Interactive personalized information pushing method and wearable device

Abstract

The invention discloses an interactive personalized information pushing method, and the method comprises the steps: obtaining user data, building a dynamic user portrait for a user based on the user data, and enabling the dynamic user portrait to comprise a demand tag, a preference weight coefficient and an economic capability evaluation index; user requirements are analyzed in real time through natural language interaction, and dynamic features and intention keywords are extracted; in combination with multi-dimensional information of a commodity database, the system adopts a priority matching algorithm to generate a recommended commodity list, and generates personalized push content through a natural language generation engine, including commodity core selling points, matching degree analysis and interactive question and answer guidance; and finally, the system sends the content to a user terminal through multi-channel pushing, monitors user behaviors in real time to update a user portrait, and continuously optimizes a recommendation strategy.

Description

Technical Field

[0001] This invention relates to the field of personalized information push technology, and more specifically, to an interactive personalized information push method and wearable device. Background Technology

[0002] In today's digital age, with the rapid development of artificial intelligence technology, intelligent real estate sales robots are gradually becoming an important tool for marketing and services in the real estate industry. Modern consumers have increasingly diverse and personalized demands for real estate. They not only focus on the basic attributes of a house, such as area, layout, and price, but also have higher requirements for surrounding facilities, community environment, and property services. Accurately grasping customer needs and providing real estate information that meets their expectations has become a significant challenge for real estate companies.

[0003] However, in actual use, it still has some drawbacks. For example, traditional real estate sales methods often rely on manual telemarketing, offline activities, and simple online registration. These methods are not only inefficient, but also difficult to accurately grasp customer needs, resulting in the loss of a large number of potential customers. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an interactive personalized information push method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Step A1: Obtain user data, which includes user identity information, browsing history, offline consultation records, and purchase qualification data authorized by third-party platforms;

[0007] Step A2: Build dynamic user profiles based on user data. User profiles include purchase demand tags, preference weight coefficients, and economic capacity assessment indicators.

[0008] Step A3: Engage in real-time dialogue with users through natural language interaction, parse the user's voice or text input, and extract dynamic demand features and intent keywords;

[0009] Step A4: Based on the dynamic user profile and the dynamic demand features extracted in real time, combined with multi-dimensional product information in the product database, a priority matching algorithm is used to generate a list of recommended products.

[0010] Step A5: Based on the recommended product list, call the natural language generation engine to generate personalized push content. The push content includes the core selling points of the products, the matching degree with user needs, and interactive Q&A guidance.

[0011] Step A6: Send the personalized content to the user's terminal through multi-channel push, and monitor user clicks, feedback and subsequent interaction behavior in real time to update the user profile.

[0012] Preferably, in step A1, by acquiring user data, the system first integrates user identity information, historical browsing records, offline consultation records, and encrypted third-party platform authorization data. Based on cross-validation and data cleaning techniques, the algorithm model will perform in-depth modeling of user purchase intention strength, economic capacity threshold, and regional preference heatmap. Simultaneously, it will perform semantic analysis on offline consultation recordings using natural language processing to extract implicit demand tags. Finally, an interactive report containing purchase qualification pre-screening conclusions, personalized product matching indices, and simulated financial solutions will be generated through a dynamic knowledge graph.

[0013] Preferably, in step A2, a dynamic user profile is constructed based on the acquired user data. This process encompasses multiple stages, including data cleaning, feature extraction, model calculation, and real-time updates. First, the system cleans and preprocesses user identity information, historical browsing records, offline consultation records, and purchase qualification data, removing duplicate and invalid data. Natural language processing technology is then used to parse and classify the unstructured data. Next, the system extracts key features from the cleaned data to generate purchase demand tags, preference weight coefficients, and economic capacity assessment indicators.

[0014] Preferably, in step A3, the user's voice input is converted into text data using speech recognition technology, or the user's input text information is received directly. Then, the system uses natural language processing technology to perform word segmentation, part-of-speech tagging, and syntactic analysis on the text, extracting key information and identifying the user's intent.

[0015] Preferably, in step A4, key features are extracted from the dynamic user profile, including purchase demand tags, preference weight coefficients, and economic capacity assessment indicators. Simultaneously, the system further optimizes the user profile by combining dynamic demand features extracted from real-time dialogue. For example, if a user mentions "wanting the house to come with renovations" in a dialogue, the system will include this demand in the real-time features.

[0016] Next, the system extracts multi-dimensional product information from the product database, including the product's basic attributes, geographical location, supporting facilities, and market dynamics; this information is stored in a structured manner and represented by feature vectors.

[0017] Preferably, in step A5, the system extracts the core selling points of each product from the recommended product list, including key information such as apartment type, price, location, and supporting facilities.

[0018] Preferably, in step A6, the system automatically selects the push channel based on the user's terminal type and usage habits; for users who are accustomed to using instant messaging tools, the system will send card-style messages via WeChat or WhatsApp, including product images, key selling points, and interactive buttons; for users who prefer email, the system will generate structured emails with detailed product information and links; for users who use mobile apps, the system will send concise product summaries and jump links via push notifications.

[0019] In terms of push content design, in instant messaging tools, the system will generate card messages with pictures and text, highlighting the core selling points of the product and matching analysis, and embed interactive buttons such as "Click to view details" and "Schedule a viewing"; in emails, the system will generate detailed product descriptions, with high-definition pictures and interactive links; in mobile apps, the system will send concise product summaries via push notifications and support one-click jump to the details page.

[0020] The technical effects and advantages of this invention are as follows:

[0021] This invention constructs dynamic user profiles by acquiring user data; it analyzes user needs in real time through natural language interaction, extracting dynamic demand features; it combines a product database and uses a priority matching algorithm to generate a recommended product list; subsequently, it calls a natural language generation engine to generate personalized push content based on the core selling points of the products and the matching degree of user needs, embedding interactive Q&A guidance. The system pushes personalized content to user terminals through multiple channels according to the user's terminal type and usage habits. The push content is designed with rich graphics and text, highlighting product selling points and matching degree, and providing interactive buttons. Finally, the system monitors user clicks, feedback, and subsequent interaction behaviors in real time, continuously updating the user profile to optimize recommendation strategies and improve user experience. This invention features interactive personalized information push, improves marketing efficiency and service quality, accurately grasps customer needs, and provides customers with personalized real estate information and services. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the interactive personalized information push method provided in the embodiments of this application;

[0023] Figure 2 This is a schematic diagram illustrating the architecture for information sharing and transmission between a wearable device and a field robot, as provided in an embodiment of this application.

[0024] Figure 3 This is a schematic diagram of an interactive personalized information push system provided in an embodiment of this application. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 As shown, the interactive personalized information push method provided by the present invention includes:

[0027] The on-site robot receives requests from multiple terminals, including user data.

[0028] Step A1: Obtain user data, which includes user identity information, browsing history, offline consultation records, and qualification data authorized by third-party platforms.

[0029] In step A1, user data is acquired. The system first integrates user identity information, historical browsing records, offline consultation records, and encrypted third-party platform authorization data. Based on cross-validation and data cleaning techniques, the algorithm model will perform in-depth modeling of user purchase intention strength, economic capacity threshold, and regional preference heatmap. At the same time, semantic analysis of offline consultation recordings is performed through natural language processing to extract implicit demand tags. Finally, an interactive report containing purchase qualification pre-approval conclusions, personalized product matching index, and financial solution simulation calculations is generated through a dynamic knowledge graph.

[0030] The user identity information includes name, contact information, and occupation; browsing history includes product page dwell time, search keywords, and favorites behavior data; offline consultation records include structured data such as frequency of visits to the sales office, apartment type preferences, and price sensitivity; third-party platform authorized data includes purchase qualification verification results such as social security payment records, credit scores, and housing provident fund loan limits; and implicit demand tags include school district priority and commuting radius tolerance.

[0031] Step A2: Construct a dynamic user profile for the user based on the user data. The dynamic user profile includes demand tags, preference weight coefficients, and economic capacity assessment indicators.

[0032] In step A2, a dynamic user profile is constructed based on the acquired user data. This process encompasses multiple stages, including data cleaning, feature extraction, model calculation, and real-time updates. First, the system cleans and preprocesses user identity information, historical browsing records, offline consultation records, and purchase qualification data, removing duplicate and invalid data. Natural language processing technology is then used to parse and classify the unstructured data. Next, the system extracts key features from the cleaned data to generate purchase demand tags, preference weight coefficients, and economic capacity assessment indicators.

[0033] User profiling is the labeling of user information, that is, tagging users. After collecting and analyzing data on consumers' social attributes, lifestyle habits, and consumption behavior, companies abstract a complete commercial profile of a user.

[0034] For example, by analyzing the product pages browsed by users, the system can extract features such as apartment type, region, and price to generate tags such as "preference for three-bedroom apartments" or "demand for school district housing"; by quantifying user behavior data (such as clicks, favorites, and dwell time), it can calculate the user's preference weight for different features; and by using purchase qualification data authorized by third-party platforms (such as credit scores and income certificates), it can assess the user's economic capacity.

[0035] In the process of building dynamic user profiles, the system uses machine learning algorithms to model user characteristics and monitors user behavior data in real time, dynamically adjusting the labels and weight coefficients in the profile.

[0036] For example, if a user clicks on products in a particular area multiple times, the system will increase the preference weight for that area; if a user mentions "needing a house near a school" in their inquiry, the system will generate a "school district housing demand" tag and optimize the recommendation strategy. Furthermore, the system analyzes the user's economic capacity indicators to filter products that match their budget and provide personalized purchasing plans; these economic capacity indicators include income level and loan amount.

[0037] Step A3 involves engaging in real-time dialogue with the user through natural language interaction, parsing the user's input voice or text information, and extracting dynamic demand features and intent keywords.

[0038] In step A3, the user's voice input is converted into text data using speech recognition technology, or the system directly receives the user's text input. Next, the system uses natural language processing technology to perform word segmentation, part-of-speech tagging, and syntactic analysis on the text, extracting key information and identifying the user's intent.

[0039] For example, when a user enters "I want to find a three-bedroom apartment, preferably near the subway", the system will extract "three-bedroom apartment" and "near the subway" as key demand features and identify the user's intent as "searching for products".

[0040] During the intent recognition process, the system uses a pre-trained intent recognition model in the real estate field, combined with knowledge graphs and dialogue scenario templates.

[0041] For example, the system dynamically adjusts its intent recognition strategy based on contextual information from user input (such as previous conversations). If a user previously mentioned a budget of "under 5 million," the system will prioritize recommending products that fit that budget in subsequent conversations.

[0042] The system also triggers a request dialogue process by monitoring users' negative feedback in real time, further refining user needs.

[0043] For example, when a user is dissatisfied with a recommended product, the system will proactively ask, "What aspects do you value most? Is it price, location, or apartment type?" to obtain more specific information about your needs.

[0044] In the dynamic demand extraction stage, the system not only focuses on the user's explicitly expressed needs but also uncovers implicit needs through semantic analysis. For example, when a user mentions "my child is about to start school," the system infers that the user may have a demand for "school district housing" and includes it in the dynamic user profile. At the same time, the system dynamically adjusts the weight of demands based on the user's real-time feedback.

[0045] For example, if a user clicks on products in a certain area multiple times, the system will increase the preference weight of that area and prioritize displaying related products in subsequent recommendations.

[0046] Step A4: Based on the dynamic user profile, the real-time extracted dynamic demand features, and the multi-dimensional product information in the product database, a priority matching algorithm is used to generate a recommended product list.

[0047] In step A4, key features are extracted from the dynamic user profile, including purchase demand tags, preference weight coefficients, and economic capacity assessment indicators. Simultaneously, the system further optimizes the user profile by combining dynamic demand features extracted from real-time conversations. For example, if a user mentions "wanting the house to come furnished" in a conversation, the system will include this demand in the real-time features.

[0048] Next, the system extracts multi-dimensional product information from the product database, including the product's basic attributes, geographical location, supporting facilities, and market dynamics; this information is stored in a structured manner and represented by feature vectors.

[0049] During the matching process, the system uses a priority matching algorithm to calculate the degree of matching between product features and user needs; a set of products is represented by a vector: V = [v1, v2, v3, ..., v n In this context, v1 represents the apartment type, v2 represents the price, v3 represents the distance to the subway, and so on, up to v n.

[0050] The specific method for calculating feature similarity is as follows:

[0051] Where A(U,V) represents the feature similarity, U·V represents the dot product of the vectors, and ||U|| and ||V|| represent the magnitudes of the vectors.

[0052] The matching score calculation weights are dynamically adjusted based on the user's preference weight coefficients and real-time demand characteristics. The specific matching score calculation method is as follows:

[0053] Where B(U, V) represents the matching degree, β i U is represented as the weight of the i-th feature. i Let V be the i-th dimension of user needs. i This is represented as the i-th dimension of the product features;

[0054] The system sorts the products based on the adjusted matching degree and generates a list of recommended products; at the same time, it will further optimize the sorting results by taking into account the market dynamics of the products.

[0055] Step A5: Based on the recommended product list, a natural language engine is invoked to generate personalized push content. The push content includes the core features of the products, the matching degree with user needs, and interactive question-and-answer guidance.

[0056] In step A5, the system extracts the core selling points of each product from the recommended product list, including key information such as apartment type, price, location, and supporting facilities.

[0057] For example, for a three-bedroom apartment, the system will extract its core selling points such as "close to the subway," "school district," and "fully furnished." The system combines dynamic user profiles and real-time demand characteristics to analyze the matching degree between the product and user needs.

[0058] For example, if a user prefers a school district property and has a budget of less than 5 million, the system will calculate the match between the property's school district advantages and price range, and generate a match analysis.

[0059] When generating push content, the NLG engine uses a pre-trained text generation model for the real estate sector, combined with templated and rule-based generation strategies. The system automatically generates push content based on product characteristics and user needs. Furthermore, the system embeds interactive Q&A prompts within the push content, such as "Click to view detailed floor plans" or "Schedule an offline viewing," to further engage users.

[0060] Step A6: Send the personalized content to the user terminal through multi-channel push, and monitor user clicks, feedback and subsequent interaction behavior in real time in order to update the dynamic user profile.

[0061] In step A6, the system automatically selects the push channel based on the user's terminal type and usage habits. For users who are accustomed to using instant messaging tools, the system will send card-style messages via WeChat or WhatsApp, including product images, key selling points, and interactive buttons. For users who prefer email, the system will generate structured emails with detailed product information and links. For users who use mobile apps, the system will send concise product summaries and redirect links via push notifications.

[0062] In terms of push content design, in instant messaging tools, the system will generate card messages with pictures and text, highlighting the core selling points of the product and matching analysis, and embed interactive buttons such as "Click to view details" and "Schedule a viewing"; in emails, the system will generate detailed product descriptions, with high-definition pictures and interactive links; in mobile apps, the system will send concise product summaries via push notifications and support one-click jump to the details page.

[0063] After the push content is sent, the system will monitor the user's clicks, feedback and subsequent interaction behavior in real time; for example, the system will record whether the user clicked on the product details, whether they saved the product, whether they initiated an inquiry or scheduled a viewing.

[0064] This application also provides a wearable device, such as Figure 2 The diagram shows the architecture for information sharing and transmission between the wearable device and the on-site robot. The wearable device and the on-site robot share and transmit information; receive request information from multiple terminals, which includes request data and location information; control the on-site robot and send tasks to the on-site robot according to the request information so that the on-site robot can reach the corresponding location according to the request information.

[0065] Specifically, the location layout of the sales office is shown in Table 1, and the requested information is shown in Table 2.

[0066] Table 1

[0067] Location A Position D Position B Position E Position C Position F

[0068] Table 2

[0069] Request information Request data and location information User A sent request 1 Men's clothing, position A User B sent request 2 Cosmetics, Location D User C sent request 3 Buying a house, location F

[0070] Optionally, the shopping mall robot receives multiple requests sent by users on their terminals. For example, request 3 is a request from user A at location F in the mall to purchase a house. The wearable device controls the shopping mall robot and sends a task to the robot based on the information in request 3. This task can be an instruction to complete request 3, so that the shopping mall robot can reach location F in the mall, i.e., the location of user C, based on the information in request 3.

[0071] Optionally, the wearable device also includes generating control commands based on path planning data and sending the control commands to the robot; the control commands include the robot's location information to be reached and its path planning data.

[0072] Control commands are generated based on path planning data to drive the robot to move along the planned route.

[0073] Wearable devices also include wireless communication modules that support Bluetooth / Wi-Fi / 5G, used to establish data links with external robots and to share and transmit information with them.

[0074] Wearable devices also include a compass that detects the direction of the Earth's magnetic field in real time and generates a direction signal; a navigation module that integrates a GPS positioning unit to obtain the user's location information and plan a route for the user; and a route planning module that plans and updates the route planning data for the user in real time based on the direction signal and route planning data so that the user can travel along the correct planned route.

[0075] Wearable devices also include temperature sensors to detect ambient temperature and send it to the administrator's terminal in real time.

[0076] Specifically, wearable devices will transmit the collected mall temperature data to the mall management system's server or the manager's terminal in real time, so that the mall manager can monitor and manage the mall temperature in real time, such as turning on the air conditioner and adjusting the temperature.

[0077] Wearable devices also include a health monitoring module that collects the user's heart rate and blood oxygen data in real time.

[0078] Specifically, by emitting light of a specific wavelength onto the surface of human skin, the system receives the light signal reflected back from the blood flow within the blood vessels and calculates the heart rate based on the periodic changes in the light signal. The blood oxygen sensor employs dual-wavelength (red light and infrared light) technology, utilizing the difference in absorption characteristics of oxyhemoglobin and deoxyhemoglobin to different wavelengths of light, and calculates blood oxygen saturation by measuring the intensity of reflected light.

[0079] The wearable device uses built-in accelerometer and gyroscope sensors to identify the user's type of exercise (such as walking, running, cycling, etc.) and number of steps, and combines heart rate and blood oxygen data to perform sports and health analysis.

[0080] Figure 3 This is a schematic diagram of an interactive personalized information push system provided in an embodiment of this application. The interactive personalized information push system can be a system embedded in a sales robot. The system includes: an acquisition unit 201, used to acquire user data, including user identity information, historical browsing records, offline consultation records, and purchase qualification data authorized by a third-party platform; a construction unit 202, used to construct a dynamic user profile for the user based on the user data, the dynamic user profile including purchase demand tags, preference weight coefficients, and economic capacity assessment indicators; and an interaction unit 203, used to engage in real-time dialogue with the user through natural language interaction and parse user input. Voice or text information is used to extract dynamic demand features and intent keywords; the recommendation unit 204 is used to generate a recommended product list based on the dynamic user profile and the real-time extracted dynamic demand features, combined with multi-dimensional product information in the product database, using a priority matching algorithm; the generation unit 205 is used to generate personalized push content based on the recommended product list by calling a natural language engine, the push content including the core selling points of the product, the matching degree analysis with user needs, and interactive question-and-answer guidance; the update unit 206 is used to send the personalized content to the user terminal through multi-channel push, and monitor user clicks, feedback and subsequent interaction behaviors in real time to update the dynamic user profile.

[0081] Optionally, the robot also includes a navigation module, which integrates a GPS positioning unit to obtain the robot's location and plan a path for the robot, and shares the path planning data with wearable devices.

[0082] Optionally, the robot receives control commands sent by the wearable device and navigates to the target location in the control commands, wherein the control commands are generated by the wearable device based on path planning data.

[0083] The aforementioned interactive personalized information push system is a system embedded in the on-site robot, corresponding one-to-one with the aforementioned interactive personalized information push method. Each module in the aforementioned interactive personalized information push system executes the respective steps in the aforementioned interactive personalized information push method.

[0084] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An interactive personalized information push method, characterized in that, The method includes: Step A1: Obtain user data, which includes user identity information, browsing history, offline consultation records, and qualification data authorized by third-party platforms; Step A2: Construct a dynamic user profile for the user based on the user data. The dynamic user profile includes demand tags, preference weight coefficients, and economic capacity assessment indicators. Step A3: Engage in real-time dialogue with users through natural language interaction, parse the user's voice or text input, and extract dynamic demand features and intent keywords; Step A4: Based on the dynamic user profile, the real-time extracted dynamic demand features, and the multi-dimensional product information in the product database, a priority matching algorithm is used to generate a recommended product list; Step A5: Based on the recommended product list, call the natural language engine to generate personalized push content. The push content includes the core features of the products, the matching degree with user needs, and interactive question-and-answer guidance. Step A6: Send the personalized content to the user terminal through multi-channel push, and monitor user clicks, feedback and subsequent interaction behavior in real time in order to update the dynamic user profile.

2. The interactive personalized information push method according to claim 1, characterized in that, In step A1, user data is acquired, and user identity information, historical browsing records, offline consultation records, and encrypted third-party platform authorization data are integrated from the user data. Based on cross-validation and data cleaning algorithm models, we conduct in-depth modeling of users' purchase intention strength, economic capacity threshold, and preference heatmap. We use natural language processing to perform semantic analysis on offline consultation recording text to extract implicit demand tags. We generate interactive reports containing personalized product matching index and purchase method simulation calculations through dynamic knowledge graph.

3. The interactive personalized information push method according to claim 1, characterized in that, In step A2, the construction of dynamic user profiles includes data cleaning, feature extraction, model calculation, and real-time updates. The system cleans and preprocesses user identity information, historical browsing records, offline consultation records, and purchase qualification data to remove duplicate and invalid data, and parses and classifies unstructured data through natural language processing. The system extracts key features from the cleaned data and generates purchase demand tags, preference weight coefficients, and economic capacity assessment indicators. In the process of constructing the dynamic user profile, the system uses machine learning algorithms to model user characteristics and monitors user behavior data in real time, dynamically adjusting the labels and weight coefficients in the profile.

4. The interactive personalized information push method according to claim 1, characterized in that, In step A3, the user's voice input is converted into text data through speech recognition, or the text information input by the user is received; natural language processing is used to segment the text, perform part-of-speech tagging and syntactic analysis to extract key information, and a pre-trained product intent recognition model is used in combination with knowledge graph and dialogue scenario templates to identify the user's purchase intent.

5. The interactive personalized information push method according to claim 1, characterized in that, In step A4, key features are extracted from the dynamic user profile, including purchase demand tags, preference weight coefficients, and economic capacity assessment indicators; the user profile is then optimized by combining the dynamic demand features extracted from real-time dialogue. The system extracts multi-dimensional product information from the product database, including the product's basic attributes, geographical location, supporting facilities, and market dynamics; the product information is stored in a structured manner and represented by feature vectors.

6. The interactive personalized information push method according to claim 5, characterized in that, The method further includes: during the matching process, the system uses a priority matching algorithm to calculate the matching degree between product features and user needs; the product is represented by a vector as: V = [v1, v2, v3, ..., v n In this context, v1 represents the style, v2 represents the price, v3 represents the feature, and so on, up to v n ; The specific method for calculating feature similarity is as follows: Where A(U,V) represents the feature similarity, U·V represents the dot product of the vectors, and ||U|| and ||V|| represent the magnitudes of the vectors.

7. The interactive personalized information push method according to claim 6, characterized in that, The method further includes: The matching score calculation weights are dynamically adjusted based on the user's preference weight coefficients and real-time demand characteristics. The specific matching score calculation method is as follows: Where B(U, V) represents the matching degree, β i U is represented as the weight of the i-th feature. i Let V be the i-th dimension of user needs. i This is represented as the i-th dimension of the product features; The products are sorted according to the adjusted matching degree to generate a recommended product list.

8. The interactive personalized information push method according to claim 6, characterized in that, In step A5, the system extracts the core selling points of each product from the recommended product list, including key information such as style, price, geographical location, and supporting facilities; the system combines the dynamic user profile and real-time demand characteristics to analyze the matching degree between the product and the user's needs; The NLG engine uses a pre-trained text generation model for the real estate sector, combined with templated and rule-based generation strategies, to automatically generate push content based on product characteristics and user needs.

9. The interactive personalized information push method according to claim 1, characterized in that, In step A6, the push channel is automatically selected based on the user's terminal type and usage habits; card-style messages containing product images, key selling points, and interactive buttons are sent to the communication terminal via WeChat and WhatsApp platforms; the system generates a structured email containing detailed product information and links and sends it to the email address. Send concise product summaries and redirect links to mobile devices via push notifications; The system generates illustrated card messages that highlight the product's core selling points and matching analysis, and embeds an interactive "Click to view details" button; in emails, the system generates detailed product descriptions, along with high-resolution images and interactive links; In the mobile app, the system will send a concise product summary via push notification and support one-click jump to the details page.

10. A wearable device, using an interactive personalized information push method as described in any one of claims 1-9, characterized in that, The device includes: The wearable device shares and transmits information with the on-site robot; it receives request information from multiple terminals, including request data and location information; it controls the on-site robot, sending tasks to the robot according to the request information so that the robot can reach the corresponding location; the wearable device also includes: The acquisition unit is used to acquire user data, which includes user identity information, historical browsing records, offline consultation records, and purchase qualification data authorized by third-party platforms. A construction unit is used to build a dynamic user profile for the user based on the user data. The dynamic user profile includes purchase demand tags, preference weight coefficients, and economic capacity assessment indicators. The interaction unit is used to engage in real-time dialogue with users through natural language interaction, parse the voice or text information input by users, and extract dynamic demand features and intent keywords. The recommendation unit is used to generate a list of recommended products based on the dynamic user profile and the dynamic demand features extracted in real time, combined with multi-dimensional product information in the product database, using a priority matching algorithm. The generation unit is used to generate personalized push content based on the recommended product list by calling a natural language engine. The push content includes the core selling points of the products, the matching degree with user needs, and interactive question-and-answer guidance. The update unit is used to send the personalized content to the user terminal through multi-channel push, and to monitor user clicks, feedback and subsequent interaction behaviors in real time to update the dynamic user profile.

Images