the user input of the user terminal. The control used in the user terminal to trigger the display of the dialog box for receiving the user input is any one of a plurality of designated controls preset in a variety of business pages.

[0136] The agent determination module 120 is used to determine the target agent applicable to the identified intention from multiple optional agents. Different optional agents are configured at different business ends, and different business ends are used to execute different types of business functions.

[0137] The reply content generation module 130 is used to generate reply content corresponding to the user input through the target agent.

[0138] The reply module 140 generates reply data displayed in the dialog box based on the reply content.

[0139] The interactive response device may be in the form of computer software, and each module of the interactive response device may be implemented by a computer software module. The implementation process of the functions and effects of each module in the interactive response device is detailed in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0140] The execution subject of the interactive response method in the specific implementation of the present disclosure can be an electronic device such as a computer or a server.

[0141] Therefore, based on any of the above embodiments, the present disclosure further provides an electronic device, which can execute the interactive response method of any of the embodiments described above in the present disclosure.

[0142] Figure 12 1 is a schematic block diagram of the structure of an electronic device 1000 according to an embodiment of the present disclosure.

[0143] The hardware structure of the electronic device 1000 can be implemented using a bus architecture. The bus architecture can include any number of interconnecting buses and bridges, depending on the specific application and overall design constraints of the hardware. The bus 1100 connects various circuits including one or more processors 1200, memory 1300, and / or hardware modules. The bus 1100 can also connect various other circuits 1400 such as peripheral devices, voltage regulators, power management circuits, external antennas, etc.

[0144] Bus 1100 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component Architecture (EISA) bus. Buses can be classified as address buses, data buses, control buses, and the like. For ease of illustration, this figure shows only one connecting line, but this does not imply that there is only one bus or only one type of bus.

[0145] The processor 1200 may be a central processing unit (CPU). The processor 1200 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of these chips.

[0146] Memory 1300 can be used as a non-transitory computer-readable storage medium to store non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions of the computer program in the embodiments of the present disclosure. Processor 1200 implements the interactive response method by executing the non-transitory software programs, instructions, and modules stored in memory 1300.

[0147] The memory 1300 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created by the processor 1200. In addition, the memory 1300 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 1300 may optionally include a memory remotely located relative to the processor 1200, and these remote memories may be connected to the processor 1200 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0148] The present disclosure also provides a readable storage medium having a computer program stored therein, which is used to implement the above-mentioned method when the computer program is executed by a processor. "Readable storage medium" can be any device that can contain, store, communicate, propagate or transmit a program for use in an instruction execution system, device or equipment or in combination with these instruction execution systems, devices or equipment. More specific examples of readable storage media include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable read-only memory (CDROM), etc.

[0149] The present disclosure also provides a computer program product. The method of the present disclosure can be implemented in whole or in part through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed, the process or function of the present disclosure is executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, a core network device, an OAM, or other programmable device.

[0150] A computer program or instruction can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instruction can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any accessible medium or a data storage device such as a server or data center that integrates one or more accessible media. The accessible medium can be a magnetic medium such as a floppy disk, hard disk, or magnetic tape; an optical medium such as a digital video disk; or a semiconductor medium such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or can include both volatile and non-volatile types of storage media.

[0151] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0153] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0154] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0155] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" means that the specific features, structures, or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present disclosure. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, or characteristics described may be combined in a suitable manner in any one or more embodiments / methods or examples. In addition, those skilled in the art may combine and combine different embodiments / methods or examples described in this specification and the features of different embodiments / methods or examples, unless they are contradictory.

[0156] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0157] Those skilled in the art will appreciate that the above embodiments are merely intended to clearly illustrate the present disclosure and are not intended to limit the scope of the present disclosure. Other changes or modifications may be made based on the above disclosure, and such changes or modifications are still within the scope of the present disclosure.

Claims

1. An interactive response method, characterized in that: include: Performing intention recognition on a user input on a user terminal, wherein a control in the user terminal for triggering display of a dialog box for receiving the user input is any one of a plurality of designated controls preset on a plurality of business pages; Determine a target agent to which the recognized intent applies from a plurality of optional agents, wherein different optional agents are configured at different business terminals, and different business terminals are used to perform different types of business functions, and different types of business interfaces correspond to different types of business terminals; generating, by the target agent, a response content corresponding to the user input; as well as Reply data displayed in the dialog box is generated based on the reply content.

2. The interactive response method according to claim 1, characterized in that: Identify the intent of user input on the user side, including: The user input at the user end and the intent information of the predefined available intents are input into the language model to obtain the recognized intent output by the language model.

3. The interactive response method according to claim 1 or 2, characterized in that: Identify the intent of user input on the user side, including: Perform content moderation on user input on the user side; and If the content is approved, the user input is subjected to intent recognition, otherwise the user terminal is directly replied with designated content indicating that the answer cannot be given.

4. The interactive response method according to claim 2, characterized in that: The method further comprises: Configure the intent information sent by the business end and create available intents.

5. The interactive response method according to claim 4, characterized in that: The intention configuration information includes one or more information of the intention name, intention description, and information of the agent to which the intention applies; The intent information includes an intent name and an intent description.

6. The interactive response method according to claim 1 or 5, characterized in that: Determines the target agent to which the identified intent applies from among the available agents, specifically: The optional agent to which the identified intent applies is used as the target agent.

7. The interactive response method according to claim 1, characterized in that: Generating, by the target agent, a reply content corresponding to the user input, comprising: Sending, to the service end, information related to the user corresponding to the user end according to a request from the service end, the information including one or more of the user's location, historical conversations with the user, and historical consultation information of the user; and Receive the return data from the business end, and generate reply content corresponding to the user input based on the return data, wherein the return data is obtained by the target agent performing natural language processing on the relevant information and the user input.

8. The interactive response method according to claim 7, characterized in that: The return data is a model output obtained by the target agent inputting the relevant information, the user input and the relevant business data into the language model.

9. An electronic device, characterized in that: include: a memory storing execution instructions; as well as A processor, wherein the processor executes the execution instructions stored in the memory, so that the processor executes the method according to any one of claims 1 to 8.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program is used to implement the method according to any one of claims 1 to 8.

Images