Recommendation content generation method and device based on user real-time data and storage medium

Abstract

The invention discloses a recommendation content generation method and device based on user real-time data and a storage medium, and the method comprises the steps: collecting first real-time data corresponding to a target user, and generating a first message list corresponding to the target user based on the first real-time data; searching first content information conforming to the interest of the target user according to the first message list; and generating second content information recommended to the target user based on the first content information by using a content creation model based on a large language model.

Description

Technical Field

[0001] The present application relates to the technical field of content recommendation, and in particular to a method, device and storage medium for recommending content based on real-time user data. Background Art

[0002] At present, with the rapid growth of the number of users and the rapid development of Internet information technology, content recommendation systems have become an important bridge between users and the Internet. Traditional content recommendation systems mainly rely on users' historical behavior data and predict users' interests and needs through collaborative filtering and content recommendation algorithms. Among them, users' historical behavior data includes, for example, user browsing records, user click records, and user purchase records.

[0003] In addition, with the popularity and continuous updating of instant messaging tools (such as chat applications and social media, etc.), users are using these instant messaging tools more and more. Therefore, the instant messaging tools contain a large amount of chat records, status information and interactive information related to users, and these chat records, status information and interactive information related to users contain real-time interest information and real-time emotional change information of users.

[0004] However, traditional recommendation content generation systems mostly use historical data information or data information crawled from the web to generate recommendation content. It is difficult to efficiently use the above-mentioned user instant information (for example, real-time interest information and real-time emotional change information) to generate recommendation content for users, resulting in the recommendation results being unable to reflect the user's latest interests and needs, and unable to provide users with more accurate and personalized recommendation services.

[0005] There is no effective solution to the technical problem that the traditional recommendation content generation system in the above-mentioned prior art is difficult to efficiently use the user's instant information to generate recommendation content for the user, resulting in the recommendation results failing to reflect the user's latest interests and needs. Summary of the invention

[0006] The embodiments of the present disclosure provide a method, device and storage medium for generating recommended content based on real-time user data, so as to at least solve the technical problem in the prior art that the traditional recommended content generation system is difficult to efficiently use the user's instant information to generate recommended content for the user, resulting in the recommendation results being unable to reflect the user's latest interests and needs.

[0007] According to one aspect of an embodiment of the present disclosure, a method for generating recommended content based on user real-time data is provided, comprising: collecting first real-time data corresponding to a target user, and generating a first message list corresponding to the target user based on the first real-time data; searching for first content information that matches the interests of the target user based on the first message list; and generating second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model.

[0008] According to another aspect of an embodiment of the present disclosure, a storage medium is further provided, the storage medium including a stored program, wherein when the program is running, a processor executes any one of the methods described above.

[0009] According to another aspect of an embodiment of the present disclosure, there is also provided a device for generating recommended content based on user real-time data, including: a real-time data acquisition module, used to acquire first real-time data corresponding to a target user, and generate a first message list corresponding to the target user based on the first real-time data; a first content information search module, used to search for first content information that meets the interests of the target user according to the first message list; and a second content information output module, used to generate second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model.

[0010] According to another aspect of an embodiment of the present disclosure, there is also provided a device for generating recommended content based on user real-time data, including: a processor; and a memory connected to the processor, for providing the processor with instructions for processing the following processing steps: collecting first real-time data corresponding to a target user, and generating a first message list corresponding to the target user based on the first real-time data; searching for first content information that meets the interests of the target user based on the first message list; and generating second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model.

[0011] The present application provides a method for generating recommended content based on user real-time data. First, first real-time data corresponding to a target user is collected, and a first message list corresponding to the target user is generated based on the first real-time data. Then, based on the first message list, first content information that meets the interests of the target user is searched. Finally, based on the first content information and using a content creation model based on a large language model, second content information for recommendation to the target user is output.

[0012] With reference to the above-mentioned contents, it can be known that in the present application, by collecting the first real-time data corresponding to the target user, the chat information and communication information generated by the target user using the instant messaging tool can be obtained. And when the first real-time data corresponding to the target user is collected, the first real-time data can be input into the pre-trained content recommendation model in the form of a first message list, thereby outputting the recommended content corresponding to the target user.

[0013] Therefore, since the present application utilizes the instant information of the target user to generate recommended content for the target user, the recommended content finally generated meets the interests and needs of the target user in the current state.

[0014] This solves the technical problem in the prior art that the traditional recommendation content generation system is difficult to efficiently use the user's real-time information to generate recommendation content for the user, resulting in the recommendation results being unable to reflect the user's latest interests and needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present application. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0016] Figure 1 is a hardware structure block diagram of a computing device for implementing the method according to Embodiment 1 of the present disclosure;

[0017] Figure 2 is a schematic diagram of a system for generating recommended content based on user real-time data according to Embodiment 1 of the present disclosure;

[0018] Figure 3 is a flow chart of the method for generating recommended content based on user real-time data according to Embodiment 1 of the present disclosure;

[0019] Figure 4 is a schematic diagram of training an agent and a first language model according to Embodiment 1 of the present disclosure;

[0020] Figure 5 is a modular schematic diagram of a device for generating recommended content based on user real-time data according to Embodiment 2 of the present disclosure;

[0021] Figure 6 It is a modular schematic diagram of the device for generating recommended content based on real-time user data according to Embodiment 3 of the present disclosure. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0024] Example 1

[0025] According to this embodiment, a method embodiment of generating recommended content based on user real-time data is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] The method embodiment provided in this embodiment can be executed in a mobile terminal, a computer terminal, a server or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computing device for implementing a method for generating recommended content based on real-time user data. Figure 1 As shown, the computing device may include one or more processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, the transmission device, and the input / output interface are connected to the processor via a bus. In addition, it may also include: a display, a keyboard, and a cursor control device connected to the input / output interface. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0027] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the computing device. As involved in the embodiments of the present disclosure, the data processing circuitry acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0028] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for generating recommended content based on user real-time data in the embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the method for generating recommended content based on user real-time data of the above-mentioned application program is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0029] The transmission device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computing device. In one example, the transmission device includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet wirelessly.

[0030] The display may be, for example, a touch screen liquid crystal display (LCD) that may enable a user to interact with a user interface of the computing device.

[0031] It should be noted that, in some optional embodiments, the above Figure 1 The computing device shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. Figure 1This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computing devices described above.

[0032] Figure 2 is a schematic diagram of a system for generating recommended content based on real-time user data according to this embodiment. Figure 2 As shown, the system includes: a real-time data collection and processing module, an intelligent agent, and a first language model. The real-time data collection and processing module is connected to the chat application server for collecting real-time data corresponding to the target user (e.g., chat content, status information, and interactive information, etc.) from the chat application server, and generating a message list corresponding to the target user based on the real-time data. The message list can be, for example, a real-time data list arranged in time sequence. In addition, reference Figure 2 As shown, in this embodiment, the intelligent agent and the first language model together constitute a content recommendation model.

[0033] The real-time data collection and processing module is respectively connected to the first language model and the intelligent agent. The first language model is used to predict the topics that the target user is interested in based on the message list corresponding to the target user. The intelligent agent is used to generate content recommended to the target user by combining the topics that the target user is interested in and the message list corresponding to the target user.

[0034] In addition, reference Figure 4 As shown, the system also includes a second language model, and the second language model is used to determine the user portrait corresponding to the online real user, the label information of the user portrait of the online real user, simulate the behavior of the virtual user and generate the corresponding behavior data sample, and determine the correlation between the user portrait of the online real user and the behavior data sample. The second language model is a pre-trained general multimodal large model.

[0035] Further, refer to Figure 2 The content shown, the intelligent agent also includes a user portrait determination module, a search module and a content generation module. Among them, the user portrait determination module is used to generate a user portrait of the target user based on the message list corresponding to the target user. The search module is connected in communication with the user portrait determination module, and is used to search from the Internet according to the user portrait of the target user and the topic of interest to the target user, and determine the multimodal content that meets the user portrait of the target user and the topic of interest to the target user. The content generation module is connected in communication with the search module, and is used to generate prompt information based on the multimodal content, and output the content recommended to the target user by using the content creation model and the prompt information.

[0036] It should be noted that the real-time data acquisition and processing module, the first language model and the intelligent agent in the system can all be applied to the hardware structure described above.

[0037] Under the above operating environment, according to the first aspect of this embodiment, a method for generating recommended content based on user real-time data is provided. The method comprises: Figure 2 The system implementation shown in . Figure 3 A schematic diagram showing the process of the method is shown in FIG. Figure 3 As shown, the method includes:

[0038] S302: collecting first real-time data corresponding to the target user, and generating a first message list corresponding to the target user based on the first real-time data;

[0039] S304: searching for first content information that matches the interests of the target user according to the first message list; and

[0040] S306: Generate second content information corresponding to the target user based on the first content information and using a content creation model based on a large language model.

[0041] Specifically, first, the real-time data collection and processing module collects the first real-time data corresponding to the target user from the chat application server, and then generates the first message list corresponding to the target user based on the collected first real-time data (S302). Among them, the first real-time data includes but is not limited to the chat content, status information and interactive information of the target user. And it is worth noting that the time interval for the real-time data collection and processing module to collect the first real-time data can be set in the system in advance by the management personnel, so that the real-time data collection and processing module can collect the first real-time data corresponding to the target user from the chat application server at a certain time interval. In addition, the real-time data collection and processing module can also collect the first real-time data corresponding to the target user from the chat application server from time to time, for example, as long as the first real-time data of the target user changes, the real-time data collection and processing module will collect the first real-time data from the chat application server.

[0042] For example, at 9:00-9:01, the real-time data collection and processing module collects the chat content and status information of the target user (ie, the first real-time data). Then, the real-time data collection and processing module generates the first message list shown in Table 1 below based on the collected chat content and status information of the target user.

[0043] Table 1

[0044]

[0045] As shown in reference table 1, the first message list includes real-time data corresponding to the time interval 8:30-8:31, and the real-time data in this period includes the chat content and interactive information of the target user; real-time data corresponding to the time interval 8:31-8:32, and the real-time data in this period includes the chat content, status information and interactive information of the target user; ...; real-time data corresponding to the time interval 9:00-9:01, and the real-time data in this period includes the chat content and status information of the target user. Therefore, according to the above content, it can be known that the first message list is a list arranged in chronological order and used to display the real-time data corresponding to the target user.

[0046] In addition, referring to the contents shown in Table 1 above, it can be known that the first message list corresponding to the first real-time data includes not only the first real-time data corresponding to the target user collected by the real-time data acquisition and processing module (i.e., the real-time data corresponding to the current time interval), but also the historical data corresponding to the target user collected by the real-time data acquisition and processing module. Among them, the historical data corresponds to the historical time period. And the first message list is collected and updated on the basis of the first real-time data corresponding to the target user and the message list of the previous time period. Therefore, it is worth noting that the first message list input to the content recommendation model includes not only the real-time data corresponding to the time interval 9:00-9:01, but also the real-time data before 9:00-9:01.

[0047] For example, Table 2 shows a message list corresponding to the previous period.

[0048] Table 2

[0049]

[0050] As shown in reference table 2, the message list corresponding to the previous time period includes the real-time data corresponding to the time period 8:30-8:31, and the real-time data of this time period includes the chat content and interactive information of the target user; the real-time data corresponding to the time period 8:31-8:32, and the real-time data of this time period includes the chat content, status information and interactive information of the target user; ...; the real-time data corresponding to the time period 8:59-9:00, and the real-time data of this time period includes the chat content and status information of the target user. Thus, when the real-time data acquisition and processing module acquires the first real-time data corresponding to the target user, the above table 2 is updated, and the above table 1 (i.e., the first message list) is generated.

[0051] Afterwards, the content recommendation model searches the Internet for first content information that matches the interests of the target user based on the first message list (S304). Specifically, the content recommendation model collects relevant content that matches the interests of the target user from the massive multimodal data on the Internet based on the first message list. The specific operation of how the content recommendation model searches for the first content information that matches the interests of the target user will be described later, so it will not be repeated here.

[0052] Finally, when the content recommendation model searches the Internet for the first content information corresponding to the target user, the content recommendation model generates the second content information corresponding to the target user based on the first content information and using the content creation model based on the large language model (S306). The second content information is the content recommended to the target user. Figure 2 As shown, when the content recommendation model searches the Internet for the first content information corresponding to the target user, the first content information is filled into the preset prompt template and the corresponding prompt information is generated. Afterwards, the content recommendation model inputs the prompt information into the content creation model, so that the content creation model can output the content recommended to the target user based on the prompt information. The above content will be described in detail later, so it will not be repeated here.

[0053] As described in the background technology, traditional recommendation systems rely too much on content-related behavioral data (for example, the click-through rate and collection rate of the content by other users, etc.) to make recommendations, resulting in content information related to users' real-time interests and real-time emotional changes being difficult to be accurately and timely recommended to users, thereby failing to provide more accurate and personalized recommendation services.

[0054] In view of this, the present application provides a method for generating recommended content based on user real-time data. And in the present application, by collecting the first real-time data corresponding to the target user, the chat information and communication information generated by the target user using the instant messaging tool can be obtained. And in the case of collecting the first real-time data corresponding to the target user, the first real-time data can be input into the pre-trained content recommendation model in the form of a first message list, thereby outputting the recommended content corresponding to the target user. In addition, the present application is completely independent of the recommendation with behavioral data (like rate, collection rate), and dynamically predicts the content of interest to the user through the content of the user's message list, thereby realizing an efficient recommendation algorithm for instant interest recommendation.

[0055] Therefore, since the present application utilizes the instant information of the target user to generate recommended content for the target user, the recommended content finally generated meets the interests and needs of the target user in the current state.

[0056] This solves the technical problem in the prior art that the traditional recommendation content generation system is difficult to efficiently use the user's real-time information to generate recommendation content for the user, resulting in the recommendation results being unable to reflect the user's latest interests and needs.

[0057] Optionally, the operation of outputting second content information corresponding to the target user based on the first content information and utilizing a content creation model based on a large language model includes: filling the first content information into a pre-set prompt template and generating prompt information; and inputting the prompt information into the content creation model and outputting the second content information corresponding to the target user.

[0058] Specifically, when the content recommendation model searches for the first content information from the Internet, the first content information is filled into a preset prompt template and prompt information is generated. After that, the content recommendation model further inputs the prompt information into a content creation model based on a large language model and receives the second content information (i.e., the content recommended to the target user) generated by the content creation model. Finally, the content recommendation model outputs the second content information corresponding to the target user.

[0059] For example, enter the prompt information in the content recommendation model: You are an article writing assistant, please write an article on the topic of "Philosophy and Psychology-Psychology-Positive Psychology-Theory and Practical Application of Positive Psychology" (format: first-level category-second-level category-entity-topic). The first user portrait of the target user is: age 28; gender male; occupation may be related to games, IT or marketing; living in China, possibly in Sichuan. The long-term interests of the target user are: long-term interest in horror mobile games and movies; long-term attention to hiking and related equipment; strong interest in ancient Chinese history and culture. The short-term interests of the target user are: recent interest in the matching of modern minimalist style bedroom furniture; certain interest in handmade crafts, especially dough processing; interest in the quality of olive oil and the best way to eat it. The searched related web pages (i.e., the first content information) are: the content included in web page 1, the content included in web page 2, and the content included in web page 3.

[0060] The content recommended to the target user by the content recommendation model is: "Taoism and Confucianism merge into positive psychology: modern application of ancient wisdom". In recent years, a wave of "positive psychology" has been quietly rising. From bestseller lists to hot topics on social media, from corporate training courses to school mental health education, positive psychology seems to be everywhere. So, what is positive psychology? How does it relate to the traditional Chinese culture we are familiar with, especially Taoism and Confucianism? ...In this fast-paced era, let us re-examine these ancient wisdoms. Perhaps, we will find the long-lost inner peace, the love of life, and the hope for the future. After all, true wisdom can stand the test of time.

[0061] Thus, by filling the first content information into a preset prompt template, corresponding prompt information is generated, and by utilizing the content creation model and based on the prompt information, recommended content that intersects the long-term / short-term interests in the target user portrait is output, thereby achieving the technical effect of personalized content recommendation.

[0062] Optionally, the operation of searching for the first content information that matches the interest of the target user based on the first message list includes: inputting the first message list into a pre-trained first language model, and using the first language model to predict a first interest topic corresponding to the target user; and inputting the first interest topic into a pre-trained intelligent agent, and using the intelligent agent and based on the first message list, searching for the first content information that matches the interest of the target user. Further optionally, the operation of inputting the first interest topic into a pre-trained intelligent agent, and using the intelligent agent and based on the first message list, searching for the first content information that matches the interest of the target user includes: inputting the first message list into a pre-trained intelligent agent, and using the intelligent agent to generate a first user portrait of the target user; and using the intelligent agent and based on the first interest topic and the first user portrait, searching for the first content information that matches the first interest topic and the first user portrait.

[0063] Specifically, when the real-time data acquisition and processing module generates a first message list corresponding to the target user, the first message list is input into the pre-trained first language model, and the first language model is used to predict the first interest topic corresponding to the target user. For example, the real-time data acquisition and processing module inputs the first message list shown in Table 1 above into the first language model, and then the first language model outputs the topic of interest to the target user (i.e., the first interest topic). Among them, the topic structure of the first interest topic can be, for example, "first interest category-secondary interest category-entity-topic content", so that the detailed topic structure can help the subsequent intelligent agent to accurately generate the second content information.

[0064] In addition, it is worth noting that since the first message list contains the first real-time data corresponding to the target user, and the first language model can predict the topic of interest to the target user at the current time and state based on the first real-time data corresponding to the target user, the topic of interest to the target user predicted by this application is not static, but is updated based on the change of time and data. Therefore, the first topic of interest with timeliness can more accurately represent the interest orientation of the target user at the current time and state, and help to provide accurate recommendation services for the target user in the future.

[0065] At the same time, when the real-time data collection and processing module inputs the first message list into the agent, the user portrait determination module in the agent can further determine the first user portrait of the target user according to the first message list. The agent is a pre-trained user personalized agent. The first user portrait includes but is not limited to gender, age, mobile phone model, occupation, geographic location, hobbies and occupation, etc.

[0066] In addition, it is worth noting that since the first message list contains the first real-time data corresponding to the target user, and the user portrait determination module can determine the first user portrait of the target user at the current time and state based on the first real-time data corresponding to the target user, the user portrait of the target user determined by this application is not static, but is updated based on the changes in time and data. Therefore, the user portrait with timeliness can more accurately locate the characteristics of the target user and help provide accurate recommendation services for the target user in the future.

[0067] Finally, when the intelligent agent receives the topic of interest to the target user (i.e., the first topic of interest) sent by the first language model and generates a first user portrait of the target user, it can search for relevant content (i.e., the first content information) from the Internet in combination with the topic of interest to the target user and the first user portrait of the target user.

[0068] Therefore, since this application utilizes the first language model to predict topics of interest to target users and utilizes the agent double verification mechanism, it can effectively improve the accuracy of content recommendations and provide target users with more accurate and personalized recommendation services.

[0069] Optionally, the operation of using an intelligent agent and, based on the first interest topic and the first user portrait, searching the Internet for first content information that matches the first interest topic and the first user portrait includes: using the intelligent agent to search the Internet for third content information that matches the first interest topic; and using tag information in the first user portrait to filter the third content information and determine the first content information, wherein the tag information includes the long-term interests and short-term interests of the target user.

[0070] Specifically, refer to Figure 2 As shown, when the user portrait determination module generates a first user portrait based on the first message list corresponding to the target user, and the first language model determines the first interest topic corresponding to the target user based on the first message list corresponding to the target user, the search module in the intelligent agent can first search from the Internet based on the first interest topic corresponding to the target user (i.e., the topic that the target user is interested in) and obtain the third content information.

[0071] For example, the first language model determines that the topic of interest to the target user is "Vietnamese history" based on the first message list corresponding to the target user. Therefore, the search module searches the Internet based on "Vietnamese history" and obtains content related to "Vietnamese history" (i.e., third content information). Among them, the content related to "Vietnamese history" includes: "What role does golf play in Vietnamese history", "The development of Vietnamese history", and "What is the relationship between Vietnamese history and American history", etc.

[0072] Afterwards, the search module uses the tag information in the first user portrait to filter the third content information and determine the first content information. The tag information in the first user portrait includes but is not limited to gender, age, mobile phone model, occupation, address, long-term interests, and short-term interests. It is worth noting that the search module does not use all the tags in the tag information of the first user portrait for filtering, but adaptively selects multiple tags from all tags for filtering. For example, the search module uses the gender, age, mobile phone model, occupation, address, and long-term interests in the first user portrait to filter the third content information. For another example, the search module can also use the gender, age, mobile phone model, occupation, address, and short-term interests in the first user portrait to filter the third content information.

[0073] For example, when the search module searches and obtains content related to "Vietnamese history" from the Internet, some tag information in the first user portrait (including the target user's long-term interests) shows that the target user is more interested in "golf". Therefore, the search module can filter the determined content related to "Vietnamese history" and retain the content related to "Vietnamese history" and further related to "golf". That is, retain the above "What role does golf play in Vietnamese history".

[0074] Thus, by utilizing the first interest topic to search for the third content information that meets the requirements from the Internet, and then screening the third content information according to the tag information in the first user portrait, and determining the operation of the first content information, the technical effect of ensuring that the first content information is strongly related to the user portrait of the target user, thereby ensuring that the second content information ultimately received by the target user meets the interests of the target user, and improving the target user's satisfaction with the ultimately generated second content information is achieved.

[0075] Optionally, the operation of using an intelligent agent and, based on the first interest topic and the first user portrait, searching the Internet for first content information that matches the first interest topic and the first user portrait includes: using tag information in the first user portrait to fine-tune the first interest topic and generate a second interest topic; and using the intelligent agent to search the Internet for first content information that matches the second interest topic.

[0076] Specifically, refer to Figure 2 As shown, when the user portrait determination module generates a first user portrait based on the first message list corresponding to the target user, and the first large language model determines the first interest topic corresponding to the target user based on the first message list corresponding to the target user, the search module in the intelligent agent can first use the label information in the first user portrait to fine-tune the first interest topic.

[0077] It is worth noting that the search module does not use all the tags in the tag information of the first user portrait to fine-tune the first interest topic, but adaptively selects multiple tags from all the tags to fine-tune the first interest topic. For example, the search module uses the gender, age, mobile phone model, occupation, address location and long-term interests in the first user portrait to fine-tune the first interest topic. For another example, the search module can also use the gender, age, mobile phone model, occupation, address location and short-term interests in the first user portrait to fine-tune the first interest topic.

[0078] For example, the first language model determines that the target user is interested in "American History" based on the first message list corresponding to the target user. And the search module determines that the target user is more interested in "baseball" based on part of the tag information in the first user portrait (including the short-term interest of the target user). Therefore, the search module fine-tunes "American History" and adjusts it to an interest topic related to "baseball" (i.e., the second interest topic). Among them, "American History" is fine-tuned, and the interest topic related to "baseball" is "What role does baseball play in American history?"

[0079] Afterwards, when the search module determines the second interest topic corresponding to the target user, the first content information matching the second interest topic is searched from the Internet.

[0080] For example, in the above-mentioned content, the search module determines that the second topic of interest is "What role does baseball play in American history?" Then the search module searches and obtains content related to "What role does baseball play in American history" (ie, first content information) from the Internet.

[0081] Thus, by searching the Internet for third content information that matches the first topic of interest, and using the tag information in the first user portrait to screen the third content information, and determining the operation of the first content information, the technical effect of ensuring that the first content information is strongly correlated with the user portrait of the target user, thereby ensuring that the second content information ultimately received by the target user matches the interests of the target user, and improving the target user's satisfaction with the ultimately generated second content information is achieved.

[0082] Optionally, it also includes: collecting second real-time data corresponding to the target user, and generating a second message list based on the second real-time data, wherein the second real-time data is used to indicate the updated first real-time data, and the second message list is used to indicate the updated first message list; when generating the second message list, inputting the second message list into the intelligent agent, and using the intelligent agent to generate a second user portrait of the target user; and using the second user portrait to update the first user portrait.

[0083] Specifically, refer to Figure 2 As shown, since the target user continuously updates the chat history, status information, and interaction information on the chat application server, the real-time data corresponding to the target user is not static, but changes over time.

[0084] Thus, when the chat records, status information, and interactive information of the chat application server corresponding to the target user are constantly changing, the real-time data collection and processing module collects the second real-time data corresponding to the target user from the chat application server, and generates a second message list based on the second real-time data. The second real-time data may be, for example, data updated relative to the first real-time data.

[0085] Table 3 shows a second message list generated based on the second real-time data.

[0086] Table 3

[0087]

[0088] Referring to the contents shown in Table 1 and Table 3, when the real-time data acquisition and processing module collects the second real-time data corresponding to the time period 9:01-9:02, the second real-time data is used to update the above Table 1 (i.e., the first message list), thereby generating the above Table 3 (i.e., the second message list).

[0089] And when the real-time data acquisition and processing module generates the second message list, the second message list is input into the intelligent agent, and the second user profile of the target user is generated by the intelligent agent. Then, the first user profile is updated by the second user profile. For example, when the real-time data acquisition and processing module generates the above Table 3, and the above Table 1 is updated by the above Table 3, the above Table 3 is further input into the user profile determination module in the intelligent agent. Thus, the user profile determination module can generate the second user profile based on the above Table 3.

[0090] Thus, by collecting the chat records, status information and interaction information generated by the target user on the chat application server in real time, and using the collected information to dynamically update the message list, and further using the updated message list to dynamically update the user portrait of the target user, the technical effect of being able to accurately generate a user portrait and ensuring that the ultimately generated second content information is in line with the changes in the interests of the target user is achieved.

[0091] Optionally, it also includes: pre-training a user portrait determination module for determining the first user portrait in the intelligent agent, wherein the operation of pre-training the user portrait determination module includes: determining first behavior data samples corresponding to multiple first real users, and generating a third message list based on the first behavior data samples; using a pre-trained second largest language model to determine the third user portrait corresponding to each first real user, thereby constructing multiple first sample pairs based on the third message list corresponding to each first real user and the third user portrait corresponding to each first real user; determining second behavior data samples corresponding to multiple second real users, and generating a fourth message list based on the second behavior data samples; determining fourth user portraits corresponding to each second real user by manual labeling, thereby constructing multiple second sample pairs based on the fourth message list corresponding to each second real user and the fourth user portrait corresponding to each second real user; and inputting the first sample pairs and the second sample pairs into the user portrait determination module, and training the user portrait determination module.

[0092] Specifically, Figure 4 Schematic diagram of training the user portrait determination module and the first language model of the intelligent agent according to the embodiment of the present application. Figure 4 As shown, before using the user portrait determination module to generate the second content information recommended to the target user, the user portrait determination module needs to be trained in advance.

[0093] Thus, first, the chat application server calls the first behavior data samples corresponding to the multiple first real users, and generates a third message list based on the first behavior data samples corresponding to each first real user. The first behavior data samples may be, for example, historical chat records, historical status information, and historical interaction information of the multiple first real users. The historical chat records, historical status information, and historical interaction information of the multiple first real users are stored in the chat application server.

[0094] Then, the second largest language model receives the third message list corresponding to each first real user sent by the chat application server, and further determines the third user portrait corresponding to each first real user. Among them, the prompt information input into the second largest language model includes: "Please build a third user portrait of the first real user based on the third message list of the first real user. Input: The third message list corresponding to each first real user, arranged in chronological order. Output: The description of the third user portrait of each first real user is specifically divided into three components. That is, output one: basic information of the first real user, including basic information such as age, gender and occupation of the first real user; output two: long-term interests of the first real user; output three: short-term interests of the first real user". Among them, the second largest language model can be, for example, a general large language model.

[0095] For example, a third message list is input into the second largest language model, wherein the third message list is shown in Table 4 below.

[0096] Table 4

[0097]

[0098]

[0099] Based on Table 4 above, the basic profile of the first real user is: age 28; gender male; occupation may be related to games, IT or marketing; living in China, possibly in Sichuan. The long-term interests of the first real user are: long-term interest in horror mobile games and movies; long-term attention to hiking and related equipment; strong interest in ancient Chinese history and culture. The short-term interests of the first real user are: recent interest in the matching of modern minimalist bedroom furniture; certain interest in handmade crafts, especially dough processing; interest in the quality of olive oil and the best way to eat it.

[0100] Furthermore, the second largest language model constructs a plurality of first sample pairs based on the third message list corresponding to each first real user and the third user portrait corresponding to each first real user. The format of the plurality of first sample pairs synthesized by the second largest language model is: [message list, user portrait].

[0101] In addition, the chat application server calls the second behavior data samples corresponding to the plurality of second real users, and generates a fourth message list based on the second behavior data samples corresponding to each second real user. The second behavior data samples may be, for example, historical chat records, historical status information, and historical interaction information of the plurality of second real users. The historical chat records, historical status information, and historical interaction information of the plurality of second real users are stored in the chat application server.

[0102] Then, the portraits of the second real users are annotated by manual annotation, so as to determine the fourth user portraits corresponding to the second real users.

[0103] Furthermore, the chat application server constructs a plurality of second sample pairs: [message list, user portrait] using the fourth message list and the fourth user image corresponding to each second real user.

[0104] Thus, the user portrait determination module can be trained using a sample set consisting of a plurality of first sample pairs and a plurality of second sample pairs.

[0105] For example, the above-mentioned multiple first sample pairs and multiple second sample pairs can be input into the user portrait determination module, and the following can be input into the intelligent agent: "You are an intelligent recommendation assistant, please analyze the third message list and analyze the corresponding third user portrait of the first real user". The third message list is shown in Table 4.

[0106] The user portrait determination module outputs: "Age 28; gender male; occupation may be related to games, IT or marketing; lives in China, possibly in Sichuan. The first real user's long-term interests are: long-term interest in horror mobile games and movies; long-term attention to hiking and related equipment; strong interest in ancient Chinese history and culture. The first real user's short-term interests are: recent interest in the combination of modern minimalist style bedroom furniture; certain interest in handmade craftsmanship, especially dough processing; interest in the quality of olive oil and the best way to consume it."

[0107] For example, the above-mentioned multiple first sample pairs and multiple second sample pairs are input into the user portrait determination module, and in this embodiment, the input of the user portrait determination module does not calculate the loss, and only the loss of the answer of the user portrait determination module is calculated. In which, the parameter of the user portrait determination module is θ1, and the training data set is in is the label mask, is the predicted output of the user profile determination module, expressed as a probability distribution. For the cross entropy loss function L with mask CE1 (θ1), can be expressed as:

[0108]

[0109] Where C represents the number of character categories. Representation sample The true label under category C (usually one-hot encoded, Indicates that the sample belongs to category C). Indicates that the user portrait determination module determines the input Predict the probability of belonging to class C. is the label mask, express efficient, express Invalid (i.e., skip loss computation for this sample).

[0110] In another training algorithm, Not just 0 or 1, in, Indicates that the input of the user portrait determination module is lossless. w1 represents the weight value corresponding to the long-term interest in the user portrait training result. w2 represents the weight value corresponding to the short-term interest in the user portrait training result. And so on. w n Indicates the weight value corresponding to the occupation in the user portrait training result.

[0111] For example, in And when w1=0.8, it means that the weight of long-term interest in the user portrait training result is 0.8. And when w2 = 0.9, it means that the weight of short-term interest in the user portrait training result is 0.9. And w n =1, it means that the weight of the occupation in the user portrait training result is 1.

[0112] Therefore, the importance of each user portrait tag information can be determined based on the weight value of each user portrait tag information in the above training results. And when the weight value is larger, the importance of the corresponding user portrait tag information is greater; when the weight value is smaller, the importance of the corresponding user portrait tag information is smaller.

[0113] Then, the user portrait determination module is fine-tuned according to the user portrait output by the user portrait determination module and the user portrait in the sample pair.

[0114] Thus, by using the second largest language model to synthesize multiple first sample pairs and artificially synthesizing multiple second sample pairs, and using the multiple first sample pairs and multiple second sample pairs to train the user portrait determination module, the technical effect of providing the necessary basis for subsequently providing the second content information to the target user is achieved.

[0115] Optionally, it also includes: pre-training the first large language model, wherein the operation of pre-training the first large language model includes: collecting corresponding different label information for each label item of the user portrait, and arranging and combining the collected label information of each label item to generate multiple fifth user portraits, and simulating and generating corresponding virtual users based on the fifth user portraits; using the second large language model and based on the virtual users, generating third behavior data samples and corresponding topics of interest, and generating a fifth message list based on the third behavior data samples, thereby constructing multiple third sample pairs based on the fifth message list and the corresponding topics of interest; generating a sixth message list based on fourth behavior data samples and topics of interest corresponding to multiple third real users, and constructing multiple fourth sample pairs based on the sixth message list and the corresponding topics of interest; and training the first large language model using the multiple third sample pairs and the multiple fourth sample pairs.

[0116] Specifically, refer to Figure 4 As shown, before using the first language model to generate topics of interest to the target user, the first language model needs to be trained in advance.

[0117] Therefore, first, the second largest language model collects label information corresponding to the label items of each user portrait online. The label items corresponding to each fifth user portrait include but are not limited to gender, age, mobile phone model used, occupation, geographic location, long-term interests and short-term interests. Different label information is collected for each label item. For example, for the gender label item, "male" and "female" are collected. For the age label item, different ages are collected as label information; for the mobile phone model used, different mobile phone models are collected as label information; and so on.

[0118] The different tag information of each tag item is arranged and combined, so as to simulate and generate multiple user portraits (ie, the fifth user portrait), wherein each generated user portrait corresponds to a virtual user.

[0119] For example, the tag information of the fifth user portrait 1 is gender 1, age 1, mobile phone model 1, occupation 1, geographic location 1, long-term interest 1, and short-term interest 1; the tag information of the fifth user portrait 2 is gender 2, age 2, mobile phone model 2, occupation 2, geographic location 2, long-term interest 2, and short-term interest 2. Thus, different fifth user portraits can be generated based on the arrangement and combination of each tag information, and corresponding virtual users can be generated.

[0120] Furthermore, in the case of generating multiple fifth user portraits, the second largest language model can be used to generate a fifth message list matching the fifth user portrait and an interesting topic corresponding to the fifth message list based on the fifth user portrait. Thus, the second largest language model further constructs multiple third sample pairs based on the fifth message list and the corresponding interesting topics.

[0121] This can be achieved by inputting prompt information into the second language model. For example, the second language model can be input with: "You are an intelligent recommendation assistant, please analyze the fifth message list of the virtual user and predict the next topic he wants to see". The fifth message list is shown in Table 5 below.

[0122] Table 5

[0123]

[0124] The topics of interest output by the first language model are: philosophy and psychology - psychology - positive psychology - the theory and practical application of positive psychology.

[0125] Finally, the second largest language model constructs multiple third sample pairs, where the format of the third sample pairs is: [message list, topic of interest to the user].

[0126] At the same time, after the real user comes online, the chat application server calls the fourth behavior data samples corresponding to each third real user, and generates a sixth message list and a corresponding topic of interest based on the fourth behavior data samples corresponding to each third real user. For example, for a series of N behaviors (such as chat messages) of a real user, the first N-1 behaviors can be used as a message list, and the Nth behavior can be used as a topic of interest, thereby constructing fourth sample pairs corresponding to multiple third real users respectively.

[0127] Finally, the first large language model may be trained using a sample set constructed by the plurality of third sample pairs and the plurality of fourth sample pairs.

[0128] Among them, the algorithm for training the first language model in the embodiment of the present application, the input information and output information are native modal information (for example, text, pictures, and videos, etc.). Compared with the traditional content recommendation system that relies on user behavior (for example, a posteriori features such as likes or collections) to assist in improving the accuracy of recommendations, the present application only relies on the continuity of the content to make recommendations. Furthermore, the first language model is based on the modeling paradigm of native modal information, and the parameters of the first language model can be hot-started from a general multimodal large model (ie, the second language model), and continues a large amount of knowledge that has been trained.

[0129] For example, the above-mentioned plurality of third sample pairs and the plurality of fourth sample pairs are input into the first large language model, and in this embodiment, the input loss of the first large language model is not calculated, and only the loss of the answer of the first large language model is calculated. Wherein, the parameter of the first large language model is θ2, and the training data set is in is the label mask, is the predicted output of the first language model, expressed as a probability distribution. For the cross entropy loss function L with mask CE2 (θ2), can be expressed as:

[0130]

[0131] Where C represents the number of character categories. Representation sample The true label under category C (usually one-hot encoded, Indicates that the sample belongs to category C). Indicates that the first language model is for input Predict the probability of belonging to class C. is the label mask, express efficient, express Invalid (i.e., skip loss computation for this sample).

[0132] Thus, by utilizing the second largest language model and constructing multiple third sample pairs based on cold start data, and constructing multiple fourth sample pairs using behavioral data samples corresponding to real online users, and using multiple third sample pairs and multiple fourth sample pairs to train the first largest language model, the technical effect of providing the necessary basis for subsequently providing the second content information to the target user is achieved.

[0133] Optionally, a sixth message list is generated according to fourth behavior data samples corresponding to multiple third real users, and based on the sixth message list, multiple fourth sample pairs are constructed, including: collecting fifth behavior data samples corresponding to multiple fourth real users, and using a second largest language model to analyze the correlation between user portraits of the multiple fourth real users and the corresponding fifth behavior data samples, wherein the number of the multiple fourth real users is greater than or equal to the number of the multiple third real users; when the correlation is greater than a preset threshold, selecting corresponding multiple fourth behavior data samples from the multiple fifth behavior data samples, and constructing a sixth message list based on the multiple fourth behavior data samples; and constructing multiple fourth sample pairs based on the sixth message list.

[0134] Specifically, refer to Figure 4 As shown, before the second largest language model constructs a plurality of fourth sample pairs based on the fourth behavior data samples respectively corresponding to the plurality of third real users, the second largest language model first collects fifth behavior data samples corresponding to the plurality of fourth real users. The number of the plurality of fourth real users is greater than or equal to the number of the plurality of third real users.

[0135] Then, the second largest language model determines the user portrait corresponding to each fourth real user, and determines the correlation between the user portraits of multiple fourth real users and the corresponding fifth behavior data samples. And from the multiple fifth behavior data samples, multiple fourth behavior data samples with correlation greater than a preset threshold are determined, and the multiple fourth behavior data samples with correlation greater than the preset threshold are retained, and the behavior data samples with correlation less than or equal to the preset threshold are eliminated.

[0136] Finally, the second largest language model constructs a plurality of fourth sample pairs according to the determined sixth message list corresponding to the plurality of fourth behavior data samples.

[0137] Thus, by performing data cleaning on the behavioral data samples corresponding to the real users and filtering out the behavioral data samples with low correlation with the user portraits of the real users, and retaining only the behavioral data samples with high correlation with the user portraits of the real users, the technical effect of improving the accuracy of the constructed multiple fourth sample pairs and thereby ensuring the accuracy of the trained first language model is achieved.

[0138] In addition, reference Figure 1 As shown, according to the second aspect of this embodiment, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0139] Therefore, according to this embodiment, the technical effect of providing users with more accurate and personalized recommendation services can be achieved.

[0140] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0141] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0142] Example 2

[0143] Figure 5 FIG. 5 shows a device 500 for generating recommended content based on user real-time data according to this embodiment, which corresponds to the method according to embodiment 1. Figure 5 As shown, the device 500 includes: a real-time data acquisition module 510, which is used to collect first real-time data corresponding to the target user, and generate a first message list corresponding to the target user based on the first real-time data; a first content information search module 520, which is used to search for first content information that meets the interests of the target user according to the first message list; and a second content information output module 530, which is used to generate second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model.

[0144] Optionally, the second content information output module 530 includes: a prompt information generation module, used to fill the first content information into a pre-set prompt template and generate prompt information; and a second content information output sub-module, used to input the prompt information into the content creation model and output the second content information for recommendation to the target user.

[0145] Optionally, the first content information search module 520 includes: a first interest topic prediction module, which is used to input the first message list into a pre-trained first language model, and use the first language model to predict a first interest topic corresponding to the target user; and a first content information search sub-module, which is used to input the first interest topic into a pre-trained intelligent agent, and use the intelligent agent and based on the first message list, search for the first content information that meets the interests of the target user.

[0146] Optionally, the first content information search submodule includes: a first user portrait generation module, used to input the first message list into a pre-trained intelligent agent, and use the intelligent agent to generate a first user portrait of the target user; and an intelligent agent search module, used to use the intelligent agent and, based on the first interest topic and the first user portrait, search for the first content information that matches the first interest topic and the first user portrait.

[0147] Optionally, the first content information search module 520 includes: a third content information search module, which is used to use an intelligent agent to search for third content information that matches the first interest topic from the Internet; and a screening module, which is used to use the tag information in the first user portrait to screen the third content information and determine the first content information, wherein the tag information includes the long-term interests and short-term interests of the target user.

[0148] Optionally, the first content information search module 520 includes: a fine-tuning module, which is used to use the tag information in the first user portrait to fine-tune the first interest topic and generate a second interest topic; and a first content information search sub-module, which is used to use an intelligent agent to search for the first content information that matches the second interest topic from the Internet.

[0149] Optionally, the device 500 also includes: a first data acquisition module, used to collect second real-time data corresponding to the target user, and generate a second message list based on the second real-time data, wherein the second real-time data is used to indicate the updated first real-time data, and the second message list is used to indicate the updated first message list; a second user portrait generation module, used to input the second message list into the intelligent agent when generating the second message list, and use the intelligent agent to generate a second user portrait of the target user; and a user portrait updating module, used to use the second user portrait to update the first user portrait.

[0150] Optionally, the device 500 also includes: a first model training module, which is used to pre-train the user portrait determination module for determining the first user portrait in the intelligent agent, wherein the first model training module includes: a second data acquisition module, which is used to determine the first behavior data samples corresponding to multiple first real users, and generate a third message list based on the first behavior data samples; a first sample pair construction module, which is used to determine the third user portrait corresponding to each first real user by using the pre-trained second largest language model, thereby constructing multiple first sample pairs based on the third message list corresponding to each first real user and the third user portrait corresponding to each first real user; a third data acquisition module, which is used to determine the second behavior data samples corresponding to multiple second real users, and generate a fourth message list based on the second behavior data samples; a second sample pair construction module, which is used to determine the fourth user portraits corresponding to each second real user by manual annotation, thereby constructing multiple second sample pairs based on the fourth message list corresponding to each second real user and the fourth user portrait corresponding to each second real user; and a first model training submodule, which is used to input the first sample pair and the second sample pair into the user portrait determination module, and train the user portrait determination module.

[0151] Optionally, the device 500 also includes: a second model training module, which is used to pre-train the first large language model, wherein the second model training module includes: a fifth portrait generation module, which is used to collect corresponding different label information for each label item of the user portrait, and arrange and combine the collected label information of each label item to generate multiple fifth user portraits; a third sample pair construction module, which is used to use the second large language model and generate a fifth message list and corresponding topics of interest based on the fifth user portrait, and construct multiple third sample pairs based on the fifth message list and the corresponding topics of interest; a fourth sample pair construction module, which is used to construct multiple fourth sample pairs based on the sixth message list and the corresponding topics of interest corresponding to multiple third real users; and a second model training submodule, which is used to train the first large language model using multiple third sample pairs and multiple fourth sample pairs.

[0152] Optionally, the fourth sample pair construction module includes: a correlation determination module, used to collect fifth behavior data samples corresponding to multiple fourth real users, and use the second largest language model to analyze the correlation between the user portraits of the multiple fourth real users and the corresponding fifth behavior data samples, wherein the number of the multiple fourth real users is greater than or equal to the number of the multiple third real users; a sixth message list construction module, used to select the corresponding multiple fourth behavior data samples from the multiple fifth behavior data samples when the correlation is greater than a preset threshold, and construct a sixth message list based on the multiple fourth behavior data samples; and a fourth sample pair construction sub-module, used to construct multiple fourth sample pairs based on the sixth message list.

[0153] Therefore, according to this embodiment, the technical effect of providing users with more accurate and personalized recommendation services can be achieved.

[0154] Example 3

[0155] Figure 6 FIG. 6 shows a device 600 for generating recommended content based on user real-time data according to this embodiment, which corresponds to the method according to embodiment 1. Figure 6 As shown, the device 600 includes: a processor 610; and a memory 620, which is connected to the processor 610 and is used to provide the processor 610 with instructions for processing the following processing steps: collecting first real-time data corresponding to a target user, and generating a first message list corresponding to the target user based on the first real-time data; searching for first content information that meets the interests of the target user according to the first message list; and generating second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model.

[0156] Optionally, the operation of generating second content information for recommendation to target users based on the first content information and utilizing a content creation model based on a large language model includes: filling the first content information into a pre-set prompt template and generating prompt information; and inputting the prompt information into the content creation model and outputting the second content information for recommendation to the target user.

[0157] Optionally, the operation of generating second content information corresponding to the target user based on the first content information and utilizing a content creation model based on a large language model includes: filling the first content information into a pre-set prompt template and generating prompt information; and inputting the prompt information into the content creation model and outputting the second content information corresponding to the target user.

[0158] Optionally, the operation of searching for the first content information that matches the interest of the target user based on the first message list includes: inputting the first message list into a pre-trained first language model, and using the first language model to predict a first interest topic corresponding to the target user; and inputting the first interest topic into a pre-trained intelligent agent, and using the intelligent agent and based on the first message list, searching for the first content information that matches the interest of the target user.

[0159] Optionally, the operation of using the agent and searching the Internet for first content information that matches the first interest topic and the first user portrait based on the first interest topic and the first user portrait includes: using the agent to search the Internet for third content information that matches the first interest topic; and

[0160] The third content information is screened using the tag information in the first user portrait, and the first content information is determined, wherein the tag information includes the long-term interests and short-term interests of the target user.

[0161] Optionally, the operation of using an intelligent agent and, based on the first interest topic and the first user portrait, searching the Internet for first content information that matches the first interest topic and the first user portrait includes: using tag information in the first user portrait to fine-tune the first interest topic and generate a second interest topic; and using the intelligent agent to search the Internet for first content information that matches the second interest topic.

[0162] Optionally, the device 600 also includes: collecting second real-time data corresponding to the target user, and generating a second message list based on the second real-time data, wherein the second real-time data is used to indicate the updated first real-time data, and the second message list is used to indicate the updated first message list; when generating the second message list, inputting the second message list into the intelligent agent, and using the intelligent agent to generate a second user portrait of the target user; and using the second user portrait to update the first user portrait.

[0163] Optionally, the device 600 also includes: pre-training the intelligent agent, wherein the operation of pre-training the user portrait determination module for determining the first user portrait in the intelligent agent includes: determining first behavior data samples corresponding to multiple first real users, and generating a third message list based on the first behavior data samples; using the pre-trained second largest language model to determine the third user portrait corresponding to each first real user, thereby constructing multiple first sample pairs based on the third message list corresponding to each first real user and the third user portrait corresponding to each first real user; determining second behavior data samples corresponding to multiple second real users, and generating a fourth message list based on the second behavior data samples; determining fourth user portraits corresponding to each second real user by manual labeling, thereby constructing multiple second sample pairs based on the fourth message list corresponding to each second real user and the fourth user portrait corresponding to each second real user; and inputting the first sample pairs and the second sample pairs into the user portrait determination module, and training the user portrait determination module.

[0164] Optionally, the device 600 also includes: pre-training the first large language model, wherein the operation of pre-training the first large language model includes: collecting corresponding different label information for each label item of the user portrait, and arranging and combining the collected label information of each label item to generate multiple fifth user portraits; using the second large language model and based on the fifth user portrait, generating a fifth message list and corresponding topics of interest, and constructing multiple third sample pairs based on the fifth message list and the corresponding topics of interest; according to the sixth message list and topics of interest corresponding to multiple third real users, respectively, and based on the sixth message list and the corresponding topics of interest, constructing multiple fourth sample pairs; and using the multiple third sample pairs and the multiple fourth sample pairs to train the first large language model.

[0165] Optionally, a sixth message list is generated according to fourth behavior data samples corresponding to multiple third real users, and based on the sixth message list, multiple fourth sample pairs are constructed, including: collecting fifth behavior data samples corresponding to multiple fourth real users, and using a second largest language model to analyze the correlation between user portraits of the multiple fourth real users and the corresponding fifth behavior data samples, wherein the number of the multiple fourth real users is greater than or equal to the number of the multiple third real users; when the correlation is greater than a preset threshold, selecting corresponding multiple fourth behavior data samples from the multiple fifth behavior data samples, and constructing a sixth message list based on the multiple fourth behavior data samples; and constructing multiple fourth sample pairs based on the sixth message list.

[0166] Therefore, according to this embodiment, the technical effect of providing users with more accurate and personalized recommendation services can be achieved.

[0167] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0168] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0169] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0170] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0171] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0172] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0173] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for generating recommended content based on real-time user data, characterized in that: include: Collecting first real-time data corresponding to a target user, and generating a first message list corresponding to the target user based on the first real-time data; Searching, according to the first message list, for first content information that matches the interest of the target user; as well as Based on the first content information and using a content creation model based on a large language model, second content information is generated for recommendation to the target user.

2. The method according to claim 1, characterized in that The operation of generating second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model includes: Filling the first content information into a preset prompt template and generating prompt information; and The prompt information is input into the content creation model, and second content information for recommendation to the target user is output.

3. The method according to claim 1, characterized in that The operation of searching, according to the first message list, for first content information that matches the interest of the target user includes: Inputting the first message list into a pre-trained first language model, and using the first language model to predict a first interest topic corresponding to the target user; and The first topic of interest is input into a pre-trained intelligent agent, and the intelligent agent is used to search for first content information that matches the interest of the target user based on the first message list, wherein The operation of inputting the first interest topic into a pre-trained intelligent agent, and using the intelligent agent and based on the first message list, searching for first content information that matches the interest of the target user includes: Inputting the first message list into a pre-trained intelligent agent, and using the intelligent agent to generate a first user profile of the target user; and Using the agent, and based on the first interest topic and the first user portrait, searching for first content information that matches the first interest topic and the first user portrait, wherein The operation of using the agent and searching the Internet for first content information that matches the first interest topic and the first user portrait according to the first interest topic and the first user portrait includes: Using the agent, searching the Internet for third content information that matches the first interest topic; and The third content information is screened using the tag information in the first user portrait, and the first content information is determined, wherein the tag information includes the long-term interests and short-term interests of the target user.

4. The method according to claim 3, characterized in that The operation of using the agent and searching the Internet for first content information that matches the first interest topic and the first user portrait according to the first interest topic and the first user portrait includes: Using the tag information in the first user portrait, fine-tuning the first interest topic and generating a second interest topic; and The agent is used to search for first content information that matches the second interest topic.

5. The method according to claim 4, characterized in that Also includes: Collecting second real-time data corresponding to the target user, and generating a second message list based on the second real-time data, wherein the second real-time data is used to indicate the updated first real-time data, and the second message list is used to indicate the updated first message list; When the second message list is generated, the second message list is input into the agent, and a second user portrait of the target user is generated by using the agent; as well as The first user portrait is updated using the second user portrait.

6. The method according to claim 3, characterized in that Also includes: Pre-training a user portrait determination module for determining the first user portrait in the agent, wherein the operation of pre-training the user portrait determination module includes: Determining first behavior data samples corresponding to a plurality of first real users, and generating a third message list based on the first behavior data samples; Determine a third user portrait corresponding to each first real user by using a pre-trained second language model, thereby constructing a plurality of first sample pairs based on a third message list corresponding to each first real user and a third user portrait corresponding to each first real user; Determining second behavior data samples corresponding to a plurality of second real users, and generating a fourth message list based on the second behavior data samples; Determine fourth user portraits corresponding to the respective second real users by manual labeling, thereby constructing a plurality of second sample pairs based on fourth message lists corresponding to the respective second real users and fourth user portraits corresponding to the respective second real users; and The first sample pair and the second sample pair are input into the user portrait determination module, and the user portrait determination module is trained.

7. The method according to claim 6, characterized in that Also includes: The first large language model is pre-trained, wherein the operation of pre-training the first large language model includes: Collecting corresponding different tag information for each tag item of the user portrait, and arranging and combining the collected tag information of each tag item to generate multiple fifth user portraits; Generate a fifth message list and corresponding topics of interest using the second largest language model and according to the fifth user portrait, and construct a plurality of third sample pairs according to the fifth message list and the corresponding topics of interest; According to sixth message lists and interested topics corresponding to a plurality of third real users respectively, and based on the sixth message lists and the corresponding interested topics, constructing a plurality of fourth sample pairs; and The first language model is trained using the plurality of third sample pairs and the plurality of fourth sample pairs, wherein The operation of generating a sixth message list according to fourth behavior data samples respectively corresponding to a plurality of third real users, and constructing a plurality of fourth sample pairs based on the sixth message list includes: collecting fifth behavior data samples corresponding to the plurality of fourth real users, and using the second largest language model to analyze the correlation between the user portraits of the plurality of fourth real users and the corresponding fifth behavior data samples, wherein the number of the plurality of fourth real users is greater than or equal to the number of the plurality of third real users; In the case where the correlation is greater than a preset threshold, selecting the corresponding plurality of fourth behavior data samples from the plurality of fifth behavior data samples, and constructing the sixth message list according to the plurality of fourth behavior data samples; and The plurality of fourth sample pairs are constructed according to the sixth message list.

8. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is run, the processor executes the method according to any one of claims 1 to 7.

9. A device for generating recommended content based on real-time user data, characterized in that: include: A real-time data collection module, used to collect first real-time data corresponding to a target user, and generate a first message list corresponding to the target user based on the first real-time data; A first content information search module, configured to search, according to the first message list, for first content information that matches the interests of the target user; as well as The second content information output module is used to generate second content information for recommendation to the target user based on the first content information and using a content creation model based on a large language model.

10. A device for generating recommended content based on real-time user data, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Collecting first real-time data corresponding to a target user, and generating a first message list corresponding to the target user based on the first real-time data; Searching, according to the first message list, for first content information that matches the interest of the target user; as well as Based on the first content information and using a content creation model based on a large language model, second content information is generated for recommendation to the target user.

Images