Information interaction method and device, electronic equipment and medium

Abstract

Embodiments of the present application disclose an information interaction method and device, electronic equipment and medium, and relate to the technical field of artificial intelligence. A specific implementation manner of the method comprises: pushing recommended information to a user through a native message application of a user terminal; receiving feedback information corresponding to the recommended information sent by the user through the native message application; performing intent recognition on the feedback information to obtain structured intent data corresponding to the feedback information; generating target response information corresponding to the feedback information according to the structured intent data, and pushing the target response information to the native message application. The method realizes information interaction relying on the native message application of the user terminal, does not require the user to additionally install an application program, reduces the use threshold, supports rich media content transmission, and breaks through the limitation that short messages can only transmit text. Meanwhile, the application analyzes user feedback through intent recognition technology to generate accurately adapted response information, and realizes efficient and intelligent information transmission.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to an information interaction method, device, electronic device, and medium. Background Technology

[0002] In information exchange scenarios, the demand for information transmission between users and service providers and enterprises is growing. There is an urgent need for efficient, convenient, and diverse interaction methods that can meet the communication needs of various scenarios such as daily notifications, service inquiries, and business promotions, thereby improving information transmission efficiency and user experience.

[0003] Currently, the mainstream information interaction solutions in the industry are mainly divided into two categories: one is to rely on traditional SMS services to send text messages to user terminals through the operator's network to achieve one-way or simple two-way communication; the other is to rely on the push function of specific applications, where enterprises or service providers push information to users who have installed and authorized the application through self-developed or cooperative applications to complete the interaction loop.

[0004] However, SMS interaction solutions can only carry text content, which cannot meet the diverse needs of content dissemination and the interaction methods are limited; push notifications for specific applications require users to pre-install and authorize the application, which raises the barrier to entry, and push information is easily blocked or ignored by users, resulting in low information transmission efficiency. Summary of the Invention

[0005] This application provides an information interaction method, device, electronic device, and medium to solve the problems in related technologies where information interaction relies on SMS or specific applications, resulting in a single interaction form, high usage threshold, and poor information delivery effect.

[0006] In a first aspect, embodiments of this application provide a method for information interaction, the method comprising: pushing recommendation information to a user through a native messaging application on a user terminal; receiving feedback information corresponding to the recommendation information sent by the user through the native messaging application; performing intent recognition on the feedback information to obtain structured intent data corresponding to the feedback information; generating target response information corresponding to the feedback information based on the structured intent data, and pushing the target response information to the native messaging application.

[0007] In some embodiments, performing intent recognition on feedback information to obtain structured intent data corresponding to the feedback information includes: extracting key features from the feedback information to obtain first key feedback information; supplementing the key feedback information with contextual information based on historical interaction feedback information corresponding to the first key feedback information to obtain supplemented second key feedback information; performing sentiment analysis on the second key feedback information to obtain the sentiment state corresponding to the second key feedback information; performing polysemous word resolution on the second key feedback information according to the sentiment state to obtain resolved target key feedback information; and performing intent recognition on the target key feedback information to obtain structured intent data.

[0008] In some embodiments, generating target response information corresponding to feedback information based on structured intent data includes: performing data retrieval on structured intent data based on a preset knowledge graph and / or a preset vector database to determine initial response information corresponding to the structured intent data; and integrating the initial response information based on a preset structure template to obtain target response information.

[0009] In some embodiments, data retrieval of structured intent data based on a preset knowledge graph and / or a preset vector database to determine the initial response information corresponding to the structured intent data includes: performing a first-level matching retrieval of the structured intent data based on the preset knowledge graph to determine whether there is matching information in the preset knowledge graph that matches the structured intent data; if there is no matching information in the preset knowledge graph, performing a second-level semantic retrieval of the structured intent data based on the preset vector database to determine whether there is semantic information in the preset vector database that is semantically related to the structured intent data; if there is no semantic information in the preset vector database, generating initial response information based on a preset response generation model and the structured intent data.

[0010] In some embodiments, generating initial response information based on a preset response generation model and structured intent data includes: acquiring system operation status indicators during information interaction; adjusting the model data or model type of the preset response generation model according to the system operation status indicators and preset adjustment rules; and generating initial response information based on the adjusted preset response generation model and structured intent data.

[0011] In some embodiments, the system operation status indicators include at least one of concurrent request volume, intent confidence, and response latency. Adjusting the model data or model type of the preset response generation model according to the system operation status indicators and preset adjustment rules includes: when the concurrent request volume is greater than or equal to a first preset threshold, calling the elastic scaling component to adjust the number of processing nodes of the preset response generation model; when the intent confidence is less than a second preset threshold, adjusting the number of models of the preset response generation model; and when the response latency is greater than or equal to a third preset threshold, adjusting the model type of the preset response generation model to a lightweight generation model.

[0012] Secondly, embodiments of this application provide an information interaction device, the device comprising: The first interaction unit is used to push recommendation information to users through the native messaging application on the user terminal; The receiving unit is used to receive feedback information corresponding to the recommendation information sent by the user through the native messaging application; The intelligent processing unit is used to identify the intent of the feedback information and obtain the structured intent data corresponding to the feedback information. The second interaction unit is used to generate target response information corresponding to the feedback information based on the structured intent data, and push the target response information to the native messaging application.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a processor and a memory for storing a computer program capable of running on the processor, wherein, when the processor runs the computer program, it performs the method described in any embodiment of the first aspect.

[0014] Fourthly, embodiments of this application provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described in any embodiment of the first aspect.

[0015] Fifthly, embodiments of this application provide a computer program product including a computer program that, when executed by a processor, performs the method described in any embodiment of the first aspect.

[0016] This application provides an information interaction method that pushes recommendation information to a user through the native messaging application on a user terminal; receives feedback information corresponding to the recommendation information sent by the user through the native messaging application; performs intent recognition on the feedback information to obtain structured intent data corresponding to the feedback information; generates target response information corresponding to the feedback information based on the structured intent data, and pushes the target response information to the native messaging application. This method enables information interaction based on the native messaging application on the user terminal, eliminating the need for users to install additional applications, thus lowering the barrier to entry. It also supports rich media content transmission, overcoming the limitation of SMS only transmitting text. Furthermore, this application uses intent recognition technology to analyze user feedback and generate accurately tailored response information, improving the accuracy and applicability of information interaction while avoiding the problem of push notifications from specific applications being easily blocked, achieving efficient and intelligent information delivery.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are merely embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort, and this application can be applied to other similar scenarios based on the provided drawings.

[0019] Figure 1 A flowchart illustrating an information interaction method provided in an embodiment of this application; Figure 2 A flowchart illustrating the second information interaction method provided in this application embodiment; Figure 3 This is a schematic diagram illustrating a specific example of intent recognition of feedback information provided in an embodiment of this application; Figure 4 A schematic diagram of a specific information interaction system provided in an embodiment of this application; Figure 5 This is a schematic diagram illustrating information interaction in a specific information interaction system provided in an embodiment of this application. Figure 6 This is a schematic diagram of the structure of an information interaction device 600 provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. The described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0021] It should be noted that the terms "system," "device," "unit," and / or "module" used in this application are methods of distinguishing different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they can be replaced by other expressions.

[0022] Hereinafter, the terms "first," "second," and "third" 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 indicated. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include one or more of that feature.

[0023] In digital communication and information interaction scenarios, the demand for information transmission between users and service providers and enterprises is growing. There is an urgent need for efficient, convenient, and diverse interaction methods that can support diverse content to meet the communication needs of various scenarios such as daily notifications, service inquiries, and business promotions, thereby improving information transmission efficiency and user experience.

[0024] Currently, the mainstream information interaction solutions in the industry are mainly divided into two categories: one is to rely on traditional SMS services to send text messages to user terminals through the operator's network to achieve one-way or simple two-way communication; the other is to rely on the push function of specific applications, where enterprises or service providers push information to users who have installed and authorized the application through self-developed or cooperative applications to complete the interaction loop.

[0025] However, both of the above-mentioned existing solutions have obvious limitations: for the SMS solution, its interaction form is extremely limited, and the content can only be presented in plain text, which cannot carry diverse content such as pictures, audio and video, and rich media cards, making it difficult to meet the current diverse content dissemination and interaction needs; for the application push solution, its usage threshold is high, requiring users to download and install the corresponding application in advance and grant push permissions, and due to users' resistance to application push or permission management settings, push information is easily blocked or ignored, ultimately resulting in a significant reduction in the effectiveness of information delivery and failure to achieve efficient communication.

[0026] To address the problems in existing technologies, this application proposes an information interaction method that utilizes the native entry point direct access function of 5G messaging (i.e., through the native messaging application on the user terminal) to achieve direct information delivery without requiring users to download or install applications, significantly improving information reach. Simultaneously, it integrates service functions into the message interaction interface of the native messaging application, supporting diverse rich media content formats such as text, images, audio, video, and cards, enriching information dissemination forms to meet diverse user needs and enhancing the interactive experience. Furthermore, it leverages the semantic analysis and intent matching capabilities of an intelligent agent engine to achieve intelligent and precise message delivery, further improving information adaptability and user satisfaction.

[0027] The following section provides a detailed description of an information interaction method provided in this application, with reference to the accompanying drawings.

[0028] Figure 1 A flowchart illustrating an information interaction method provided in an embodiment of this application is shown. Figure 1 As shown, the information interaction method includes steps 101-104.

[0029] Step 101: Push recommendation information to the user through the user terminal's native messaging application.

[0030] In the embodiments of this application, the application relies on the native messaging application (such as the phone's native inbox) built into the user terminal as the information carrier. Based on 5G messaging technology, the recommended information is directly pushed to the native entry point without requiring the user to download, install or authorize any third-party application. This reduces the threshold for users to receive information from the source and supports rich media format transmission, laying the foundation for subsequent diversified interactions.

[0031] Among them, user terminals refer to smart devices with communication functions used by users, such as smartphones and tablets, which need to support the 5G messaging protocol.

[0032] Native messaging applications are the default message receiving and interaction applications (i.e., the phone's native inbox) that are pre-installed on user terminals at the factory. They do not require manual installation by the user and have basic message sending and receiving, rich media rendering, and other functions. They are the system-level native entry point.

[0033] Recommended information is the core information content that the service provider pushes to users. It can cover various types such as notifications, inquiries, promotions, and service guidance, and supports diverse formats such as text, images, audio and video, and rich media cards.

[0034] Step 102: Receive feedback information corresponding to the recommendation information sent by the user through the native messaging application.

[0035] In the embodiments of this application, users do not need to switch to other third-party applications; they can directly respond to recommended information within the native messaging application using multimodal formats such as text, voice, and images (e.g., inquiring about details, expressing needs, rejecting recommendations, etc.). The system links with the native messaging application through the operator gateway interface to receive users' uplink feedback information in real time, ensuring the continuity and convenience of interaction and preventing users from interrupting their interaction due to switching applications.

[0036] Feedback information refers to the user's response to the recommended information. It is multimodal input data, including but not limited to text replies (such as "check ticket prices"), voice responses (such as verbally saying "recommend nearby restaurants"), and image feedback (such as uploading screenshots of attractions to inquire about routes).

[0037] Sending / receiving refers to the information transmission process between the 5G network and the operator's gateway. Users submit feedback information through native messaging applications (uplink transmission), and the system receives this information through a dedicated channel to ensure the stability and real-time nature of the transmission.

[0038] Step 103: Perform intent recognition on the feedback information to obtain structured intent data corresponding to the feedback information.

[0039] In the embodiments of this application, the system can transform ambiguous multimodal feedback information from users into explicit instructions that are understandable and executable by machines. The system first standardizes the feedback information (e.g., speech is converted to text via ASR, images are converted to text via CV object detection, and then word segmentation and part-of-speech tagging are performed). Then, key features are extracted using BERT semantic embedding, and the context is supplemented by historical interaction records. Ambiguity is eliminated through sentiment analysis and polysemous word resolution. Finally, relying on a hierarchical reinforcement learning framework and industry knowledge graph, intent slot filling is completed, and structured intent data is output, providing a basis for accurately generating response information.

[0040] Intent recognition refers to the process of analyzing the real needs behind user feedback information by integrating multiple technologies such as natural language processing (NLP), computer vision (CV), and speech recognition (ASR). It includes sub-steps such as semantic analysis, context completion, ambiguity resolution, and slot filling.

[0041] Structured intent data is the final output of intent recognition. It is standardized data organized in a key-value slot format, which clearly defines the user's core needs dimensions, such as [action=query][object=ticket price][location=Forbidden City][time=tomorrow], [action=recommendation][object=food][emotion=positive], etc., which makes it easy for the system to directly call resources to generate responses.

[0042] Step 104: Based on the structured intent data, generate the target response information corresponding to the feedback information, and push the target response information to the native messaging application.

[0043] In the embodiments of this application, the system can generate accurate and appropriate feedback content and push it based on the user's explicit intent. The system uses a two-level retrieval strategy to call the industry knowledge base (first knowledge graph matching, then vector database retrieval) based on structured intent data. If no match is found, a large model is activated to generate a fallback response. Simultaneously, the system combines the user's emotional attitude (e.g., negative emotions paired with reassuring language) to assemble the response content into rich media formats (e.g., text + cards, audio / video + operation guidance chains). Finally, it pushes the response to the user's native messaging application via a 5G channel, completing a full closed loop of push, feedback, parsing, and response.

[0044] Targeted response information is precise response content generated based on user feedback. It supports rich media formats and is both practical and interactive. For example, after querying tickets, it returns a card message with "price and reservation link"; after inquiring about routes, it returns a combined message with "navigation link and text and image guide".

[0045] In summary, the information interaction method proposed in this application uses the user terminal's native messaging application as the interaction carrier. Through a closed-loop process of pushing recommended information, receiving multimodal feedback, intelligent intent recognition, and generating accurate responses, it avoids the shortcomings of traditional SMS's monotonous content and the high threshold of APP push notifications. It also improves information reach through the native direct delivery characteristics of 5G messages, and enhances interaction accuracy and user experience by leveraging multimodal interaction and intelligent intent recognition technologies. Ultimately, it achieves efficient, convenient, and intelligent information interaction, applicable to multiple vertical fields such as culture and tourism, government affairs, and commercial services.

[0046] based on Figure 1 The embodiment shown, Figure 2 A flowchart of the second information exchange method is further shown. Figure 2 based on Figure 1 The illustrated embodiment further defines steps 103 and 104. Figure 2 In the illustrated embodiment, step 103 includes steps 203, 204, 205, 206, and 207, and step 104 includes steps 208 and 209. Figure 2 As shown, the method includes the following steps: Step 201: Push recommendation information to the user through the user terminal's native messaging application.

[0047] In the embodiments of this application, the application can complete the preparation of recommendation information materials, selection of target users, template construction and channel distribution through the 5G message operation management platform. The specific implementation process is as follows: 1) Material selection: Operations personnel can edit and select the content materials to be distributed through the material management module of the 5G Messaging Operations Management Platform. The material types cover rich media formats such as text, images, audio and video. 2) Customer Selection: Based on the customer management module of the 5G messaging operation and management platform, and combined with data such as user tags and behavioral preferences, the operations staff accurately select the target group for push notifications of recommended information; 3) Template generation: Operators can use the platform's visual drag-and-drop interface to flexibly combine selected materials to build 5G message templates that conform to the GSMA RCS UP 2.4 protocol, in order to generate recommended information and ensure that messages are rendered normally in the user terminal's native messaging application; 4) Channel delivery: Operators configure message sending strategies (including sending time, push frequency, etc.) within the platform. After configuration, 5G messages (i.e. recommendation information) are automatically distributed to the target user's mobile phone native inbox (i.e. the user terminal's native messaging application) via the operator gateway, realizing direct push of recommendation information.

[0048] Step 202: Receive feedback information corresponding to the recommendation information sent by the user through the native messaging application.

[0049] In the embodiments of this application, this application can render and present 5G messages based on the user terminal and perform multimodal interaction operations with the user. The specific process is as follows: 1) Message rendering and presentation: After receiving the 5G recommendation information sent by the operator's gateway, the user terminal will parse and render the message according to the GSMA RCS UP 2.4 protocol, restore it to the preset 5G message template style (including rich media cards, interactive menus, etc.), and display it in the native messaging application (the phone's native inbox) for the user to view; 2) User interaction feedback: After viewing the recommended information, users can directly reply to the upstream information in the native messaging application in a multimodal manner without switching to other applications, forming feedback information. The reply forms include, but are not limited to, text input, voice dictation, and image upload. 3) Feedback information reception: The system connects with the native messaging application through the operator gateway to receive uplink feedback information sent by the user in real time, providing data input for subsequent intent recognition and response generation.

[0050] Step 203: Extract key features from the feedback information to obtain the first key feedback information.

[0051] In this embodiment, feedback information (text, voice, images, and other multimodal forms) sent by the user through the native messaging application is synchronized to the semantic analysis module of the intelligent agent engine, initiating a real-time semantic parsing process: First, a unified preprocessing process is performed on the multimodal feedback information, namely word segmentation and part-of-speech tagging. The continuous input content is broken down into meaningful basic language units, and the grammatical attributes of each unit (such as verbs, nouns, time words, etc.) are labeled to unify the input format. Key features are extracted based on the type of feedback information: semantic features are extracted from text input using BERT semantic embedding technology; semantic features are extracted from speech input after it is converted into a text sequence using ASR (Automatic Speech Recognition) technology; and features are extracted from image input after it is identified by CV (Computer Vision) object detection algorithm and converted into text. After the above processing, key information reflecting the user's core needs (such as actions, objects, time, location, etc.) is selected to form the first key feedback information.

[0052] Step 204: Based on the historical interaction feedback information corresponding to the first key feedback information, supplement the key feedback information with contextual information to obtain the supplemented second key feedback information.

[0053] In the embodiments of this application, the semantic analysis module of the agent engine calls the historical interaction feedback information (i.e., cross-turn dialogue data) corresponding to the user, and combines it with the LSTM time sequence analysis unit to supplement the context of the first key feedback information.

[0054] Specifically, the system automatically links the user's past interaction records related to the current feedback information and completes the missing contextual information in the current input (for example, if the user's past feedback is "planning to travel to Beijing", and the current first key feedback information is "check tomorrow's weather", then it is completed as "check Beijing's weather tomorrow"). At the same time, user feedback information will be synchronized to the data platform, where the intelligent data annotation module will classify and annotate the feedback content. The annotation results will be used to assist in the accurate supplementation of the context. After context completion, a second key feedback message containing complete contextual information is obtained, avoiding misjudgment of needs due to missing context.

[0055] Step 205: Perform sentiment analysis on the second key feedback information to obtain the sentiment state corresponding to the second key feedback information.

[0056] In the embodiments of this application, based on the supplemented second key feedback information, a sentiment tendency recognition algorithm that integrates semantic analysis and contextual reasoning is used to complete the sentiment state determination.

[0057] Specifically, this application can use sentiment analysis algorithms to analyze the tone words, emotional vocabulary, and expression logic in the second key feedback information, and identify the user's emotional attitude in combination with the context. The emotional state includes positive (e.g., "The recommended attractions are really good"), negative (e.g., "This reservation process is too cumbersome"), neutral (e.g., "Check ticket prices"), and urgent (e.g., "Reply to my inquiry quickly") types. The sentiment analysis results will be synchronized to the data platform to provide a reference for subsequent user tag updates and customer group management adjustments.

[0058] Step 206: Based on the emotional state, perform polysemous resolution on the second key feedback information to obtain the resolved target key feedback information.

[0059] In the embodiments of this application, polysemous word resolution is carried out through a closed-loop process of multi-dimensional information fusion, knowledge graph-driven approach, and step-by-step verification. Combining second key feedback information, emotional state, and historical interaction feedback information, and relying on industry knowledge graphs, the true semantic meaning of polysemous words is accurately determined. The specific implementation steps are as follows: The polysemous word recognition and candidate semantic generation system first performs grammatical parsing and word filtering on the second key feedback information. Using a pre-trained polysemous word recognition model (based on a variant BERT architecture), it locates words with semantic ambiguity in the text (such as "apple," "garbage," and "chrysanthemum"). Then, it calls the industry knowledge graph to query all candidate entity nodes and associated semantics corresponding to the polysemous word. For example, for "apple," the knowledge graph returns two core candidate semantics: "fruit (entity ID: F001)" and "technology company (entity ID: T003)," along with their respective attribute tags (e.g., fruit corresponds to "edible," "orchard," and "vitamins," while technology company corresponds to "mobile phone," "iOS," and "electronic products").

[0060] Contextual feature matching extracts contextual keywords from the second key feedback information (such as "purchase," "battery life," and "after-sales service" in the sentence containing "apple"), and simultaneously retrieves relevant context from the user's historical interaction feedback information (such as the user's past queries for "new iPhone" and "phone repair"), constructing a feature set of "current context + historical context." This feature set is then matched with the associated attribute tags of each candidate semantic in the knowledge graph for similarity, and a matching score is calculated (e.g., "apple" + "battery life" scores 85 points with the attribute "technology company" and 12 points with the attribute "fruit"), initially filtering out candidate semantics with high matching degrees.

[0061] The sentiment state weighted screening, combined with the sentiment state obtained in step 205, adjusts the weights of the candidate semantics after the initial screening to clarify the priority of semantic orientation: If the emotional state is negative, such as "garbage" corresponding to the context of "product evaluation" + negative emotion, the candidate semantics of "garbage" in the knowledge graph are "waste (L001)" and "poor quality (E005)". At this time, the weight coefficient of "poor quality" and "negative emotion + product evaluation context" increases to 0.8, and the weight coefficient of "waste" decreases to 0.2. "Poor quality" is prioritized as the target semantic. If the emotional state is positive, such as "chrysanthemum" corresponding to the context of "travel recommendation" + positive emotion, the weight of the candidate semantic "plant flowers (F008)" is higher than that of "brand logo (B012)", ultimately pointing to the meaning of "flower". If the emotional state is neutral, such as "apple" corresponding to the context of "query price" with neutral sentiment, then the context feature matching score is used as the core basis to avoid emotional interference.

[0062] Semantic confidence verification and confirmation integrates "context matching score and sentiment weighting coefficient" to calculate the final confidence score of each candidate semantic (confidence score = matching score × sentiment weight + historical context fit score), and sets a confidence threshold (e.g., 0.7). If the confidence score of a candidate semantic exceeds the threshold, it is directly confirmed as the true semantic; if multiple candidate semantics have similar confidence scores (difference < 0.1), a secondary verification is initiated, which calls user tag data (e.g., user tags include "technology enthusiast" and "electronic product consumer") and combines the correlation between the tags and the candidate semantics to finally determine the unique true semantic.

[0063] After confirming the true semantics through output and reverse optimization, polysemous words are replaced with explicit semantic expressions (e.g., "garbage" is replaced with "poor quality," and "apple" is replaced with "Apple Technology Company"), resulting in unbiased target key feedback information. Simultaneously, the polysemous word resolution results are synchronized to the data platform to update the accuracy of user tags (e.g., because "apple" refers to a technology company, a new tag "electronic product demand" is added for the user), and the 5G messaging operation management platform is notified to adjust the associated customer groups. Subsequent interactions can then optimize semantic parsing efficiency based on the updated tags.

[0064] Step 207: Perform intent recognition on the key feedback information of the target to obtain structured intent data.

[0065] In the embodiments of this application, the intent matching module of the intelligent agent engine completes intent recognition and slot filling based on key feedback information of the target and relying on the hierarchical reinforcement learning framework (DQN+PPO) and the three-level mapping system of "domain, scenario, and function".

[0066] Specifically, the intent matching module performs in-depth analysis of key feedback information from the target, clarifying the user's core needs dimensions (action, object, limiting conditions, etc.); the analysis results are filled into preset core slots to form standardized structured intent data. Slot types include action slots (such as query, recommendation, reservation), object slots (such as weather, tickets, attractions), limiting slots (such as time, location, quantity), and sentiment slots (such as positive, negative, urgent). The final output of structured intent data is an example of [action=query][object=Forbidden City tickets][time=tomorrow][location=Beijing][sentiment=neutral]. This data will serve as the core basis for generating subsequent target response information, ensuring the accuracy and adaptability of the response.

[0067] Furthermore, to more clearly understand the intent recognition process for feedback information, this application... Figure 3 The demonstration showcases the specific logic behind intent recognition of feedback information: the entire process begins with "user input," first performing "word segmentation and part-of-speech tagging" on the input content to standardize the processing format; then, it is parsed according to the input type (text, speech, image)—text input is extracted for semantic features through "BERT semantic embedding," speech input is converted into a text sequence through "ASR speech-to-text," and image input relies on "CV object detection" to identify visual information and translate it; next, "contextual reasoning" is used to integrate dialogue history to complete the context, followed by "sentiment analysis" (identifying user emotions) and "polysemous word resolution" (eliminating semantic ambiguity), and finally, "intent slot filling" transforms the user input into a machine-executable structured intent, completing the entire intent recognition process.

[0068] Understandable, Figure 3 The specific implementation steps can be found in steps 203 to 207, and will not be repeated here.

[0069] Step 208: Based on a preset knowledge graph and / or a preset vector database, perform data retrieval on the structured intent data to determine the initial response information corresponding to the structured intent data.

[0070] In the embodiments of this application, the application can achieve accurate generation of response information through a two-level retrieval and large model fallback strategy. The specific process is as follows: Two-level search strategy execution: First, a first-level search is performed: based on a preset knowledge graph (i.e., an industry knowledge graph), the structured intent data is matched and searched to determine whether there is matching information in the knowledge graph that corresponds to the intent (e.g., if the structured intent is "query the price of tickets to the Forbidden City", then the entity relationship of "Forbidden City, tickets, and prices" in the knowledge graph is matched).

[0071] If no information is found in the first-level search, the second-level search is initiated: based on the preset vector database, semantic search is performed on the structured intent data to find information related to the intent semantics (such as matching historical Q&A content related to "Forbidden City tickets" through vector similarity).

[0072] If any level of retrieval matches information, the initial response information is generated directly based on the matching result. That is, if the first-level retrieval matches information, the matched information is determined as the initial response information; if the second-level retrieval matches information, the semantic information is determined as the initial response information.

[0073] If no information is found in either of the two levels of retrieval, the fallback process of the large model is triggered.

[0074] Large model fallback process (executed when no match is found in the two-level search): First, obtain system operation status indicators (including concurrent request volume, intent confidence, response latency, etc.) during the information interaction process, and monitor the SLI indicators such as QPS, P99 latency, and error rate of the intelligent agent engine in real time.

[0075] Based on system operating status indicators and preset adjustment rules, the resource configuration or model type of the preset response generation model is dynamically adjusted. Specific adjustment rules are shown in Table 1. Table 1

[0076] Referring to Table 1, if the number of concurrent requests is greater than or equal to the first preset threshold (e.g., >1000 QPS), the Kubernetes HPA (Elastic Scaling Component) will be invoked to automatically scale up the number of GPU processing nodes for the model; if the intent confidence is less than the second preset threshold (e.g., <0.7), a multi-model voting mechanism will be initiated, integrating the results of models such as GPT-4 and Claude 3 for weighted fusion; if the response latency is greater than or equal to the third preset threshold (e.g., >500ms), the model type will be switched to a lightweight generative model (e.g., TinyBERT). Based on the adjusted preset response generation model and structured intent data, initial response information is generated.

[0077] This process enables efficient and accurate response calls by leveraging knowledge graphs and vector databases, while also ensuring response availability and efficiency in extreme scenarios through large-scale model backup and dynamic resource adjustments.

[0078] Step 209: Integrate the initial response information based on the preset structure template to obtain the target response information.

[0079] In the embodiments of this application, the intelligent agent engine uses the initial response information generated in step 208 as the core content and calls the template generation module of the 5G message operation management platform through a preset interface. This module dynamically assembles the information based on preset multi-type message structure templates (covering formats such as plain text, graphic cards, and rich media interactive menus), combined with the content attributes of the initial response information and the user interaction scenario, to ensure that the target response information conforms to the GSMA RCS UP 2.4 protocol standard and can be rendered normally in the user terminal's native messaging application. After assembly, the target response information is generated and then pushed to the user terminal's native messaging application via the operator gateway, completing a complete interaction loop. After receiving the target response information, the user can reply again through the native messaging application, triggering a new round of information interaction.

[0080] In summary, this application achieves barrier-free information delivery through native messaging applications, enhances the accuracy of intent recognition by relying on multimodal parsing, context fusion, and sentiment analysis, efficiently matches response content through two-level retrieval using knowledge graphs and vector databases, and combines a large-scale model as a backup and a resource scheduling mechanism that dynamically adapts to the system's operating status. This not only breaks through the content limitations of traditional SMS and the usage barriers of third-party application push notifications, but also achieves low latency, high accuracy, and high adaptability of interaction through precise semantic parsing, multi-level retrieval optimization, and flexible resource adjustment, significantly improving information reach, response efficiency, and user experience.

[0081] based on Figures 1 to 2 The illustrated embodiments, such as Figure 4 As shown, this application provides a schematic diagram of a specific information interaction system.

[0082] Reference Figure 4 The information interaction method of this application can be applied to a 5G messaging intelligent agent propagation system (i.e., an information interaction system) based on the mobile phone's native inbox. The information interaction system specifically includes a 5G messaging operation management platform, an intelligent agent engine, and a data platform. The specific functions of the 5G messaging operation management platform, the intelligent agent engine, and the data platform are as follows: In some embodiments, the 5G messaging operation management platform is one of the core components of the 5G messaging intelligent agent propagation system based on the mobile phone's native inbox in this application. This platform provides a range of functions designed to meet users' comprehensive needs for 5G messaging editing, sending, management, and data insights.

[0083] Detailed feature introduction: (1) Centralized management of materials Function Description: Establish a unified interface for centralized management of four categories of material resources: images, videos, audio, and URL links.

[0084] Technical benefits: Facilitates efficient retrieval and application, and improves the efficiency and personalization of message editing.

[0085] (2) Menu Management Function Description: Allows users to flexibly configure the bottom menu, including convenient functions such as setting up automatic card replies, URL redirection links, and one-click dialing.

[0086] Technical benefits: Improves user experience and interactivity, enabling users to access the services or information they need more quickly.

[0087] (3) Template Management Feature Description: Provides a template editor that combines static and dynamic elements, supporting various formats such as text, multimedia, and rich media cards.

[0088] Technical benefits: It meets users' diverse message editing needs and enhances the attractiveness and reach of messages.

[0089] (4) Message Management Feature Description: Supports bulk messaging and personalized one-on-one replies, enabling the creation of rich conversation flows. Also compatible with video MMS and SMS formats.

[0090] Technical benefits: Ensures that 5G messages can be delivered to users efficiently and accurately, improving information delivery efficiency and user satisfaction.

[0091] (5) User Management Function description: Provides functions such as customer group management, user grouping, and blacklists.

[0092] Technical effect: Provides capability support for user-tiered and hierarchical management.

[0093] (6) Data Management Function Description: Provides data analysis and insight functions to help operations personnel understand key metrics such as message sending effectiveness and user behavior.

[0094] Technical impact: Provides data support for optimizing message push strategies and improving user experience.

[0095] In some embodiments, the intelligent agent engine is the core processing hub of the 5G message intelligent agent propagation system, responsible for realizing the intelligentization and automation of human-computer interaction. This module provides real-time decision support for message interaction through multimodal fusion analysis and service scheduling capabilities.

[0096] Detailed feature introduction: (1) Dynamic semantic analysis module Functional Description: Employing a Transformer-based hybrid neural network architecture, integrating the BERT semantic understanding model and LSTM temporal analysis unit, it enables contextual reasoning, sentiment identification, and polysemous word resolution in cross-turn dialogues.

[0097] Technical benefits: Dialect recognition covers 7 major dialect regions including Cantonese and Minnan; supports the understanding of professional terms in 15 vertical fields such as culture and tourism, government affairs, and sports.

[0098] (2) Multi-dimensional intent matching module Function Description: Construct a hierarchical reinforcement learning framework (DQN+PPO), combined with a 200,000-node industry knowledge graph (containing 8,000+ business entity relationships), to realize a three-level mapping system of demand-service (domain → scenario → function).

[0099] Technical performance: In complex scenarios such as cultural tourism and government affairs, the accuracy rate of intent recognition is 98.7% (test set: ICSL-2025); the peak concurrent processing capacity is 3000 QPS.

[0100] (3) Multi-model collaborative decision-making module Function Description: Develop a model scheduler based on Bayesian optimization to intelligently allocate 12 basic models, including NLP (GPT-4Omni / ERNIE4.0) and CV (YOLOv10), and achieve elastic allocation of computing resources through Kubernetes.

[0101] Technical effect: Intelligent combination of different modal model outputs significantly improves the accuracy of comprehensive judgment.

[0102] (4) Feedback Information Combination Module Function Description: Design a multimodal information assembly pipeline, and integrate multi-source data (text / charts / API call results) with a structured output engine; support multimodal response generation (voice broadcast + visual dashboard + operation guidance chain) and update relevant knowledge base.

[0103] Technical effect: Enables the assembly of complex information in seconds, significantly improving user operation efficiency.

[0104] In some embodiments, the data platform serves as the data asset hub of the 5G messaging intelligent agent propagation system, enabling in-depth data value mining and secure data transfer through end-to-end data processing capabilities. This module constructs a full-lifecycle management system from data collection to intelligent applications, providing high-quality data support for upper-layer services.

[0105] Detailed feature introduction: (1) Intelligent data annotation module Function Description: Employing a collaborative mechanism of AI pre-annotation and manual quality inspection, this system annotates various types of data, enabling efficient production of structured data.

[0106] Technical effects: Significantly reduces the workload of manual annotation and improves the quality of data annotation.

[0107] (2) User profiling engine module Function Description: Aggregates user touchpoint data and behavioral links from multiple sources in real time, and generates a 360° user profile through a dynamic tagging system.

[0108] Technical benefits: Supports flexible configuration and real-time activation of tag strategies to meet the needs of precision marketing.

[0109] (3) Industry Knowledge Base Module Function Description: It comes pre-built with standardized data models for industries such as sports and cultural tourism, and provides scenario-based data package API services.

[0110] Technical effect: Supports quick access to industry knowledge assets.

[0111] Furthermore, in some embodiments, a schematic diagram illustrating the information interaction implemented by the system is shown below. Figure 5 As shown.

[0112] Reference Figure 5 This system manages materials, selects customer groups, generates templates, and distributes messages through a 5G messaging operation and management platform. Messages are pushed to the user's native inbox, where users can interact and provide feedback. The feedback information is synchronized to the intelligent agent engine for semantic analysis and intent matching. Then, it generates a response by searching through knowledge graphs and knowledge bases (if a match is found, the feedback is directly integrated; otherwise, a large model is activated as a fallback). Simultaneously, the feedback information also flows to the data platform for data annotation and user tag management. Finally, it is synchronized to the customer group management module of the 5G messaging operation and management platform, forming a closed-loop information interaction process of message push, user interaction, intelligent parsing, data accumulation, and strategy optimization. This achieves efficient reach, intelligent interaction, and continuous data-driven optimization of 5G messaging through the native entry point.

[0113] It should be noted that the specific implementation process of each module in the information interaction system can be found by referring to... Figures 1 to 2 The embodiments shown will not be described in detail here.

[0114] To implement the above embodiments, this application also provides an information interaction device. Figure 6 This is a schematic diagram of the structure of an information interaction device 600 provided in an embodiment of this application. Figure 6 As shown, the device includes: The first interaction unit 610 is used to push recommendation information to the user through the user terminal's native messaging application; The receiving unit 620 is used to receive feedback information corresponding to the recommendation information sent by the user through the native messaging application; The intelligent processing unit 630 is used to perform intent recognition on the feedback information and obtain the structured intent data corresponding to the feedback information; The second interaction unit 640 is used to generate target response information corresponding to the feedback information based on the structured intent data, and push the target response information to the native messaging application.

[0115] In some embodiments, the intelligent processing unit 630 is configured to: extract key features from the feedback information to obtain first key feedback information; supplement the key feedback information with contextual information based on the historical interaction feedback information corresponding to the first key feedback information to obtain supplemented second key feedback information; perform sentiment analysis on the second key feedback information to obtain the sentiment state corresponding to the second key feedback information; perform polysemous word resolution on the second key feedback information according to the sentiment state to obtain resolved target key feedback information; and perform intent recognition on the target key feedback information to obtain structured intent data.

[0116] In some embodiments, the second interaction unit 640 is configured to: perform data retrieval on structured intent data based on a preset knowledge graph and / or a preset vector database to determine the initial response information corresponding to the structured intent data; and integrate the initial response information based on a preset structure template to obtain target response information.

[0117] In some embodiments, the second interaction unit 640 is configured to: perform a first-level matching retrieval on the structured intent data based on a preset knowledge graph to determine whether there is matching information in the preset knowledge graph that matches the structured intent data; if there is no matching information in the preset knowledge graph, perform a second-level semantic retrieval on the structured intent data based on a preset vector database to determine whether there is semantic information in the preset vector database that is semantically related to the structured intent data; if there is no semantic information in the preset vector database, generate initial response information based on a preset response generation model and the structured intent data.

[0118] In some embodiments, the second interaction unit 640 is configured to: acquire system operation status indicators during the information interaction process; adjust the model data or model type of the preset response generation model according to the system operation status indicators and preset adjustment rules; and generate initial response information based on the adjusted preset response generation model and structured intent data.

[0119] In some embodiments, the system operation status indicators include at least one of concurrent request volume, intent confidence, and response latency. The second interaction unit 640 is configured to: call the elastic scaling component to adjust the number of processing nodes of the preset response generation model when the concurrent request volume is greater than or equal to a first preset threshold; adjust the number of models of the preset response generation model when the intent confidence is less than a second preset threshold; and adjust the model type of the preset response generation model to a lightweight generation model when the response latency is greater than or equal to a third preset threshold.

[0120] The methods and apparatus provided in the embodiments of this application have been described above. To implement the functions of the methods provided in the embodiments of this application, the electronic device may include a hardware structure and software modules, and may implement the above functions in the form of a hardware structure, software modules, or a hardware structure plus software modules. One of the above functions may be executed in the form of a hardware structure, software modules, or a hardware structure plus software modules.

[0121] Figure 7 This is a block diagram illustrating an electronic device 700 for implementing the above-described information interaction method, according to an exemplary embodiment. For example, the electronic device 700 may be a mobile phone, computer, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0122] Reference Figure 7 The electronic device 700 may include a communication interface 701, capable of interacting with other devices; a processor 702, connected to the communication interface 701 to enable interaction with other devices, used to execute the methods provided by one or more of the above-described technical solutions when running a computer program; and a memory 703, on which the computer program is stored. Specifically, the specific processing procedure of the processor 702 can refer to the information interaction method described in the above embodiments of this disclosure.

[0123] Of course, in practical applications, the various components in electronic device 700 are coupled together through bus system 704. It can be understood that bus system 704 is used to realize the connection and communication between these components. In addition to a data bus, bus system 704 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in... Figure 7 The general designated all buses as Bus System 704.

[0124] The memory 703 in this embodiment is used to store various types of data to support the operation of the electronic device 700. Examples of such data include any computer program used to operate on the electronic device 700.

[0125] The methods disclosed in the embodiments of this application can be applied to processor 702, or implemented by processor 702. Processor 702 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 702 or by instructions in the form of software. The processor 702 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 702 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in memory 703. Processor 702 reads the information in memory 703 and combines its hardware to complete the steps of the aforementioned method.

[0126] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned method.

[0127] Embodiments of this disclosure also propose a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the information interaction methods described in the above embodiments of this disclosure.

[0128] Embodiments of this disclosure also provide a computer program product, including a computer program that is executed by a processor using the information interaction method described in the above embodiments of this disclosure.

[0129] Embodiments of this disclosure also propose a chip including one or more interface circuits and one or more processors; the interface circuits are used to receive signals from the memory of an electronic device and send signals to the processors, the signals including computer instructions stored in the memory, which, when executed by the processor, cause the electronic device to perform the information interaction method described in the above embodiments of this disclosure.

[0130] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0131] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0132] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0133] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (control method), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0134] It should be understood that various parts of the embodiments of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0135] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0136] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc.

[0137] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An information exchange method, characterized in that, include: Push recommendation information to users through the native messaging application on the user's terminal; Receive feedback information corresponding to the recommendation information sent by the user through the native messaging application; The feedback information is subjected to intent recognition to obtain structured intent data corresponding to the feedback information; Based on the structured intent data, target response information corresponding to the feedback information is generated, and the target response information is pushed to the native messaging application.

2. The method according to claim 1, characterized in that, The step of performing intent recognition on the feedback information to obtain structured intent data corresponding to the feedback information includes: Key features are extracted from the feedback information to obtain the first key feedback information; Based on the historical interaction feedback information corresponding to the first key feedback information, the key feedback information is supplemented with contextual information to obtain the supplemented second key feedback information. Sentiment analysis is performed on the second key feedback information to obtain the emotional state corresponding to the second key feedback information; Based on the emotional state, the second key feedback information is decrypted to obtain the decrypted target key feedback information; The target key feedback information is subjected to intent recognition to obtain the structured intent data.

3. The method according to claim 1, characterized in that, The step of generating the target response information corresponding to the feedback information based on the structured intent data includes: Based on a preset knowledge graph and / or a preset vector database, data retrieval is performed on the structured intent data to determine the initial response information corresponding to the structured intent data; The initial response information is integrated based on a preset structural template to obtain the target response information.

4. The method according to claim 3, characterized in that, The step of retrieving data from the structured intent data based on a preset knowledge graph and / or a preset vector database to determine the initial response information corresponding to the structured intent data includes: Based on the preset knowledge graph, a first-level matching retrieval is performed on the structured intent data to determine whether there is matching information in the preset knowledge graph that matches the structured intent data; If the matching information does not exist in the preset knowledge graph, a second-level semantic retrieval is performed on the structured intent data based on the preset vector database to determine whether there is semantic information related to the semantics of the structured intent data in the preset vector database. If the semantic information is not present in the preset vector database, the initial response information is generated based on the preset response generation model and the structured intent data.

5. The method according to claim 4, characterized in that, The generation of the initial response information based on the preset response generation model and the structured intent data includes: Acquire system operation status indicators during information exchange; Based on the system operating status indicators and preset adjustment rules, adjust the model data or model type of the preset response generation model; The initial response information is generated based on the adjusted preset response generation model and the structured intent data.

6. The method according to claim 5, characterized in that, The system operation status indicators include at least one of concurrent request volume, intent confidence, and response latency. The step of adjusting the model data or model type of the preset response generation model according to the system operating status indicators and preset adjustment rules includes: If the number of concurrent requests is greater than or equal to a first preset threshold, the elastic scaling component is invoked to adjust the number of processing nodes of the preset response generation model. If the confidence level of the intent is less than the second preset threshold, adjust the number of models in the preset response generation model; If the response delay is greater than or equal to a third preset threshold, the model type of the preset response generation model is adjusted to a lightweight generation model.

7. An information interaction device, characterized in that, The device includes: The first interaction unit is used to push recommendation information to users through the native messaging application on the user terminal; A receiving unit is configured to receive feedback information corresponding to the recommendation information sent by the user through the native messaging application; An intelligent processing unit is used to perform intent recognition on the feedback information to obtain structured intent data corresponding to the feedback information; The second interaction unit is used to generate target response information corresponding to the feedback information based on the structured intent data, and push the target response information to the native messaging application.

8. An electronic device, characterized in that, include: The processor and the memory used to store computer programs that can run on the processor. When the processor is used to run the computer program, it performs the method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images