Personalized user video recommendation system and method based on large model, and terminal
By extracting multimodal features and training the BERT model, combined with user feedback optimization, the problem of insufficient user interest capture in existing recommendation systems has been solved, thereby improving the accuracy and timeliness of personalized video recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-27
- Publication Date
- 2026-03-27
AI Technical Summary
Existing recommendation systems struggle to fully capture users' potential interests and respond to real-time changes in user preferences, resulting in inaccurate and untimely recommendations.
Features are extracted from explicit and implicit user data through a multimodal user feature extraction module and a video feature extraction module, respectively. A classification model is trained using the BERT model to generate a video recommendation list. The model is then optimized using user feedback data to achieve personalized recommendations.
It enables more accurate and timely personalized video recommendations, and can update and optimize user preferences in real time to improve recommendation quality.
Smart Images
Figure CN121743531A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of recommendation systems, and in particular to a large model-based personalized user video recommendation system. BACKGROUND
[0002] With the rapid development of the Internet, the amount of information users face is growing explosively, and how to effectively filter out the content of interest to users from the vast amount of information has become an important challenge. Although traditional recommendation systems have solved the problem of information overload to some extent, they still have deficiencies in dealing with complex user behavior, capturing potential interests, and responding to real-time changes in user preferences. SUMMARY
[0003] In view of the above-mentioned shortcomings of the prior art, the purpose of the present application is to provide a large model-based personalized user video recommendation system, method and terminal, which solves the problem that the existing recommendation system cannot fully capture the potential interests of users and respond to real-time changes in user preferences.
[0004] To achieve the above-mentioned purpose and other related purposes, the present application provides a large model-based personalized user video recommendation system, which includes the following contents: a multi-modal user feature extraction module for extracting multi-modal interest features of users; a video feature extraction module for extracting video features; a model training module connected to the multi-modal user feature extraction module and the video feature extraction module, for constructing a user-video interaction pair set according to the multi-modal interest features of the users and the video features, and inputting it into a BERT model for training to obtain a final classification model; a user recommendation module connected to the model training module, for predicting the potential preferences of a target user based on the final classification model, and generating a video recommendation list according to the prediction result; and a user feedback data acquisition module connected to the user recommendation module and the multi-modal user feature extraction module, for acquiring feedback data of users according to the video recommendation list and inputting it into the multi-modal feature extraction module to refine the multi-modal interest features of users, so as to optimize the classification model and generate a new video recommendation list.
[0005] In some embodiments of the first aspect of the present application, the multi-modal user feature extraction module includes: a data acquisition unit for acquiring explicit data and implicit data of different users; a data preprocessing unit for preprocessing the explicit data and implicit data of each user to obtain a multi-modal data-label pair set; and a cross-modal fusion unit for inputting the data-label pair set into a multi-modal large model for user feature extraction and multi-modal feature fusion to obtain multi-modal interest features of users.
[0006] In some embodiments of the first aspect of this application, the preprocessing method of the data preprocessing unit includes: data deduplication, data denoising, standardization processing, and data-tag pair formation operation.
[0007] In some embodiments of the first aspect of this application, the multimodal user feature extraction module is further configured to update and optimize user multimodal interest features based on the latest user interaction behavior acquired in real time.
[0008] In some embodiments of the first aspect of this application, updating and optimizing user multimodal interest features based on the latest user interaction behavior acquired in real time includes: preprocessing the latest user interaction behavior data collected in real time through the data preprocessing unit to obtain a new set of data-tag pairs; extracting features from the data-tag pair set through the cross-modal fusion unit to obtain the latest user interest features; fusing the latest user interest features with the user multimodal interest features to obtain optimized user multimodal interest features; and using the optimized user multimodal interest features to fine-tune the original classification model to update the user's video recommendation list in real time.
[0009] In some embodiments of the first aspect of this application, constructing a set of user-video interaction pairs based on the user's multimodal interest features and the video features includes: constructing interaction relationships between each user's multimodal interest features and each video feature to form multiple user-video interaction pairs.
[0010] In some embodiments of the first aspect of this application, the user feedback data collection module is an interactive recommendation interface, which allows users to perform feedback operations based on the video recommendation list to generate user feedback data.
[0011] In some embodiments of the first aspect of this application, the step of collecting user feedback data based on the video recommendation list and inputting it into a multimodal feature extraction module to refine user multimodal interest features in order to optimize the classification model and generate a new video recommendation list includes: inputting user feedback data into the multimodal feature extraction module, performing data preprocessing on it through the data preprocessing unit to obtain new data-label pairs; performing feature extraction on the data-label pairs through a cross-modal fusion unit to obtain the latest user interest features, fusing the latest user interest features with the user multimodal interest features to obtain optimized user multimodal interest features, and using the optimized user multimodal interest features to fine-tune the original classification model, thereby generating a new video recommendation list.
[0012] To achieve the above and other related objectives, a second aspect of this application provides a personalized user video recommendation method based on a large model. The method includes: extracting user multimodal interest features and video features respectively; constructing a user-video interaction pair set based on the user multimodal interest features and the video features, and inputting it into a BERT model for training to obtain a final classification model; predicting the potential preferences of target users based on the final classification model, and generating a video recommendation list based on the prediction results; collecting user feedback data based on the video recommendation list; extracting multimodal features from the collected user feedback data, refining the user multimodal interest features to optimize the classification model, and generating a new video recommendation list.
[0013] The present invention provides an electronic terminal, comprising: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory, so that the terminal executes the personalized user video recommendation method based on a large model.
[0014] As described above, the personalized user video recommendation system, method, and terminal based on a large model of this application have the following beneficial effects:
[0015] The multimodal feature extraction module and video feature extraction module extract user multimodal interest features and video features from explicit and implicit data and video data of different users, respectively, to deeply mine the complex historical behavior and preference information of users. Then, the data is input into the BERT model for training to obtain a trained classification model. Based on the trained classification model, the potential preferences of target users are predicted to generate a video recommendation list. At the same time, user feedback data is collected based on the video recommendation list and input into the multimodal feature extraction module to optimize the user multimodal interest features and generate a new video recommendation list, thereby achieving more accurate and timely personalized video recommendations. Attached Figure Description
[0016] Figure 1 The diagram shown is a framework schematic of a personalized user video recommendation system based on a large model, according to one embodiment of this application.
[0017] Figure 2 The diagram shown is a flowchart of a personalized user video recommendation method based on a large model in one embodiment of this application.
[0018] Figure 3 The diagram shown is a structural schematic of an electronic terminal according to an embodiment of this application. Detailed Implementation
[0019] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.
[0020] It should be noted that in the following description, reference is made to the accompanying drawings, which illustrate several embodiments of this application. It should be understood that other embodiments may also be used, and changes in mechanical composition, structure, electrical system, and operation may be made without departing from the spirit and scope of this application. The following detailed description should not be considered limiting, and the scope of the embodiments of this application is defined only by the claims of the published patent. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. Spatially related terms, such as “upper,” “lower,” “left,” “right,” “below,” “below,” “lower part,” “above,” “upper part,” etc., may be used herein to illustrate the relationship between one element or feature shown in the figures and another element or feature.
[0021] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," "fixing," and "holding" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.
[0022] Furthermore, as used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context indicates otherwise. The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data used can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising,” “including,” indicate the presence of the stated features, operations, elements, components, items, kinds, and / or groups, but do not exclude the presence, occurrence, or addition of one or more other features, operations, elements, components, items, kinds, and / or groups. It should be further understood that the terms “or” and “and / or” as used herein are interpreted as inclusive, or mean any one or any combination thereof. Thus, “A, B, or C” or “A, B, and / or C” means “any one of: A; B; C; A and B; A and C; B and C; A, B, and C.” An exception to this definition will only occur if the combination of elements, functions, or operations is inherently mutually exclusive in some way.
[0023] This application provides a personalized video recommendation system, method, and terminal based on a large model. It extracts multimodal interest features and video features from video data and explicit and implicit data from different users, then inputs these features into a BERT model for training, obtaining a trained classification model. Based on this model, it predicts the potential preferences of target users, thereby generating a video recommendation list. Simultaneously, it collects user feedback data from this recommendation list and inputs it into a multimodal feature extraction module to optimize the user's multimodal interest features, generating a new video recommendation list accordingly. This achieves more accurate and timely personalized video recommendations.
[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the invention.
[0025] like Figure 1 The diagram shown illustrates a framework of a large-model-based personalized user video recommendation system according to an embodiment of the present invention, comprising:
[0026] The multimodal user feature extraction module 100 is used to extract multimodal user interest features;
[0027] In one embodiment, the multimodal user feature extraction module includes:
[0028] The data acquisition unit is used to collect explicit and implicit data from different users. The explicit data includes age, gender, occupation, and personal interests. The implicit data includes historical behavioral data, text data, and image data. Historical behavioral data includes browsing history, click behavior, and purchase history. Text data includes posted text and comments. Image data includes uploaded images and video covers.
[0029] The data preprocessing unit is used to preprocess the explicit and implicit data of each user to obtain a multimodal data-label pair set;
[0030] In one embodiment, the preprocessing method of the data preprocessing unit includes: data deduplication, data denoising, standardization, and data-label pair formation. Specifically, data deduplication reduces data redundancy by removing duplicate data information from different modalities; data denoising reduces noise in explicit and implicit data, thereby selecting clean and information-rich multimodal data; standardization ensures label consistency and comparability by standardizing labels, including label normalization, synonym merging, and removal of invalid labels, to eliminate differences between different data and improve data comparability; and the data-label pair formation operation, after the data deduplication, denoising, and standardization processes, associates each data point with its corresponding label to form a data-label pair set, providing a cleaner and more information-rich dataset for model training.
[0031] The cross-modal fusion unit is used to input the data-label pair set into a multimodal large model for user feature extraction and multimodal feature fusion to obtain user multimodal interest features. The user multimodal interest features obtained by the cross-modal fusion unit have a vector dimension of 256. The multimodal large model refers to an existing open-source multimodal large model, such as Ovis, which can process and understand various data inputs such as text and images to obtain user interest features of different modalities; then, the user interest features of different modalities are fused through orthogonal sequences to finally obtain the user multimodal interest features.
[0032] The video feature extraction module 200 is used to extract video features. Specifically, it extracts video features by inputting video data into an existing open-source large model, thereby obtaining a 256-dimensional video feature vector. For example, open-source large models for extracting video features include frameworks such as ViViT (Video Vision Transformer) and video_features.
[0033] The model training module 300 is connected to the multimodal user feature extraction module 100 and the video feature extraction module 200. It is used to construct a set of user-video interaction pairs based on the user multimodal interest features and the video features, and input them into the BERT model for training to obtain the final classification model.
[0034] In one embodiment, constructing a user-video interaction pair set involves establishing interaction relationships between each user's multimodal interest features and various video features to form multiple user-video interaction pairs. The process is as follows: Multimodal feature extraction and video feature extraction modules extract user multimodal interest features and video features respectively, constructing user-video interaction pairs. The format of each user-video interaction pair is (user multimodal interest feature, video feature i, interaction probability); where video feature i is a feature extracted from any video in the video dataset, and the interaction probability is either 1 or 0, where 1 represents user interest in video i, and 0 represents no interest. This process is repeated to obtain each user's interest in each video, thus resulting in a user-video interaction pair set that includes the interest of different users in different videos.
[0035] In one embodiment, the user preference prediction function uses the sigmoid function. The output range of the sigmoid function is (0,1). By setting a threshold, the output of the sigmoid function is converted into a binary classification result, so that the user preference prediction result is 1 or 0. For example, if the threshold is set to 0.5, when the output result of the sigmoid function is greater than 0.5, it means that the target user may be interested in the video, and vice versa.
[0036] The user recommendation module 400 is connected to the model training module 300 and is used to predict the potential preferences of the target user based on the final classification model and generate a video recommendation list based on the prediction results.
[0037] In one embodiment, the recommendation process is as follows: For the target user, the multimodal user interest features of the target user are first extracted by the multimodal user feature extraction module, and then input into the trained classification model along with the video features. Through fine-tuning the model, all videos with a prediction result of 1 are used as a recommendation list to recommend to the user.
[0038] In one embodiment, the multimodal user feature extraction module is further configured to update and optimize user multimodal interest features based on the latest user interaction behavior acquired in real time, thereby improving recommendation quality. The specific process is as follows:
[0039] The data preprocessing unit performs data preprocessing on the latest user interaction data collected in real time to obtain a new set of data-tag pairs.
[0040] The data-label set is subjected to feature extraction by a cross-modal fusion unit to obtain the latest user interest features. The latest user interest features are then fused with the user multimodal interest features to obtain more accurate user multimodal interest features. The optimized user multimodal interest features are then used to fine-tune the original classification model to update the user's video recommendation list in real time, thereby obtaining more real-time recommendation results.
[0041] The user feedback data acquisition module 500 is connected to the user recommendation module 400 and the multimodal user feature extraction module 100. It is used to collect user feedback data according to the video recommendation list and input it into the multimodal feature extraction module to refine the user's multimodal interest features, thereby optimizing the classification model.
[0042] In one embodiment, the user feedback data collection module 500 is an interactive recommendation interface, which allows users to perform feedback operations based on the video recommendation list to generate user feedback data; wherein, the user feedback data includes explicit user feedback data; the explicit feedback data is the user's actions of liking or disliking, forwarding, sharing, and commenting on the videos in the video recommendation list.
[0043] In one embodiment, the process of collecting user feedback data based on the video recommendation list and inputting it into the multimodal feature extraction module to refine the user's multimodal interest features, thereby optimizing the classification model and generating a new video recommendation list, is as follows:
[0044] User feedback data is input into the multimodal feature extraction module, and the data preprocessing unit preprocesses the data to obtain new data-label pairs.
[0045] The data-label pairs are used to extract features through a cross-modal fusion unit to obtain the latest user interest features. The latest user interest features are then fused with the user multimodal interest features to obtain optimized user multimodal interest features. The optimized user multimodal interest features are then used to fine-tune the original classification model to generate a new video recommendation list.
[0046] Similar to the above embodiments, the present invention also provides a method for personalized user video recommendation based on a large model, the method flow of which is as follows: Figure 2 As shown, the method includes the following steps:
[0047] Step S201: Extract user multimodal interest features and video features respectively;
[0048] Step S202: Construct a set of user-video interaction pairs based on the user's multimodal interest features and the video features, and input them into the BERT model for training to obtain the final classification model;
[0049] Step S203: Based on the final classification model, predict the potential preferences of the target users and generate a video recommendation list based on the prediction results;
[0050] Step S204: Collect user feedback data based on the video recommendation list;
[0051] Step S205: Extract multimodal features from the collected user feedback data, refine the user's multimodal interest features, optimize the classification model, and generate a new video recommendation list.
[0052] The personalized user video recommendation method based on a large model provided in this invention can be implemented on the terminal side or the server side. Regarding the hardware structure of the electronic terminal, please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of an optional hardware structure of a personalized user video recommendation terminal 1000 based on a large model provided in an embodiment of the present invention. The terminal 1000 can be a mobile phone, computer device, tablet device, personal digital processing device, factory back-end processing device, etc. The terminal 1000 includes: at least one processor 1001, a memory 1002, at least one network interface 10010, and a user interface 1009. The various components in the device are coupled together through a bus system 1005. It is understood that the bus system 1005 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 1005 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 3 The general will label all buses as bus systems.
[0053] The user interface 1009 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.
[0054] It is understood that memory 1002 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable categories of memory.
[0055] In this embodiment of the invention, the memory 1002 is used to store various categories of data to support the operation of the terminal 1000. Examples of this data include: any executable program for operation on the terminal 1000, such as the operating system 10021 and application program 10022; the operating system 10021 contains various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. The application program 10022 may contain various applications, such as media players, browsers, etc., for implementing various application services. The implementation of the large-model-based personalized user video recommendation method provided in this embodiment of the invention can be included in the application program 10022.
[0056] The methods disclosed in the above embodiments of the present invention can be applied to or implemented by the processor 1001. The processor 1001 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 1001 or by instructions in the form of software. The processor 1001 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 1001 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor 1001 may be a microprocessor or any conventional processor, etc. The steps of the accessory optimization method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in a memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.
[0057] In an exemplary embodiment, the terminal 1000 may be used to execute the aforementioned method by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs).
[0058] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented using computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0059] In the embodiments provided in this application, the computer-readable and writable storage medium may include read-only memory, random access memory, EEPROM, CD-ROM or other optical disc storage devices, disk storage devices or other magnetic storage devices, flash memory, USB flash drive, portable hard drive, or any other medium capable of storing desired program code in the form of instructions or data structures and accessible by a computer. Additionally, any connection may be appropriately referred to as a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of the medium. However, it should be understood that computer-readable and writable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but are intended for non-transient, tangible storage media. The disks and optical discs used in the application include compact discs (CDs), laser discs, optical discs, digital multifunction discs (DVDs), floppy disks, and Blu-ray discs, where disks typically copy data magnetically, while optical discs use lasers to copy data optically.
[0060] It should be noted that the large-model-based personalized user video recommendation method in this embodiment can implement all the functions of the large-model-based personalized user video recommendation system in the above embodiments, so it will not be explained again here.
[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0062] In summary, this application provides a personalized user video recommendation system, method, and terminal based on a large model. It extracts user multimodal interest features and video features from explicit and implicit data and video data of different users through a multimodal feature extraction module and a video feature extraction module, respectively, to deeply mine complex historical behavior and preference information of users. These features are then input into a BERT model for training, resulting in a trained classification model. Based on this model, the potential preferences of target users are predicted, thereby generating a video recommendation list. Simultaneously, user feedback data is collected based on the video recommendation list and input into the multimodal feature extraction module to optimize user multimodal interest features and generate a new video recommendation list, achieving more accurate and timely personalized video recommendations.
[0063] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.
Claims
1. A personalized user video recommendation system based on a large model, characterized in that, include The multimodal user feature extraction module is used to extract multimodal user interest features; The video feature extraction module is used to extract video features; The model training module connects the multimodal user feature extraction module and the video feature extraction module. It is used to construct a set of user-video interaction pairs based on the user's multimodal interest features and the video features, and input them into the BERT model for training to obtain the final classification model. The user recommendation module, connected to the model training module, is used to predict the potential preferences of target users based on the final classification model and generate a video recommendation list based on the prediction results. The user feedback data collection module, connected to the user recommendation module and the multimodal user feature extraction module, is used to collect user feedback data based on the video recommendation list and input it into the multimodal feature extraction module to refine the user's multimodal interest features, so as to optimize the classification model and generate a new video recommendation list.
2. The personalized user video recommendation system based on a large model according to claim 1, characterized in that, The multimodal user feature extraction module includes: The data acquisition unit is used to collect explicit and implicit data from different users; The data preprocessing unit is used to preprocess the explicit and implicit data of each user to obtain a multimodal data-label pair set; The cross-modal fusion unit is used to input the data-label pair set into the multimodal large model to extract user features and fuse multimodal features to obtain user multimodal interest features.
3. The personalized user video recommendation system based on a large model according to claim 2, characterized in that, The preprocessing methods of the data preprocessing unit include: data deduplication, data denoising, standardization, and data-label pair formation.
4. The personalized user video recommendation system based on a large model according to claim 1, characterized in that, The multimodal user feature extraction module is also used to update and optimize user multimodal interest features based on the latest user interaction behavior acquired in real time.
5. The personalized user video recommendation system based on a large model according to claim 4, characterized in that, The method of updating and optimizing user multimodal interest features based on the latest user interaction behavior acquired in real time includes: The data preprocessing unit performs data preprocessing on the latest user interaction data collected in real time to obtain a new set of data-tag pairs. The data-tag pair set is subjected to feature extraction by a cross-modal fusion unit to obtain the latest user interest features. The latest user interest features are then fused with the user multimodal interest features to obtain optimized user multimodal interest features. The optimized user multimodal interest features are then used to fine-tune the original classification model to update the user's video recommendation list in real time.
6. The personalized user video recommendation system based on a large model according to claim 1, characterized in that, The step of constructing a set of user-video interaction pairs based on the user's multimodal interest features and the video features includes: Each user's multimodal interest features are used to construct interaction relationships with each video feature to form multiple user-video interaction pairs.
7. The personalized user video recommendation system based on a large model according to claim 1, characterized in that, The user feedback data collection module is an interactive recommendation interface that allows users to perform feedback operations based on the video recommendation list to generate user feedback data.
8. The personalized user video recommendation system based on a large model according to claim 1, characterized in that, The step of collecting user feedback data based on the video recommendation list and inputting it into the multimodal feature extraction module to refine the user's multimodal interest features, so as to optimize the classification model and generate a new video recommendation list, includes: User feedback data is input into the multimodal feature extraction module, and the data preprocessing unit preprocesses the data to obtain new data-label pairs. The data-label pairs are used to extract features through a cross-modal fusion unit to obtain the latest user interest features. The latest user interest features are then fused with the user multimodal interest features to obtain optimized user multimodal interest features. The optimized user multimodal interest features are then used to fine-tune the original classification model to generate a new video recommendation list.
9. A personalized user video recommendation method based on a large model, characterized in that, The method includes: Extract user multimodal interest features and video features separately; Based on the user's multimodal interest features and the video features, a set of user-video interaction pairs is constructed and input into the BERT model for training to obtain the final classification model. Based on the final classification model, the potential preferences of the target users are predicted, and a video recommendation list is generated based on the prediction results; Collect user feedback data based on the video recommendation list; Multimodal feature extraction is performed on the collected user feedback data to refine the user's multimodal interest features, thereby optimizing the classification model and generating a new video recommendation list.
10. An electronic terminal, characterized in that, include: Processor and memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory, so that the terminal performs the large-model-based personalized user video recommendation method as described in claim 9.