An information recommendation method and device

Abstract

The application discloses an information recommendation method and device, and relates to the technical field of computers. A specific embodiment of the information recommendation method comprises the following steps: obtaining a user preference description text of a target user and to-be-recommended information data; inputting the user preference description text and the to-be-recommended information data into a pre-trained information recommendation prediction model, wherein the information recommendation prediction model outputs an information recommendation prediction result corresponding to each to-be-recommended information; and performing information recommendation on the target user based on the information recommendation prediction result corresponding to each to-be-recommended information. The information recommendation prediction model can be used to perform accurate prediction according to the user preference description text of the target user and the to-be-recommended information data, so that the information recommendation result is more in line with user demand, and user experience is improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to an information recommendation method and apparatus. Background Technology

[0002] With the continuous advancement of internet technology, people are increasingly relying on internet information platforms to obtain various information. Currently, rule-based methods are mainly used to recommend information, and these rules are generally based on editorial judgments or simple business logic.

[0003] In the process of realizing this invention, the inventors discovered at least the following problems in the related technology:

[0004] Information recommendations that do not meet user needs result in a poor user experience. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an information recommendation method and apparatus that can make accurate predictions based on the user preference description text of the target user and the information data to be recommended, using an information recommendation prediction model, so that the information recommendation results are more in line with user needs and improve user experience.

[0006] To achieve the above objectives, according to one aspect of the present invention, an information recommendation method is provided, comprising:

[0007] Obtain user preference descriptions and recommended information data from target users;

[0008] The user preference description text and the information to be recommended are input into the information recommendation prediction model, and the information recommendation prediction model outputs the information recommendation prediction result for each piece of information to be recommended.

[0009] Information recommendations are made to the target user based on the information recommendation prediction results corresponding to each piece of information to be recommended.

[0010] Furthermore, the information recommendation prediction model is obtained through the following steps:

[0011] Natural language processing is performed on the acquired user profile data and historical browsing data to generate user preference description text;

[0012] A training sample set is constructed based on user preference description text, acquired information recommendation data, and information recommendation annotation results;

[0013] The information recommendation prediction model is trained based on the training sample set.

[0014] Furthermore, the acquired user profile data and historical browsing data are processed using natural language conversion to generate user preference description text, including:

[0015] Based on the acquired user profile data and historical browsing data, the characteristics of discrete data and text data are determined;

[0016] By fusing features from discrete data and text data, user preference description text is generated.

[0017] Furthermore, after training the information recommendation prediction model based on the training sample set, it also includes:

[0018] The information recommendation prediction model is deployed to the online environment to monitor online change data in real time; the online change data includes user profile change data, historical browsing change data, and information change data.

[0019] The information recommendation prediction model is adjusted and updated based on changes in online data.

[0020] Furthermore, a training sample set is constructed based on user preference description text, acquired information recommendation data, and information recommendation annotation results, including:

[0021] The user preference description text and information recommendation data are combined to form input features;

[0022] The information recommendation annotation results are mapped to the input features to construct a training sample set.

[0023] Furthermore, a pre-trained large language model is used to perform natural language conversion processing on the acquired user profile data and historical browsing data;

[0024] The information recommendation prediction model is trained based on the training sample set, including: training the newly added parameters of the pre-trained large language model using the low-rank adaptation algorithm based on the training sample set, or training all parameters of the pre-trained large language model using the full-parameter fine-tuning algorithm based on the training sample set.

[0025] According to a second aspect of the present invention, an information recommendation device is provided, comprising:

[0026] The acquisition module is used to acquire the user preference description text of the target user and the information to be recommended;

[0027] The prediction module is used to input user preference description text and information to be recommended into the information recommendation prediction model. The information recommendation prediction model outputs the information recommendation prediction result for each piece of information to be recommended.

[0028] The recommendation module is used to recommend information to target users based on the information recommendation prediction results corresponding to each piece of information to be recommended.

[0029] According to a third aspect of the present invention, an electronic device is provided, comprising:

[0030] One or more processors;

[0031] Storage device for storing one or more programs.

[0032] When one or more programs are executed by one or more processors, the one or more processors implement the methods of any of the above embodiments.

[0033] According to a fourth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method of any of the above embodiments.

[0034] According to a fifth aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method of any of the above embodiments.

[0035] One embodiment of the above invention has the following advantages or beneficial effects: by inputting user preference description text and information data to be recommended into the information recommendation prediction model, the information recommendation prediction model outputs the information recommendation prediction result corresponding to each piece of information to be recommended; information is recommended to the target user based on the information recommendation prediction result corresponding to each piece of information to be recommended; it can make accurate predictions using the information recommendation prediction model based on the user preference description text and information data to be recommended, so that the information recommendation results are more in line with user needs and improve the user experience.

[0036] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0037] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0038] Figure 1 This is a schematic diagram of the main flow of the information recommendation method according to an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of the main flow of an information recommendation method according to an optional embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of the main flow of an information recommendation method according to another optional embodiment of the present invention;

[0041] Figure 4 This is a schematic diagram of the main modules of the information recommendation device according to an embodiment of the present invention;

[0042] Figure 5 This is an exemplary system architecture diagram in which embodiments of the present invention can be applied;

[0043] Figure 6 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation

[0044] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0045] It should be noted that the acquisition, storage, and application of personal information involved in the embodiments of the present invention comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0046] With the continuous advancement of internet technology, people are increasingly reliant on online information platforms to obtain various information. Currently, rule-based methods are mainly used to recommend information, and these rules are generally based on editorial judgments or simple business logic. Existing algorithmic models have high modeling costs, inaccurate calculation results, and information recommendations that do not meet user needs lead to a poor user experience.

[0047] In view of this, according to a first aspect of the present invention, an information recommendation method is provided.

[0048] Figure 1 This is a schematic diagram illustrating the main flow of the information recommendation method according to an embodiment of the present invention. Figure 1 As shown, the information recommendation method according to an embodiment of the present invention mainly includes the following steps S101 to S103.

[0049] Step S101: Obtain the user preference description text of the target user and the information data to be recommended.

[0050] Step S102: Input the user preference description text and the information to be recommended into the information recommendation prediction model. The information recommendation prediction model outputs the information recommendation prediction result for each piece of information to be recommended.

[0051] Step S103: Recommend information to the target user based on the information recommendation prediction result corresponding to each piece of information to be recommended.

[0052] When recommending information, the system first acquires the target user's preference description text and the information to be recommended. The preference description text is generated through natural language processing of user profile data and historical browsing data, detailing the user's interests, behaviors, and preferences. The information to be recommended includes feature information of various types of information to be recommended to the user, such as title, content, and publication time. Both the preference description text and the information to be recommended are input into the information recommendation prediction model, which effectively understands the relationship between user preferences and information content. The model can generate a corresponding information recommendation prediction result for each piece of information to be recommended. This result can be expressed as a rating or ranking, reflecting the degree of matching between each piece of information and the target user's preferences, i.e., the priority of the information being recommended. Based on the recommendation prediction results for each piece of information to be recommended, the system recommends information to the target user. The system prioritizes displaying information that best matches the user's preferences, thereby improving the relevance of the recommendations and user satisfaction. Through this process, the system can dynamically provide personalized information recommendations to users, meeting their diverse needs.

[0053] This invention provides an embodiment that obtains user preference description text and recommended information data of target users; inputs the user preference description text and recommended information data into an information recommendation prediction model, and the information recommendation prediction model outputs an information recommendation prediction result for each piece of information to be recommended; and recommends information to target users based on the information recommendation prediction result for each piece of information to be recommended; thus, the information recommendation results are more in line with user needs, improving the user experience.

[0054] Figure 2 This is a schematic diagram of the main flow of an information recommendation method according to an optional embodiment of the present invention. Figure 2 As shown, the information recommendation prediction model is obtained through the following steps S201 to S203.

[0055] Step S201: Perform natural language conversion processing on the acquired user profile data and historical browsing data to generate user preference description text.

[0056] User profile data comprises various characteristic information related to a user, including basic information (such as age, gender, and region), interests, consumption habits, social network behavior, and device usage. Historical browsing data refers to a user's internet browsing behavior record within a specific time period, including information viewed, articles clicked, dwell time, reading frequency, and content saved and shared. This data can be obtained from various sources, such as user registration information, online behavior records, and social media interactions.

[0057] Natural language processing (NLP) is used to generate user preference descriptions from acquired user profile and browsing history data. This can be achieved through template-based generation. First, the user profile and browsing history data undergo structured analysis to extract key user characteristics and behavioral information. Then, based on predefined language templates, these characteristics and behavioral information are filled into the appropriate positions. For example, templates can be set for fields such as "age," "gender," and "interests," such as "This user is [age] years old [gender] and shows a high interest in [interests] type of information." Furthermore, if a user frequently browses a certain type of information within a specific time period, a more detailed text description can be generated, such as "Recently, he / she has paid close attention to [specific content] and has read related reports multiple times." This template-based generation method is simple and intuitive, suitable for processing common, structured information. Alternatively, a natural language processing model can also be used for generation. First, user profile data and browsing history data are input into a pre-trained natural language processing model. These models can be large pre-trained language models. The model can automatically generate natural language descriptive text based on the input data. For example, for a user who prefers technology-related information, the model can generate something like, "The user shows a sustained interest in the latest developments in the technology field, especially articles about artificial intelligence and big data." The user interest description text generated in this way is more natural and flexible, can capture subtle differences in user behavior, and is suitable for processing more complex and diverse user data.

[0058] User preference text is a textual description of a user's interests, hobbies, needs, and other preferences. This user preference text can provide a basis for recommendation systems, advertising, personalized content generation, and other scenarios, enabling the system to more accurately match users' needs and interests. For example, based on user preference text, "This user has a continuous interest in technology news and the latest smart devices," the system can recommend relevant news content or new product information, thereby improving user experience and satisfaction.

[0059] Step S202: Construct a training sample set based on the user preference description text, the obtained information recommendation data, and the information recommendation annotation results.

[0060] Information recommendation data refers to the various information recommended by the system to users, including the content, title, tags, publication time, author, and category of the information. Information recommendation annotation results are based on actual user feedback or manual annotation, used to indicate the recommendation effectiveness or relevance of each piece of information recommended by the system. For example, whether a user clicked, read, saved, or liked a piece of information, or whether experts scored the suitability of that piece of information.

[0061] Specifically, the system matches user preference descriptions with news recommendation data, selecting potentially interesting information as positive samples and unmatched information as negative samples. Combining this with news recommendation annotations, information actually clicked or read by the user is marked as positive samples, while information not clicked or skipped is marked as negative samples. This method constructs a training set containing both positive and negative samples to train the model, thereby improving the recommendation system's ability to accurately predict user preferences. It utilizes actual user behavior for annotation, making the training sample set more closely reflect real user preferences. Furthermore, the news recommendation annotations can be combined with the news recommendation data to generate a rated training sample set. High-rated information is used as positive samples, and low-rated or irrelevant information as negative samples. This generates a high-quality training set, ensuring the model learns the complex relationship between user preferences and news content during training. The annotation results are highly accurate, effectively improving the model's recommendation performance.

[0062] Step S203: Train the information recommendation prediction model based on the training sample set.

[0063] A news recommendation prediction model is trained using a training sample set. This set can be input into a machine learning or deep learning model for training. The model can be of various types, such as a neural network-based recommendation algorithm, a decision tree model, or a hybrid model combining user history and content features. During training, the model continuously adjusts its internal parameters, gradually approximating the labeled results, thus learning how to predict news recommendation effectiveness based on user preference descriptions and news recommendation data. After training, the model can be validated and tested on unseen data to evaluate its generalization ability and prediction accuracy. If the test results meet preset standards, the trained model is used as a news recommendation prediction model in real-world applications. In practical applications, the news recommendation prediction model can predict the recommendation score or probability of each news item in real time based on new user preference descriptions and news recommendation data, thereby providing users with personalized news recommendation services.

[0064] This invention generates user preference description text by performing natural language conversion on the acquired user profile data and historical browsing data; a training sample set is constructed based on the user preference description text, the acquired information recommendation data, and the information recommendation annotation results; and an information recommendation prediction model is trained based on the training sample set. This allows for direct conversion of natural language descriptions, eliminating the need for tedious data cleaning, feature mining, and feature engineering steps during the modeling process, thus reducing the time cost of modeling and improving the accuracy of the algorithm model.

[0065] Figure 3 This is a schematic diagram of the main flow of an information recommendation method according to another optional embodiment of the present invention. Figure 3 As shown, the acquired user profile data and historical browsing data are processed using natural language conversion to generate user preference description text, including:

[0066] Step S301: Determine discrete data features and text data features based on the acquired user profile data and historical browsing data.

[0067] Step S302: Feature fusion is performed on discrete data features and text data features to generate user preference description text.

[0068] In the natural language processing (NLP) of the acquired user profile data and historical browsing data, the first step is to extract and classify features from both. User profile data includes basic user information such as age, gender, geographic location, and occupation; these are mostly discrete data features. Historical browsing data contains past reading behaviors, including accessed news titles, content summaries, and reading times; these are mostly textual data features. These data are initially processed, separating them into discrete and textual data features. Feature fusion is then performed on these discrete and textual features. This fusion process can be achieved in various ways, such as directly concatenating discrete and textual features, or learning the correlation between them through attention mechanisms. This fusion process not only preserves the user's basic information but also incorporates their past behavioral data, forming a comprehensive description of user preferences. The user preference description text generated in this way includes not only static features (such as gender and age) but also dynamic features (such as past browsing habits), providing comprehensive user preference information for subsequent recommendation systems.

[0069] In addition, a large language model can be used to perform natural language conversion on user profile data and historical browsing data. By inputting the user profile data and historical browsing data into the large language model, the large language model can directly integrate and understand the user profile data and information browsing data, bridging the gap between tabular discrete data features and text features. At the same time, the large language model has a deeper understanding of feature fusion than the traditional feature cross-fusion understanding, ensuring the final recommendation effect.

[0070] Optionally, after training the information recommendation prediction model based on the training sample set, the method further includes: deploying the information recommendation prediction model to an online environment to monitor online change data in real time; wherein, the online change data includes user profile change data, historical browsing change data, and information change data; adjusting the information recommendation prediction model based on the online change data, and updating the information recommendation prediction model.

[0071] After deploying the information recommendation prediction model to the online environment, various online change data are monitored and captured in real time. For example, continuous monitoring of user profile changes, including updates to user age, interests, and occupation, is conducted. Simultaneously, historical browsing data is monitored to capture users' latest behavior on the platform, such as newly read articles, clicked links, and recently followed topics. Furthermore, newly published information content on the platform is tracked in real time, capturing information change data such as the publication of new articles, and the modification or deletion of information. After capturing this online change data, the information recommendation prediction model is retrained or fine-tuned based on this data to better adapt to users' latest preferences and the latest information content. The updated model is then deployed back to the online environment to continue providing users with accurate information recommendation services. This continuous updating mechanism ensures that the model can promptly reflect changes in user behavior and information content, improving recommendation accuracy and user satisfaction. In this way, the system can dynamically adapt to environmental changes, maintaining the efficiency and adaptability of the recommendation system. At the same time, it can provide reasonable explanations by observing the input and output. This explanatory ability can reveal the key factors that the information recommendation prediction model relies on when predicting recommendation scores, such as changes in users' specific preferences or the special content characteristics of a certain piece of information. This transparency is not available in traditional recommendation algorithms, which not only helps to improve the accuracy of recommendations, but also enhances users' trust in the recommendation results.

[0072] Optionally, a training sample set is constructed based on the user preference description text, the acquired information recommendation data, and the information recommendation annotation results, including: merging the user preference description text and the information recommendation data to form input features; and mapping the information recommendation annotation results to the input features to construct the training sample set.

[0073] To construct the training sample set, user preference descriptions and news recommendation data are first merged to form a complete input feature that fully represents the relationship between users and news. After obtaining the input feature, it is mapped to the news recommendation annotation results. By associating the input feature with these annotations, a corresponding label can be assigned to each input feature, representing the degree of user preference for news under that feature. By combining these labeled input features, a complete training sample set is constructed. This training sample set not only contains user preference information and news features but also incorporates actual recommendation results, providing strong data support for model training. In subsequent model training, this sample set can be used to adjust model parameters to improve the model's accuracy in predicting user interests.

[0074] Optionally, a pre-trained large language model is used to perform natural language conversion processing on the acquired user profile data and historical browsing data; an information recommendation prediction model is trained based on the training sample set, including: training the newly added parameters of the pre-trained large language model based on the training sample set using a low-rank adaptation algorithm, or training all parameters of the pre-trained large language model based on the training sample set using a full-parameter fine-tuning algorithm.

[0075] When building an information recommendation prediction model, a pre-trained large language model can be used to perform natural language conversion on the acquired user profile data and historical browsing data. This transforms the structured data into text describing user behavior and preferences, enabling subsequent models to better understand and utilize this information. Specifically, the pre-trained large language model is fine-tuned based on the constructed training sample set. The low-rank adaptation algorithm (LoRA) can be used to train the pre-trained large language model with newly added parameters. First, a small number of new parameters are introduced into the pre-trained large language model, embedded in specific layers of the model as low-rank matrices to minimize changes in the model structure. The training sample set is then input into the model, and in each iteration, the newly added parameters are updated using backpropagation. This method reduces the number of training parameters, enabling effective fine-tuning of the model without significantly increasing computational burden, allowing the model to adapt to new tasks or data distributions. Alternatively, the full parameter fine-tuning algorithm can be used to train all parameters of a pre-trained large language model. The training sample set is input into the pre-trained large language model, and all parameters of the model will be adjusted through backpropagation in each iteration. Since the full parameter fine-tuning algorithm updates the entire model, it can more thoroughly adapt to new data and task requirements.

[0076] This invention utilizes a low-rank adaptation algorithm, which effectively maintains the original model's capabilities while rapidly adapting to new tasks with a small number of newly added parameters, thereby reducing training time and resource consumption. On the other hand, a full-parameter fine-tuning algorithm is used to comprehensively update all parameters of the model, ensuring that the model can fully adapt to new data distributions and task requirements. These two methods are used to specifically train a pre-trained large language model, enabling it to accurately capture the relationship between user preferences and information content. This ultimately generates an accurate information recommendation prediction model that dynamically provides personalized information recommendations based on users' latest behavior and interests, improving user experience and platform recommendation efficiency.

[0077] According to a preferred embodiment of the present invention, in the modeling process of the information recommendation prediction model, tabular data is processed first. In this scenario, the user's tabular data is divided into two parts. An example is shown below:

[0078]

[0079]

[0080] Table 1 shows the user profile data, and Table 2 shows the user's reading history data. A Large Language Model (LLM) is used for end-to-end transformation between the tabular data. The user profile data and user browsing data are directly input into the LLM, which transforms the tabular data to obtain a score for each piece of information. This score represents the recommendation level for the user. For example, if the score range is agreed to be 1-10, then the higher the score, the more likely it is to be recommended to the user. The output is the aforementioned recommendation score. This score comes from two parts: the user interest description text (as shown in the diagram) and the input. The user interest description text represents the user's profile information and historical reading history; the input represents all the information to be recommended to the user at that moment. After understanding, learning, and summarizing the user interest description text, the model predicts the score for each piece of information in the input.

[0081] By using the above steps, a large amount of training data can be quickly generated using a large language model. Since there is a lot of user information and news information, the number of training data entries generated is no less than 100,000. After generating the training data, the large language model can be trained based on the low-rank adaptation algorithm or the full-parameter fine-tuning method to obtain the news recommendation prediction model.

[0082] According to a second aspect of the present invention, an information recommendation device is provided.

[0083] Figure 4 This is a schematic diagram of the main modules of the information recommendation device according to an embodiment of the present invention, such as... Figure 4 As shown, an information recommendation device 400 includes:

[0084] Module 401 is used to acquire the user preference description text of the target user and the information data to be recommended;

[0085] The prediction module 402 is used to input the user preference description text and the information to be recommended into the information recommendation prediction model, and the information recommendation prediction model outputs the information recommendation prediction result corresponding to each piece of information to be recommended;

[0086] The recommendation module 403 is used to recommend information to the target user based on the information recommendation prediction result corresponding to each piece of information to be recommended.

[0087] Optionally, the recommended device 400 also includes a training module, which is used for:

[0088] Natural language processing is performed on the acquired user profile data and historical browsing data to generate user preference description text;

[0089] A training sample set is constructed based on user preference description text, acquired information recommendation data, and information recommendation annotation results;

[0090] The information recommendation prediction model is trained based on the training sample set.

[0091] Optionally, the training module is also used for:

[0092] Based on the acquired user profile data and historical browsing data, the characteristics of discrete data and text data are determined;

[0093] By fusing features from discrete data and text data, user preference description text is generated.

[0094] Optionally, the recommended device 400 also includes an adjustment module for adjusting the model for:

[0095] The information recommendation prediction model is deployed to the online environment to monitor online change data in real time; the online change data includes user profile change data, historical browsing change data, and information change data.

[0096] The information recommendation prediction model is adjusted and updated based on changes in online data.

[0097] Optionally, the training module is also used for:

[0098] The user preference description text and information recommendation data are combined to form input features;

[0099] The information recommendation annotation results are mapped to the input features to construct a training sample set.

[0100] Optionally, the pre-trained large language model is used to perform natural language conversion processing on the acquired user profile data and historical browsing data; the training module is also used to: train the newly added parameters of the pre-trained large language model based on the training sample set using the low-rank adaptation algorithm, or train all parameters of the pre-trained large language model based on the training sample set using the full-parameter fine-tuning algorithm.

[0101] It should be noted that the specific implementation details of the information recommendation device in the embodiments of the present invention have been described in detail in the information recommendation method above, so the details will not be repeated here.

[0102] According to a third aspect of the present invention, an electronic device is provided, comprising:

[0103] One or more processors;

[0104] Storage device for storing one or more programs.

[0105] When one or more programs are executed by one or more processors, the one or more processors implement the method provided in the first aspect of the embodiments of the present invention.

[0106] According to a fourth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method provided in the first aspect of the present invention.

[0107] According to a fifth aspect of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the method provided in the first aspect of the present invention.

[0108] Figure 5 An exemplary system architecture 500 is shown that can be applied to the information recommendation method or information recommendation device of the present invention.

[0109] like Figure 5 As shown, system architecture 500 may include terminal devices 501, 502, and 503, a network 504, and a server 505. Network 504 serves as the medium for providing communication links between terminal devices 501, 502, and 503 and server 505. Network 504 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0110] Users can use terminal devices 501, 502, and 503 to interact with server 505 via network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, and 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0111] Terminal devices 501, 502, and 503 can be various electronic devices with displays that support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0112] Server 505 can be a server that provides various services, such as a backend management server that supports shopping websites browsed by users using terminal devices 501, 502, and 503 (for example only). The backend management server can analyze and process data such as received information recommendation requests, and feed back the processing results (such as information recommendation results - for example only) to the terminal devices.

[0113] It should be noted that the information recommendation method provided in this embodiment of the invention is generally run by server 505, and correspondingly, the information recommendation device is generally set in server 505.

[0114] It should be understood that Figure 5 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0115] The following is for reference. Figure 6 It shows a schematic diagram of the structure of a computer system 600 suitable for implementing a terminal device of the present invention. Figure 6 The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0116] like Figure 6 As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 602 or programs loaded from storage section 608 into random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the system 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0117] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0118] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is run by central processing unit (CPU) 601, it performs the functions defined above in the system of this invention.

[0119] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more operable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually operate substantially in parallel, and they may sometimes operate in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0121] The modules described in the embodiments of the present invention can be implemented in software or hardware. The described modules can also be housed in a processor. For example, a processor can be described as including an acquisition module, a prediction module, and a recommendation module. The names of these modules do not necessarily limit the specific module itself. For instance, the recommendation module can also be described as "a module for recommending information to a target user based on the information recommendation prediction result corresponding to each piece of information to be recommended."

[0122] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to include: acquiring user preference description text of a target user and information data to be recommended; inputting the user preference description text and the information data to be recommended into an information recommendation prediction model, wherein the information recommendation prediction model outputs an information recommendation prediction result corresponding to each piece of information to be recommended; and recommending information to the target user based on the information recommendation prediction result corresponding to each piece of information to be recommended.

[0123] The computer program product provided in the embodiments of the present invention includes a computer program that, when executed by a processor, implements the information recommendation method in the first aspect of the present invention.

[0124] The technical solution of the present invention has the following advantages or beneficial effects: acquiring user preference description text of the target user and information data to be recommended; inputting the user preference description text and information data to be recommended into the information recommendation prediction model, and the information recommendation prediction model outputs the information recommendation prediction result corresponding to each piece of information to be recommended; recommending information to the target user based on the information recommendation prediction result corresponding to each piece of information to be recommended; and being able to make accurate predictions using the information recommendation prediction model based on the user preference description text of the target user and information data to be recommended, so that the information recommendation results are more in line with user needs and improve user experience.

[0125] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0126] It should be noted that the acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

Claims

1. An information recommendation method, characterized in that, include: Obtain user preference descriptions and recommended information data from target users; The user preference description text and the information to be recommended are input into a pre-trained information recommendation prediction model, and the information recommendation prediction model outputs the information recommendation prediction result for each piece of information to be recommended. Information recommendations are made to the target user based on the information recommendation prediction results corresponding to each piece of information to be recommended.

2. The method according to claim 1, characterized in that, The information recommendation prediction model is obtained through the following steps: Natural language processing is performed on the acquired user profile data and historical browsing data to generate user preference description text; A training sample set is constructed based on the user preference description text, the obtained information recommendation data, and the information recommendation annotation results; An information recommendation prediction model is trained based on the training sample set.

3. The method according to claim 2, characterized in that, The acquired user profile data and historical browsing data are processed using natural language conversion to generate user preference description text, including: Based on the acquired user profile data and historical browsing data, the characteristics of discrete data and text data are determined; The discrete data features and text data features are fused to generate user preference description text.

4. The method according to claim 2, characterized in that, After training the information recommendation prediction model based on the training sample set, the model further includes: The information recommendation prediction model is deployed to an online environment to monitor online change data in real time; wherein, the online change data includes user profile change data, historical browsing change data, and information change data; The information recommendation prediction model is adjusted and updated based on the online change data.

5. The method according to claim 2, characterized in that, A training sample set is constructed based on the user preference description text, the acquired information recommendation data, and the information recommendation annotation results, including: The user preference description text and information recommendation data are combined to form input features; The information recommendation annotation results are mapped to the input features to construct a training sample set.

6. The method according to claim 2, characterized in that, The pre-trained large language model is used to perform natural language conversion on the acquired user profile data and historical browsing data; The information recommendation prediction model is trained based on the training sample set, including: training the newly added parameters of the pre-trained large language model using a low-rank adaptation algorithm based on the training sample set, or training all parameters of the pre-trained large language model using a full-parameter fine-tuning algorithm based on the training sample set.

7. An information recommendation device, characterized in that, include: The acquisition module is used to acquire the user preference description text of the target user and the information to be recommended; The prediction module is used to input the user preference description text and the information to be recommended into the information recommendation prediction model, and the information recommendation prediction model outputs the information recommendation prediction result corresponding to each piece of information to be recommended; The recommendation module is used to recommend information to the target user based on the information recommendation prediction result corresponding to each piece of information to be recommended.

8. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images