Universal intelligent personalized service method based on multi-modal large model and application

Abstract

The invention discloses a universal intelligent personalized service method based on a multi-modal large model and application. The method comprises the steps that first, second and third types of information are collected, the first type of information comprises interaction information, the second type of information comprises shooting information, equipment state information and time and place information, and the third type of information comprises personalized information and historical behavior data; identifying the current intention and style preference of the user; predicting an expected behavior of the user; generating service content, and pushing the service content to the user; and performing task planning to form a personalized task, and generating a personalized scheme. The technical scheme provided by the invention is combined with a system and an advanced large model technology, so that the defect that various existing intelligent products can only carry out specific tasks is overcome; on the basis of user intention recognition, prediction and active service of future behaviors of the user are further introduced, the service experience is further improved, and diversified and personalized service requirements are met.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent services, and particularly to a general intelligent personalized service method and application based on a multimodal large model. Background Art

[0002] With the rapid development of large language models (LLMs), the application of artificial intelligence in multiple fields has gradually increased, especially showing strong potential in providing personalized and customized services. Traditional automated systems often can only perform fixed tasks and lack the flexibility to respond to user needs and real-time status. Most of the existing technologies rely on rule presets and cannot effectively handle diverse and complex user needs. There are still significant limitations, especially in scenarios involving personalization, real-time feedback, and user interaction.

[0003] In fields such as health management and virtual assistants, how to effectively understand and respond to user needs through a system to achieve highly customized services has become an urgent problem to be solved. Existing automated services usually cannot adapt to changes in user needs in real time or perform intelligent processing through multimodal data. Therefore, a new type of system is needed that can integrate multiple information sources, make intelligent decisions through a large model, and perform personalized adjustments to provide more flexible and intelligent services.

[0004] In the prior art, although some automated devices can complete preset tasks in specific scenarios (such as voice assistants, health monitoring systems, etc.), they usually lack the ability to deeply understand user personalized needs and make dynamic adjustments. For example, existing virtual assistants can usually only respond to simple voice commands and cannot make dynamic adjustments based on the user's historical preferences or real-time feedback; although health monitoring devices can collect data, they cannot intelligently formulate long-term health management plans according to changes in personal health conditions. The understanding of user needs by many existing systems is still limited to a single modality or preset rules, lacking comprehensive analysis of multiple information sources such as voice, images, and historical feedback. Moreover, existing systems often only serve user needs and do not have an active service module; secondly, the application scenarios are also very single, without cross-scenario transfer learning.

[0005] The technical solution described in the Chinese patent "CN110960406A, a beauty instrument" automates the execution of specific makeup steps, such as applying and grooming, through the device, aiming to reduce user operation time and improve makeup efficiency. However, this technical solution has obvious limitations. One of its biggest drawbacks is the lack of flexible user interaction and personalized adjustment. Although the device can perform some basic makeup actions, due to its reliance on the preset actions of mechanical devices, it is easy to result in the makeup effect not being precisely adjusted according to the different needs of each user.

[0006] For example, personalized factors such as the user's skin color, facial features, and makeup habits are not fully considered in traditional beauty devices. For users with high personalized makeup needs, a single mechanical operation is difficult to meet their requirements for delicate makeup. In addition, this technical solution faces challenges in adapting to changing user needs during actual use. Users may have different makeup needs in different situations and moods, and the preset steps of mechanical devices often cannot adapt to these changes in a timely manner, failing to achieve true customized services. Summary of the Invention

[0007] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide a

[0008] To achieve the foregoing invention purpose, the technical solutions adopted by the present invention include:

[0009] In a first aspect, the present invention provides a general intelligent personalized service method based on a multimodal large model, which includes:

[0010] Collect at least a first type of information, a second type of information, and a third type of information, where the first type of information includes interaction information obtained by interacting with the user, the second type of information includes shooting information of the user and the environment, device status information, and time and location information, and the third type of information includes the user's personalized information and historical behavior data;

[0011] Use the multimodal large model to identify the user's current intention and style preference based on the first type of information and the third type of information;

[0012] Predict the user's expected behavior based on the historical behavior data, the current situation, and the current intention, where the current situation is generated based on the second type of information;

[0013] Generate service content based on the style preference and the expected behavior, and push the service content to the user;

[0014] Based on the historical behavior data, the current situation, and the current intention, perform task planning to form a personalized task, and use the multimodal large model to generate a personalized solution based on the personalized task.

[0015] In a second aspect, the present invention further provides a general intelligent personalized service system based on a multimodal large model, which includes:

[0016] An information collection module for collecting at least a first type of information, a second type of information, and a third type of information, where the first type of information includes interaction information obtained by interacting with the user, the second type of information includes shooting information of the user and the environment, device status information, and time and location information, and the third type of information includes the user's personalized information and historical behavior data;

[0017] An intention recognition module, configured to use a multi-modal large model to recognize the user's current intention and style preference based on the first type of information and the third type of information;

[0018] A behavior prediction module, configured to predict the user's expected behavior based on the historical behavior data, the current situation, and the current intention, wherein the current situation is generated based on the second type of information;

[0019] An active service module, configured to generate service content based on the style preference and the expected behavior, and push the service content to the user

[0020] A personalized solution module, configured to perform task planning to form a personalized task based on the historical behavior data, the current situation, and the current intention, and generate a personalized solution using the multi-modal large model based on the personalized task.

[0021] Furthermore, the system further includes:

[0022] A user feedback module, configured to continuously receive the user's feedback information and adjust the personalized solution

[0023] In a third aspect, the present invention further provides a readable storage medium, in which a computer program is stored, and when the computer program is run, it executes the steps of the above-mentioned general intelligent personalized service method.

[0024] Based on the above technical solutions, compared with the prior art, the beneficial effects of the present invention at least include:

[0025] The technical solution and system proposed by the present invention are combined with advanced large model technologies, making up for the defects of existing various intelligent products that can only perform specific tasks. After being combined with the large model, personalized and customized service planning can be completed by using the user's captured images and the captured images of items in the current environment. The user can also put forward requirements to the system and modify the service plan during the service process; on the basis of user intention recognition, the present invention further introduces the prediction of the user's future behavior and active service, further improving the service experience and meeting diverse and personalized service needs.

[0026] The above description is only an overview of the technical solution of the present invention. In order to enable those skilled in the art to understand the technical means of the present application more clearly and implement it according to the content of the specification, the following takes the preferred embodiments of the present invention and combines detailed drawings for description as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a schematic flowchart of the general intelligent personalized service method provided by a typical embodiment of the present invention;

[0028] Figure 2 It is a schematic diagram of the main composition structure of the general intelligent personalized service system provided by a typical implementation case of the present invention. Detailed implementation manners

[0029] In view of the deficiencies in the prior art, the inventors of this case, through long-term research and a large number of practices, have proposed the technical solution of the present invention. The following will further explain the technical solution, its implementation process, principles, etc.

[0030] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited by the specific embodiments disclosed below.

[0031] Moreover, relational terms such as "first" and "second" are only used to distinguish one component or method step with the same name from another, and do not necessarily require or imply any such actual relationship or order between these components or method steps.

[0032] The object of the present invention is to propose an intelligent personalized service system and method based on a multi-modal large model. The system and method can customize personalized services according to the needs of users and continuously optimize them as the user's habits change. By integrating various sensor information (such as visual, voice, tactile, etc. inputs) and combining advanced large model technology, this system can intelligently identify user needs and automatically complete service tasks. At the same time, the system will optimize according to the user feedback after each task, continuously improve the service experience, and gradually customize the most suitable personalized service plan for each user.

[0033] Based on the above object, the embodiment of the present invention first provides a general intelligent personalized service method based on a multi-modal large model (hereinafter referred to as this "method"), which includes the following steps:

[0034] Collect at least the first type of information, the second type of information, and the third type of information. The first type of information includes interaction information obtained by interacting with the user. The second type of information includes shooting information of the user and the environment, device status information, and time and location information. The third type of information includes the user's personalized information and historical behavior data;

[0035] Use the multi-modal large model to identify the user's current intention and style preference based on the first type of information and the third type of information;

[0036] Predict the user's expected behavior based on the historical behavior data, the current situation, and the current intention, where the current situation is generated based on the second type of information;

[0037] Generate service content based on the style preference and expected behavior, and push the service content to the user;

[0038] Based on the historical behavior data, current situation, and the current intention, perform task planning to form a personalized task, and generate a personalized solution using the multimodal large model based on the personalized task.

[0039] In some embodiments, the first type of information specifically includes the text, speech, and images input by the user.

[0040] In some embodiments, the second type of information specifically includes the user's facial image and the prop images and types around the user.

[0041] In some embodiments, the third type of information includes the user's historical feedback, preference settings, and preference types.

[0042] In some embodiments, the recognition process of the multimodal large model specifically includes the following process:

[0043] Analyze the features in the first type of information to generate the user's current intention;

[0044] Perform modeling based on the third type of information and analyze to generate the user's style preference.

[0045] In some embodiments, the method specifically includes:

[0046] Form an available item list according to the personalized task in combination with the second type of information;

[0047] Identify the user's emotional state based on the first type of information;

[0048] Formulate the personalized solution in combination with the available item list, emotional state, and personalized task.

[0049] In some embodiments, the method further includes:

[0050] Continuously receive the user's feedback information and adjust the personalized solution;

[0051] The feedback information includes drag operations, selection operations, and rating operations on the operations in the personalized solution.

[0052] In some embodiments, the method further includes:

[0053] After receiving the feedback information, update the third type of information.

[0054] In some implementation schemes, the method further includes: after the personalized plan is executed, receiving overall feedback from the user and updating the third type of information.

[0055] In some embodiments, the general intelligent personalized service method serves the first scenario and the second scenario, and the general intelligent personalized service method further includes:

[0056] The third type of information updated in the first scenario is updated to the second scenario through transfer learning.

[0057] The present invention introduces a user behavior prediction and proactive service module, which can predict the user's future needs and proactively provide services by analyzing the user's historical behavior data and current context. Some prior arts only mention mode switching based on interaction history, without involving user behavior prediction and proactive service generation. Although other prior arts mention personalized recommendations, their recommendation mechanisms are still based on the user's current explicit needs, lacking the ability to predict future needs and proactively provide services. The present invention designs a cross-scenario personalized transfer learning module, so that the user's preferences and behavior patterns in one scenario can be seamlessly transferred to another scenario.

[0058] In addition, none of the existing technologies mentioned cross-scenario transfer learning technology, and its personalized service is limited to a single scenario, and it is impossible to achieve seamless connection between multiple scenarios. The cross-scenario transfer learning module of the present invention can achieve seamless transition of personalized services between different scenarios, enhancing the flexibility and adaptability of the system.

[0059] Furthermore, the present invention introduces a user-participatory personalized customization module. Users can actively participate in the customization process of personalized services through simple interactions (such as dragging, selecting, and rating). For example, users can adjust the priority of recommended content or set the triggering conditions of services according to their preferences. However, personalized services in the current prior art are all automated recommendations, and users can only passively accept services, lacking the ability to actively participate and customize.

[0060] A second aspect of an embodiment of the present invention further provides a general intelligent personalized service system based on a multimodal large model (hereinafter referred to as the “system”), comprising:

[0061] An information collection module, used to collect at least first information, second information and third information, wherein the first information includes interaction information obtained through interaction with the user, the second information includes photographed information of the user and the environment, device status information and time and location information, and the third information includes personalized information and historical behavior data of the user;

[0062] An intention recognition module, configured to recognize the user's current intention and style preference based on the first type of information and the third type of information by using a multimodal large model;

[0063] A behavior prediction module, configured to predict the user's expected behavior based on the historical behavior data, the current situation, and the current intention, wherein the current situation is generated based on the second type of information;

[0064] An active service module, configured to generate service content based on the style preference and the expected behavior, and push the service content to the user;

[0065] A personalized solution module, configured to perform task planning to form a personalized task based on the historical behavior data, the current situation, and the current intention, and generate a personalized solution by using the multimodal large model based on the personalized task.

[0066] In some embodiments, the system further includes:

[0067] A user feedback module, configured to continuously receive the user's feedback information and adjust the personalized solution

[0068] The third aspect of the embodiments of the present invention further provides a readable storage medium, in which a computer program is stored, and when the computer program is run, it executes the steps of the general intelligent personalized service method provided in any of the above embodiments.

[0069] The technical solutions of the present invention will be further described in detail below through several embodiments in conjunction with the accompanying drawings. However, the selected embodiments are only used to illustrate the present invention and do not limit the scope of the present invention.

[0070] Embodiment 1

[0071] Figure 1 is a flowchart of the personalized beauty service provided in this embodiment, Figure 2 is the system framework corresponding to this method. Continuing to refer to Figure 1 As shown, the system based on the multimodal large model proposed in this embodiment includes the following steps:

[0072] (1) To ensure that diverse needs can be met, this embodiment first creates the following three types of information, and after obtaining the user's consent during system initialization, information collection is performed: In this embodiment, the three types of information created are set as an information set I = {I1, I2, I3}, where:

[0073] The first type of information I1 includes: text input by the user, voice information, and pictures. These are the main ways for the user to interact with the system. Through text and voice, the user can put forward their needs or questions, while pictures provide visual feedback on the user's current state for the system, such as personalized examples given, pictures of favorite personalized services, etc. Let the user's text information be T text , the voice information be T speech , and the picture information be T imag e.

[0074] The second type of information I2 includes: pictures of items on the desktop and pictures of the user's face. The images of items on the desktop help to identify available props or the environment, etc., ensuring that appropriate items can be selected during task execution. The pictures of the user's face are used to obtain the user's facial features, and through features such as facial expressions and makeup, a personalized plan that better meets the user's needs can be customized. Let the desktop image be D items , and the facial image be D face .

[0075] The third type of information I3 includes: the user's personalized information. This information comes from the user's historical feedback, preference settings, favorite types, etc., and can help the system make long-term adjustments to personalized services, making each personalized task more accurate and meeting the user's needs. Let the user's personalized information be P personal .

[0076] Send the first type of information and the third type of information to the multimodal large model for recognition processing. According to the multimodal fusion features of the first type of information, the large model will recognize the user's intention and make an answer or task plan based on the result.

[0077] User intention recognition: First, the first type of information and the third type of information will be sent into the multimodal large model for fusion processing. The multimodal large model analyzes the features in text, voice, and pictures to recognize the user's needs or intentions. This process not only includes understanding the user's questions but also covers modeling the user's preferences and historical data to predict the user's possible next needs. Let the recognition result of the user's intention be Then this process can be expressed by the following formula:

[0078]

[0079] Among them, f is the mapping function of the large model, which can extract features from different information sources and combine the user's historical data to predict their current needs. Based on the result of intention recognition, the system will formulate corresponding task plans. If the user's input is a question, the system will answer based on the recognition result of the model; if the user proposes a beauty task, the system will formulate specific task steps and plan the required products and tools.

[0080] Based on the recognition of user intent, this embodiment further introduces a user behavior prediction and proactive service module. This module analyzes the user's historical behavior data (such as daily activities, interaction records, preferences, etc.) and the current context (such as time, location, device status, etc.) to predict the user's future needs and proactively provide relevant services.

[0081] The core of user behavior prediction is to analyze the user's historical behavior data H and the current context C to predict the user's future needs.

[0082] D future = f predict (H, C)

[0083] where f predict is a user behavior prediction model that can extract features from historical behavior data and the current context and predict the user's future needs. For example, when the system detects that the user usually goes out to work at 8 am, it can push weather forecasts and traffic information in advance; when the user's historical shopping records show a preference for a certain product, the system can proactively recommend related products.

[0084] Based on the results of user behavior prediction, the system will generate corresponding proactive services. The process of generating proactive services can be expressed as:

[0085] S activate = f generate (D future )

[0086] f generate is a service generation function that can generate specific service content based on the predicted user needs. For example, when the system predicts that the user is about to go out, it can generate a proactive service containing weather forecasts, traffic information, and recommended routes; when the system predicts that the user has a shopping need, it can generate a proactive service containing product recommendations and discount information.

[0087] Task planning and response: After identifying the user intent, the system will perform task planning based on the recognition result. If the user's input is a question, the system will answer based on the recognition result of the model; if the user proposes a beauty task, the system will formulate specific task steps, plan the required products and tools, and prepare for subsequent operations. This process can be set as task planning T.

[0088] After receiving the personalized task, the output of the first type of information and the second type of information are sent to the multimodal large model for recognition processing.

[0089] Personalized Product Selection: Desktop image information will help the system identify all available items on the desktop. The system will automatically calculate the categories and quantities of items that require personalized customization and select a suitable solution for the user. If the desktop information is insufficient, the system can also obtain supplementary information through other means, such as asking the user or searching in historical data. Let the recognition result be P = {p1, p2,..., p n}, according to the user's needs, the system will automatically select the required product types and quantity n i , and the calculation formula is as follows:

[0090]

[0091] where g is the product recognition function, and n i is the quantity of each item. By calculating the categories and quantities of these products, the system can ensure that each product used in the task is both suitable and sufficient.

[0092] User Facial Customization Solution: The user's facial picture is used to analyze their facial features and identify key features such as the current mood, facial expression / facial structure, etc. Through these features, the large model will customize a suitable personalized solution, such as recommending the most applicable personalized service for the current mood and expression. The system can also consider the user's personalized information to further refine the recommended content. Let the user's facial picture be to generate a personalized service solution S face , and the formula is as follows:

[0093]

[0094] where h represents the mapping function for solution customization, outputting personalized service steps.

[0095] During the personalized service, the first type of information can be accepted, and the personalized task can be re-planned based on the user's facial picture in the first type of information and the second type of information.

[0096] Feedback and Re-planning: Users can provide real-time feedback through text or voice during the personalized service. If the user is not satisfied with the current solution, the system will re-plan the task according to the user's feedback. For example, the user may point out that a certain step does not meet their expectations, and the system will adjust the task according to the new feedback to ensure that the final result meets the user's needs. Let the real-time feedback be F feedback , and the feedback information will be used as input by the system to re-adjust and plan the task:

[0097] T adjusted = f adjust (T, F feedback )

[0098] where f adjustRepresents an adjustment function. The system is optimized through real-time feedback, enabling improvement in each service process to ensure that the personalized effect meets user requirements.

[0099] Real-time optimization: Through continuous feedback collection, the system will be optimized after each personalized service task. New user preferences, historical data, and other data will be added to the user's personalized information to further improve the accuracy of subsequent tasks.

[0100] After the completion of the personalized task, the system will receive the user's feedback (the first type of information), store it as the third type of information in the database, conduct comprehensive summarization, and continuously generate personalized customization for the user's preferences, etc. after each task.

[0101] User feedback storage and update: After the completion of the personalized task, the system will receive the user's feedback again. These feedbacks will be stored in the database as the third type of information. The user's feedback not only includes the satisfaction after the task completion but also evaluations on aspects such as the type and quantity of items and the suitability of the plan.

[0102] Database summarization and personalized adjustment: The feedback data after each personalized task will be summarized and analyzed and mined based on all the user's historical data. The system will identify factors such as the user's preferences and mood changes and adjust the future personalized service plan. Through continuously updated personalized data, the system can more accurately meet user requirements in each task. Let the personalized information for updating the database be Then the update formula for the feedback data is as follows:

[0103]

[0104] The feedback data after each personalized service task will be stored and analyzed based on the user's historical data. According to the user's characteristics, the system will make personalized adjustments:

[0105]

[0106] Among them, f opt is the optimization function, which optimizes the personalized service plan S through continuously updated personalized data final .

[0107] To break the limitations of a single scenario, this embodiment introduces a cross-scenario personalized transfer learning module. This module can seamlessly transfer the user's preferences and behavior patterns in one scenario to another scenario, realizing the connection of personalized services between multiple scenarios.

[0108] Let the user's preferences and behavior patterns in scenario A be P A The system transfers it to scenario B through transfer learning technology. The transfer process can be expressed as:

[0109] P B = f transfer (P A , C B )

[0110] where f transfer , is a transfer learning function that can combine the user preferences in scenario A with the context information C in scenario B B to generate a personalized service strategy applicable to scenario B.

[0111] To enhance the user's sense of participation and control, a user participation-based personalized customization module is added. This module allows users to actively participate in the customization process of personalized services through simple interactions (such as dragging, selecting, rating, etc.).

[0112] Suppose the user adjusts the personalized service through the interaction operation I, and the system will generate a customized service strategy S according to the user's input custom . The customization process can be expressed as:

[0113] S custom = f custom (S default , I)

[0114] where S default is the default personalized service strategy of the system, and f custom is a customization function that can adjust the default strategy according to the user's interaction input I to generate a customized service strategy that meets the user's needs. For example, the user can adjust the priority of recommended content by dragging or set the trigger condition of the service by selecting.

[0115] In the above embodiments, a general intelligent personalized service system based on a multimodal large model is mainly proposed. By comprehensively integrating various information sources such as text, voice, and images, it can provide highly personalized services for users. The whole process includes information acquisition, task recognition, solution formulation, and continuous update of user preference data. Through the intelligent processing and fusion of the large model, the system can accurately identify the user's intention. After the user interacts through inputs such as voice, text, and images, the system can identify their needs and automatically plan tasks based on the multimodal large model. This process not only includes answering the user's immediate questions but also planning the user's personalized tasks. The system can formulate the most suitable solution according to the user's specific needs, preferences, and real-time status to ensure the accurate execution of tasks.

[0116] By analyzing the user's historical behavior data (such as daily activities, interaction records, preferences, etc.) and the current context (such as time, location, device status, etc.), predict the user's future needs and proactively provide relevant services. For example, when the system predicts that the user is about to go out, it can push weather forecasts, traffic information or recommend travel routes in advance; when the user plans to shop, the system can automatically generate a shopping list and recommend discounted products. This proactive service model can not only significantly improve the user experience, but also reduce the user's operation burden and achieve truly intelligent services.

[0117] To break the limitations of a single scenario, this embodiment also introduces a cross-scenario personalized transfer learning module. This module can seamlessly transfer the user's preferences and behavior patterns in one scenario to another scenario, realizing personalized service connection between multiple scenarios.

[0118] To enhance the user's sense of participation and control, this embodiment also designs a user-participatory personalized customization module. This module allows users to actively participate in the customization process of personalized services through simple interactions (such as dragging, selecting, rating, etc.). For example, users can adjust the priority of recommended content, set the trigger conditions of services, or even customize the presentation mode of services according to their preferences. The system will make real-time adjustments according to the user's customization requirements to ensure that the services always meet the user's personalized needs.

[0119] Most importantly, the system of this embodiment can not only provide a single customization plan, but also provide a dynamic and continuously optimized experience for users through continuous learning and adjustment. Through continuous learning and adjustment of each task feedback, the system can deeply mine the user's preferences, skin status and makeup effects after each task, enabling the system to more accurately meet the user's needs in long-term use and improve the quality of personalized services. Through this system, users can not only obtain a solution that better meets their needs, but also enjoy the experience of continuous update and personalized adjustment.

[0120] It should be understood that the above embodiments are only used to illustrate the technical concept and features of the present invention, and their purpose is to enable those who are familiar with this technology to understand the content of the present invention and implement it accordingly, and cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit and essence of the present invention should be covered within the protection scope of the present invention.

Claims

1. A general intelligent personalized service method based on a multimodal large model, characterized in that, Including: Collect at least the first type of information, the second type of information, and the third type of information. The first type of information includes interaction information obtained through interaction with the user. The second type of information includes shooting information of the user and the environment, device status information, and time and location information. The third type of information includes the user's personalized information and historical behavior data; Use a multi-modal large model to identify the user's current intention and style preference based on the first type of information and the third type of information; Predict the user's expected behavior based on the historical behavior data, the current situation, and the current intention, where the current situation is generated based on the second type of information; Generate service content based on the style preference and the expected behavior, and push the service content to the user; Based on the historical behavior data, the current situation, and the current intention, perform task planning to form a personalized task, and use the multi-modal large model to generate a personalized solution based on the personalized task.

2. The general intelligent personalized service method according to claim 1, wherein The first type of information specifically includes text, voice, and images input by the user; And / or, the second type of information specifically includes the user's facial image and the types and images of props around the user; And / or, the third type of information includes the user's historical feedback, preference settings, and preference types.

3. The general intelligent personalized service method according to claim 1, wherein The recognition process of the multi-modal large model specifically includes: Analyze the features in the first type of information to generate the user's current intention; Perform modeling based on the third type of information and analyze to generate the user's style preference.

4. The general intelligent personalized service method according to claim 1, wherein Specifically including: Form a list of available items according to the personalized task in combination with the second type of information; Identify the user's emotional state based on the first type of information; Combine the list of available items, the emotional state, and the personalized task to draw up the personalized solution.

5. The general intelligent personalized service method according to claim 4, wherein Also including: Continuously receive the user's feedback information and adjust the personalized solution; The feedback information includes drag operations, selection operations, and rating operations on the operations in the personalized solution.

6. The general intelligent personalized service method according to claim 5, wherein, Also including: After receiving the feedback information, update the third type of information; And / or, when the personalized solution is executed, receive the overall feedback of the user and update the third type of information.

7. The general intelligent personalized service method according to claim 1, wherein The general intelligent personalized service method serves the first scenario and the second scenario. The general intelligent personalized service method also includes: Update the third type of information updated in the first scenario to the second scenario through transfer learning.

8. A general intelligent personalized service system based on a multimodal large model, characterized in that, Including: An information collection module for collecting at least the first type of information, the second type of information, and the third type of information. The first type of information includes interaction information obtained through interaction with the user. The second type of information includes shooting information of the user and the environment, device status information, and time and location information. The third type of information includes the user's personalized information and historical behavior data; An intention recognition module for using a multi-modal large model to identify the user's current intention and style preference based on the first type of information and the third type of information; A behavior prediction module for predicting the user's expected behavior based on the historical behavior data, the current situation, and the current intention, where the current situation is generated based on the second type of information; An active service module for generating service content based on the style preference and expected behavior and pushing the service content to the user; A personalized solution module for performing task planning to form personalized tasks based on the historical behavior data, current context, and current intention, and generating personalized solutions using the multimodal large model based on the personalized tasks.

9. The general intelligent personalized service system according to claim 8, characterized in that It further includes: A user feedback module for continuously receiving user feedback information and adjusting the personalized solution.

10. A readable storage medium, characterized in that, A computer program is stored in the readable storage medium, and when the computer program is run, it executes the steps of the general intelligent personalized service method described in any one of claims 1-7.

Images