Recommendation methods, apparatus, devices, and media based on retrieval-enhanced generative models
By using a recommendation method based on a retrieval-enhanced generative model, and leveraging IPTV users' viewing behavior data and knowledge graphs, a fused feature vector is generated to accurately recommend IPTV video content. This solves the problems of long recommendation times and low efficiency in existing technologies, and achieves efficient and stable video content recommendation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YOUDI NETWORK CO LTD
- Filing Date
- 2026-01-20
- Publication Date
- 2026-06-02
AI Technical Summary
Existing IPTV video content recommendation methods rely on a single user's historical records or simple collaborative filtering, resulting in long recommendation times and low efficiency, and failing to accurately determine the user's viewing needs.
A recommendation method based on retrieval-enhanced generative models is adopted. By acquiring viewing behavior data of IPTV users, encoding and concatenating feature vectors, similarity retrieval is performed using knowledge graphs to extract viewing preference information, generating fused feature vectors, and using a large language model to filter video content that meets the requirements.
It improves the efficiency and accuracy of IPTV video content recommendation, reduces computing power consumption and bandwidth usage, ensures that recommended content closely matches user needs, and reduces interference from invalid recommendations.
Smart Images
Figure CN122132594A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of big data technology and artificial intelligence technology, and in particular to recommendation methods, apparatus, devices and media based on retrieval-enhanced generative models. Background Technology
[0002] In the field of IPTV video content recommendation, current mainstream recommendation methods have significant limitations. Most methods rely solely on a single user's historical viewing history or employ relatively simple collaborative filtering algorithms to perform IPTV video content recommendation.
[0003] However, since traditional recommendation methods cannot determine users' viewing needs, in this case, the recommendation system can only perform indiscriminate filtering based on the full amount of video content. Therefore, the recommendation process needs to traverse a large number of invalid candidate contents that are irrelevant to user preferences, which leads to excessively long recommendation time for IPTV video content and is not conducive to improving the recommendation efficiency of IPTV video content. Therefore, how to recommend IPTV video content to IPTV users is a technical problem that urgently needs to be solved. Summary of the Invention
[0004] This application provides a recommendation method, apparatus, device, and medium based on a retrieval-enhanced generative model to address the technical problem of how to recommend IPTV video content to IPTV users.
[0005] In a first aspect, embodiments of this application provide a recommendation method based on a retrieval-enhanced generative model, applied to electronic devices, the recommendation method comprising: Acquire IPTV users' viewing behavior data, which includes program identifier, program name, program type, viewing time, and viewing duration. The program identifier, program content, program name, and program type are encoded separately to obtain vectors for program identifier, program name, and program type, respectively. The viewing time and viewing duration are encoded separately to obtain vectors for viewing time and viewing duration, respectively. By concatenating the vectors of program identifier, program name, program type, viewing time, and viewing duration, we obtain the viewing behavior feature vector of IPTV users. The viewing behavior feature vector of IPTV users is input into the retrieval enhancement generative model. Through the retrieval enhancement generative model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The feature vectors of IPTV users' viewing behavior and viewing preference information are fused to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0006] In one possible implementation of the first aspect, the acquisition of IPTV user viewing behavior data includes program identifier, program name, program type, viewing time, and viewing duration, comprising: Connect to the business server and send a data retrieval request to the business server. The data retrieval request carries the IPTV user's account. The service server receives the viewing behavior data of IPTV users returned based on the data acquisition request.
[0007] In one possible implementation of the first aspect, the acquisition of IPTV user viewing behavior data includes program identifier, program name, program type, viewing time, and viewing duration, comprising: Using a pre-defined word embedding model, the program identifier, program content, program name, and program type are encoded respectively, resulting in vectors for the program identifier, program name, and program type. The viewing time and viewing duration are encoded using a time-series model, resulting in vectors for viewing time and viewing duration, respectively.
[0008] In one possible implementation of the first aspect, the step of inputting the viewing behavior feature vector of IPTV users into a retrieval-enhanced generative model, performing a similarity retrieval on the viewing behavior feature vector of IPTV users in a knowledge graph through the retrieval-enhanced generative model, obtaining the similarity between each entity in the knowledge graph and the IPTV user, acquiring the viewing preference information associated with the entity with the highest similarity, and extracting features from the viewing preference information to obtain a feature vector of the viewing preference information, including: Call the input interface of the retrieval enhancement generative model, and input the viewing behavior feature vector of IPTV users into the retrieval enhancement generative model through the input interface; By using a retrieval-enhanced generative model, similarity retrieval is performed on the viewing behavior feature vectors of IPTV users in the knowledge graph. This yields the similarity between each entity in the knowledge graph and the IPTV user. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information.
[0009] In one possible implementation of the first aspect, the step of fusing the IPTV user's viewing behavior feature vector and the viewing preference information feature vector to generate a fused feature vector, processing the fused feature vector through a large language model to generate the IPTV user's viewing needs, filtering IPTV video content that meets the viewing needs from the IPTV video content library, and recommending IPTV video content that meets the viewing needs to the IPTV user includes: The IPTV user's viewing behavior feature vector and viewing preference information feature vector are input into a preset encoder. The encoder's attention mechanism is used to fuse the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector. By processing and fusing feature vectors through a large language model, the viewing needs of IPTV users are generated. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0010] In one possible implementation of the first aspect, the step of processing and fusing feature vectors through a large language model to generate viewing requirements for IPTV users, filtering IPTV video content that matches the viewing requirements from the IPTV video content library, and recommending IPTV video content that matches the viewing requirements to IPTV users includes: Obtain the loading instruction, and load the large language model from the model file using the loading instruction, and input the fused feature vector into the trained large language model; By processing and fusing feature vectors through a large language model, the viewing needs of IPTV users are generated. These viewing needs are then matched with video metadata in the IPTV video content library to obtain matching results. From these results, IPTV video content that meets the viewing needs is selected and recommended to IPTV users.
[0011] In one possible implementation of the first aspect, after fusing the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector, processing the fused feature vector through a large language model to generate the IPTV user's viewing needs, filtering IPTV video content that meets the viewing needs from the IPTV video content library, and recommending IPTV video content that meets the viewing needs to the IPTV user, the recommendation method includes: Obtain recommended IPTV video content that matches your viewing needs, and store the recommended results in a preset storage area.
[0012] Secondly, embodiments of this application provide a recommendation apparatus based on a retrieval-enhanced generative model, applied to electronic devices, comprising: The acquisition module is used to acquire viewing behavior data of IPTV users. The viewing behavior data includes program identifier, program name, program type, viewing time and viewing duration. The encoding module is used to encode the program identifier, program content, program name, and program type respectively, and to obtain vectors for the program identifier, program name, and program type respectively. It also encodes the viewing time and viewing duration respectively, and to obtain vectors for the viewing time and viewing duration respectively. The splicing module is used to splice the vectors of program identifier, program name, program type, viewing time, and viewing duration to obtain the viewing behavior feature vector of IPTV users. The extraction module is used to input the viewing behavior feature vector of IPTV users into the retrieval enhancement generation model. Through the retrieval enhancement generation model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The recommendation module is used to fuse the feature vectors of IPTV users' viewing behavior and viewing preference information to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0013] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the recommended method described in the first aspect above.
[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the recommended method described in the first aspect above.
[0015] Fifthly, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to perform the recommended method described in the first aspect.
[0016] The beneficial effects of the embodiments of this application are as follows: Firstly, the viewing behavior feature vector of IPTV users is input into the retrieval enhancement generation model. Through the retrieval enhancement generation model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and the IPTV user. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The viewing behavior feature vector of IPTV users and the feature vector of the viewing preference information are fused to generate a fused feature vector. The fused feature vector is processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is selected from the IPTV video content library and recommended to IPTV users. Secondly, since IPTV video content that meets viewing needs is selected from the IPTV video content library and recommended to IPTV users, it is not affected by human intervention, which helps to improve the stability of recommending IPTV video content that meets viewing needs to IPTV users. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 Application scenario diagram of the recommended method provided in the embodiments of this application; Figure 2 This is a flowchart illustrating the recommended method provided in the embodiments of this application; Figure 3 A flowchart illustrating the implementation of S204 provided in this application embodiment; Figure 4 A schematic block diagram of the recommended apparatus provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.
[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0021] It should be understood that in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. The terms "comprising," "including," "having," and their variations all mean "including but not limited to," unless otherwise specifically emphasized.
[0022] Furthermore, the technical solutions of the various embodiments can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed in this application.
[0023] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.
[0024] The recommended methods provided in this application can be applied to electronic devices, including but not limited to cloud servers, central servers, mobile phones, tablets, wearable devices, vehicle-mounted devices, and laptops. This application does not impose any restrictions on the specific type of electronic device.
[0025] Please see Figure 1 , Figure 1 The application scenario diagram of the recommended method provided in the embodiments of this application is described in detail below: Electronic devices connect to IPTV terminals to obtain viewing behavior data of IPTV users. The viewing behavior data includes program identifier, program name, program type, viewing time and viewing duration. The electronic device encodes the program identifier, program content, program name, and program type respectively, resulting in vectors for the program identifier, program name, and program type. It also encodes the viewing time and viewing duration respectively, resulting in vectors for the viewing time and viewing duration. The electronic device concatenates the vectors of program identifier, program name, program type, viewing time, and viewing duration to obtain the viewing behavior feature vector of the IPTV user. Electronic devices input the viewing behavior feature vector of IPTV users into the retrieval enhancement generative model. Through the retrieval enhancement generative model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. Electronic devices fuse the feature vectors of IPTV users' viewing behavior and viewing preference information to generate a fused feature vector. This fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. Finally, IPTV video content that meets these viewing needs is selected from the IPTV video content library and recommended to IPTV users.
[0026] Among them, the IPTV terminal is the terminal used by IPTV users.
[0027] In this embodiment of the application, the electronic device selects IPTV video content that meets the viewing requirements from the IPTV video content library without having to perform a full traversal, thus reducing the computing power consumption and bandwidth usage of the electronic device.
[0028] Please see Figure 2 , Figure 2 This is a flowchart illustrating the recommended method provided in the embodiments of this application, which can be applied to electronic devices.
[0029] like Figure 2 As shown, the recommended method provided in this application includes the following steps, detailed below: S201, Obtain IPTV user viewing behavior data, including program identifier, program name, program type, viewing time and viewing duration; The acquisition of IPTV user viewing behavior data includes program identifier, program name, program type, viewing time, and viewing duration, including: Connect to the business server and send a data retrieval request to the business server. The data retrieval request carries the IPTV user's account. The service server receives the viewing behavior data of IPTV users returned based on the data acquisition request.
[0030] S202, encode the program identifier, program content, program name, and program type respectively to obtain vectors of program identifier, program name, and program type respectively; encode the viewing time and viewing duration respectively to obtain vectors of viewing time and viewing duration respectively. The acquisition of IPTV user viewing behavior data includes program identifier, program name, program type, viewing time, and viewing duration, including: Using a pre-defined word embedding model, the program identifier, program content, program name, and program type are encoded respectively, resulting in vectors for the program identifier, program name, and program type. The viewing time and viewing duration are encoded using a time-series model, resulting in vectors for viewing time and viewing duration, respectively.
[0031] S203, concatenate the vectors of program identifier, program name, program type, viewing time, and viewing duration to obtain the viewing behavior feature vector of IPTV users; S204, input the IPTV user's viewing behavior feature vector into the retrieval enhancement generation model. Through the retrieval enhancement generation model, perform similarity retrieval on the IPTV user's viewing behavior feature vector in the knowledge graph to obtain the similarity between each entity in the knowledge graph and the IPTV user. Obtain the viewing preference information associated with the entity with the highest similarity. Extract features from the viewing preference information to obtain the feature vector of the viewing preference information. Among them, obtaining viewing preference information associated with entities with the highest similarity is the preference logic that best matches the behavioral characteristics and scenario needs of target users. Filtering content based on this information can effectively avoid the problems of blind matching and content mismatch in traditional recommendation.
[0032] S205: The IPTV user's viewing behavior feature vector and viewing preference information feature vector are fused to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the IPTV user's viewing needs. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to the IPTV user.
[0033] This involves selecting IPTV video content from the IPTV video content library that meets viewing needs and recommending such content to IPTV users, ensuring that the pushed video content closely matches the viewing needs of IPTV users and significantly reducing the proportion of invalid recommendations.
[0034] For ease of explanation, the following example is provided: For example, in the family evening leisure scenario, the system first obtains the user's historical viewing behavior data, including the user's long-term program on-demand records, viewing completion rate, collection marks, and other information; at the same time, it retrieves the entity association information stored in the family knowledge graph to clarify the strong association between the user and family ethics dramas and light comedy content, as well as the user's fixed viewing time habit from 8 pm to 10 pm.
[0035] Based on the above information, the system directly filters video resources updated between 8 PM and 10 PM from the IPTV video content library, and those tagged as family dramas or light comedies, to create a targeted recommendation list. After opening IPTV, users no longer need to manually search through a vast library of tens of thousands of content items, including movies, TV series, variety shows, and documentaries, or repeatedly switch channels; they can directly view the recommendation list. Simultaneously, the system automatically filters out content completely unrelated to user preferences, such as war films, horror films, and suspense thrillers, effectively avoiding interference from such invalid recommendations. This significantly reduces users' content filtering time and operational costs, allowing them to quickly get into a viewing state and improving their experience during evening leisure time.
[0036] The process of fusing the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector, processing the fused feature vector using a large language model to generate the IPTV user's viewing needs, filtering IPTV video content that meets the viewing needs from the IPTV video content library, and recommending IPTV video content that meets the viewing needs to the IPTV user includes: The IPTV user's viewing behavior feature vector and viewing preference information feature vector are input into a preset encoder. The encoder's attention mechanism is used to fuse the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector. By processing and fusing feature vectors through a large language model, the viewing needs of IPTV users are generated. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0037] For ease of explanation, the following example is provided: For example, if the IPTV user is Zhang San, retrieve Zhang San's historical viewing behavior data: In the past month, Zhang San spent 85% of his time watching family dramas and urban light comedies between 8 pm and 10 pm, and the completion rate of dramas tagged with "mother-in-law and daughter-in-law relationship" and "neighborhood mutual assistance" exceeded 90%; at the same time, the system obtains entity association information from the knowledge graph, which clarifies that Zhang San's viewing scenario is "single evening leisure", and his preferences are most similar to those of user Li Si on the platform. Li Si's historical preferences include strong association tags of "light comedy and life-like plot".
[0038] The system uses cosine similarity calculation to identify the entity-related viewing preference information of Li Si, which is associated with "light comedy and slice-of-life dramas and 8 PM to 10 PM". This information has a 92% similarity to Zhang San's preference features, making it the highest-matching preference data at present. Subsequently, this preference information is integrated with Zhang San's historical viewing data to generate standardized demand tags: 8 PM to 10 PM and family ethics light comedy and slice-of-life dramas.
[0039] Based on the generated demand tags, the system filters out three newly launched dramas with corresponding tags from the IPTV video content library, excluding content that does not meet the demand, such as war dramas and suspense dramas; at 7:50 pm, the filtered drama recommendation list is pushed to the home page of Zhang San's IPTV terminal.
[0040] After Zhang San turns on IPTV, he doesn't need to search through tens of thousands of contents; he can directly see recommended dramas that match his needs and select content to start watching within 5 minutes, completely avoiding the interference of invalid recommendations. At the same time, since electronic devices do not need to traverse the entire content library, the recommendation response time is shortened by 60%, significantly reducing computing power and bandwidth consumption.
[0041] The process of processing and fusing feature vectors through a large language model to generate viewing requirements for IPTV users, filtering IPTV video content that matches these requirements from the IPTV video content library, and recommending such content to IPTV users includes: Obtain the loading instruction, and load the large language model from the model file using the loading instruction, and input the fused feature vector into the trained large language model; By processing and fusing feature vectors through a large language model, the viewing needs of IPTV users are generated. These viewing needs are then matched with video metadata in the IPTV video content library to obtain matching results. From these results, IPTV video content that meets the viewing needs is selected and recommended to IPTV users.
[0042] The recommendation method, after fusing the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector, processing the fused feature vector through a large language model to generate the IPTV user's viewing needs, filtering IPTV video content that meets the viewing needs from the IPTV video content library, and recommending IPTV video content that meets the viewing needs to the IPTV user, includes: Obtain recommended IPTV video content that matches your viewing needs, and store the recommended results in a preset storage area.
[0043] The beneficial effects of the embodiments of this application are as follows: Firstly, the viewing behavior feature vector of IPTV users is input into the retrieval enhancement generation model. Through the retrieval enhancement generation model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and the IPTV user. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The viewing behavior feature vector of IPTV users and the feature vector of the viewing preference information are fused to generate a fused feature vector. The fused feature vector is processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is selected from the IPTV video content library and recommended to IPTV users. Secondly, since IPTV video content that meets viewing needs is selected from the IPTV video content library and recommended to IPTV users, it is not affected by human intervention, which helps to improve the stability of recommending IPTV video content that meets viewing needs to IPTV users.
[0044] Please see Figure 3 , Figure 3 The implementation flowchart of S204 provided in the embodiments of this application is described in detail below: S301, Call the input interface of the retrieval enhancement generation model, and input the viewing behavior feature vector of the IPTV user into the retrieval enhancement generation model through the input interface; S302, by using the retrieval enhancement generative model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and the IPTV user, the viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information.
[0045] In this embodiment, the viewing preference information associated with the entities with the highest similarity represents the preference logic that best matches the behavioral characteristics of the target user. By extracting feature vectors from the viewing preference information associated with the entities with the highest similarity, this implicit preference association can be transformed into structured and computable vector data, avoiding the subjective bias of manual screening and enabling the recommendation system to accurately capture the user's potential viewing needs.
[0046] For the recommended method described in the above embodiments, please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic block diagram of the recommended apparatus provided in the embodiments of this application. Figure 4 The recommended device 400 shown can be applied to, for example... Figure 1 The application scenario diagram shows electronic devices. The following section uses electronic devices as an example to illustrate this. Figure 4 The recommended device 400 shown will be described in detail. The recommended device 400 may include an acquisition module 401, an encoding module 402, a splicing module 403, an extraction module 404, and a recommendation module 405.
[0047] The acquisition module 401 is used to acquire viewing behavior data of IPTV users. The viewing behavior data includes program identifier, program name, program type, viewing time and viewing duration. The encoding module 402 is used to encode the program identifier, program content, program name, and program type respectively, to obtain vectors of program identifier, program name, and program type respectively, and to encode the viewing time and viewing duration respectively, to obtain vectors of viewing time and viewing duration respectively. The splicing module 403 is used to splice the vectors of program identifier, program name, program type, viewing time, and viewing duration to obtain the viewing behavior feature vector of IPTV users. The extraction module 404 is used to input the viewing behavior feature vector of IPTV users into the retrieval enhancement generation model. Through the retrieval enhancement generation model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The recommendation module 405 is used to fuse the feature vectors of IPTV users' viewing behavior and viewing preference information to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0048] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.
[0049] The beneficial effects of the embodiments of this application are as follows: Firstly, the viewing behavior feature vector of IPTV users is input into the retrieval enhancement generation model. Through the retrieval enhancement generation model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and the IPTV user. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The viewing behavior feature vector of IPTV users and the feature vector of the viewing preference information are fused to generate a fused feature vector. The fused feature vector is processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is selected from the IPTV video content library and recommended to IPTV users. Secondly, since IPTV video content that meets viewing needs is selected from the IPTV video content library and recommended to IPTV users, it is not affected by human intervention, which helps to improve the stability of recommending IPTV video content that meets viewing needs to IPTV users.
[0050] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.
[0051] like Figure 5 As shown, Figure 5 The electronic device 2 includes: at least one processor 20, a memory 21, and a computer program 22 stored in the memory 21 and executable on the at least one processor 20, wherein the processor 20 executes the computer program 22 to implement the steps in any of the above method embodiments.
[0052] The electronic device 2 may include, but is not limited to, a processor 20 and a memory 21. Those skilled in the art will understand that... Figure 5 This is merely an example of electronic device 2 and does not constitute a limitation on electronic device 2. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, it may also include input / output devices, network access devices, etc.
[0053] The processor 20 is used to run a computer program 22 stored in the memory 21, and performs the following steps when executing the computer program 22: Acquire IPTV users' viewing behavior data, which includes program identifier, program name, program type, viewing time, and viewing duration. The program identifier, program content, program name, and program type are encoded separately to obtain vectors for program identifier, program name, and program type, respectively. The viewing time and viewing duration are encoded separately to obtain vectors for viewing time and viewing duration, respectively. By concatenating the vectors of program identifier, program name, program type, viewing time, and viewing duration, we obtain the viewing behavior feature vector of IPTV users. The viewing behavior feature vector of IPTV users is input into the retrieval enhancement generative model. Through the retrieval enhancement generative model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The feature vectors of IPTV users' viewing behavior and viewing preference information are fused to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0054] The processor 20 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors, field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.
[0055] In some embodiments, the memory 21 may be an internal storage unit of the electronic device 2, such as a hard disk or memory of the electronic device 2. In other embodiments, the memory 21 may be an external storage device of the electronic device 2, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the electronic device 2.
[0056] Furthermore, the memory 21 may include both internal storage units and external storage devices of the electronic device 2. The memory 21 is used to store the operating system, applications, boot loader, data, and other programs, such as the program code of the computer program. The memory 21 can also be used to temporarily store data that has been output or will be output.
[0057] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0058] This application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.
[0059] The computer-readable storage medium may also be an external storage device of the recommended device or electronic device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, or non-transitory computer-readable storage medium equipped on the recommended device or electronic device.
[0060] Since the computer program stored in the computer-readable storage medium can execute any of the recommendation methods based on the retrieval enhancement generation model provided in the embodiments of this application, the computer-readable storage medium can achieve the beneficial effects that any of the recommendation methods based on the retrieval enhancement generation model provided in the embodiments of this application can achieve, as detailed in the preceding embodiments, and will not be repeated here.
[0061] This application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the recommended method described above.
[0062] When a computer program is loaded into an electronic device, it can perform the following steps: Acquire IPTV users' viewing behavior data, which includes program identifier, program name, program type, viewing time, and viewing duration. The program identifier, program content, program name, and program type are encoded separately to obtain vectors for program identifier, program name, and program type, respectively. The viewing time and viewing duration are encoded separately to obtain vectors for viewing time and viewing duration, respectively. By concatenating the vectors of program identifier, program name, program type, viewing time, and viewing duration, we obtain the viewing behavior feature vector of IPTV users. The viewing behavior feature vector of IPTV users is input into the retrieval enhancement generative model. Through the retrieval enhancement generative model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The feature vectors of IPTV users' viewing behavior and viewing preference information are fused to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
[0063] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
[0064] Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium includes: an entity or device for carrying computer program code to an electronic device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium.
[0065] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0066] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.
Claims
1. A recommendation method based on a retrieval-enhanced generative model, characterized in that, When applied to electronic devices, the recommended method includes: Acquire IPTV users' viewing behavior data, which includes program identifier, program name, program type, viewing time, and viewing duration. The program identifier, program content, program name, and program type are encoded separately to obtain vectors for program identifier, program name, and program type, respectively. The viewing time and viewing duration are encoded separately to obtain vectors for viewing time and viewing duration, respectively. By concatenating the vectors of program identifier, program name, program type, viewing time, and viewing duration, we obtain the viewing behavior feature vector of IPTV users. The viewing behavior feature vector of IPTV users is input into the retrieval enhancement generative model. Through the retrieval enhancement generative model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The feature vectors of IPTV users' viewing behavior and viewing preference information are fused to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
2. The recommended method according to claim 1, characterized in that, The acquisition of IPTV user viewing behavior data includes program identifier, program name, program type, viewing time, and viewing duration, including: Connect to the business server and send a data retrieval request to the business server. The data retrieval request carries the IPTV user's account. The service server receives the viewing behavior data of IPTV users returned based on the data acquisition request.
3. The recommended method according to claim 1, characterized in that, The acquisition of IPTV user viewing behavior data includes program identifier, program name, program type, viewing time, and viewing duration, including: Using a pre-defined word embedding model, the program identifier, program content, program name, and program type are encoded respectively, resulting in vectors for the program identifier, program name, and program type. The viewing time and viewing duration are encoded using a time-series model, resulting in vectors for viewing time and viewing duration, respectively.
4. The recommended method according to claim 1, characterized in that, The process involves inputting the IPTV user's viewing behavior feature vector into a retrieval-enhanced generative model. Through this model, a similarity search is performed on the IPTV user's viewing behavior feature vector within the knowledge graph. This yields the similarity score between each entity in the knowledge graph and the IPTV user. The viewing preference information associated with the entity with the highest similarity is then obtained. Feature extraction is performed on this viewing preference information to obtain a feature vector, including: Call the input interface of the retrieval enhancement generative model, and input the viewing behavior feature vector of IPTV users into the retrieval enhancement generative model through the input interface; By using a retrieval-enhanced generative model, similarity retrieval is performed on the viewing behavior feature vectors of IPTV users in the knowledge graph. This yields the similarity between each entity in the knowledge graph and the IPTV user. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information.
5. The recommended method according to claim 1, characterized in that, The process involves fusing the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector. This fused feature vector is then processed using a large language model to generate the IPTV user's viewing needs. Finally, IPTV video content matching these needs is selected from the IPTV video content library and recommended to the IPTV user. The IPTV user's viewing behavior feature vector and viewing preference information feature vector are input into a preset encoder. The encoder's attention mechanism is used to fuse the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector. By processing and fusing feature vectors through a large language model, the viewing needs of IPTV users are generated. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
6. The recommended method according to claim 5, characterized in that, The process involves processing and fusing feature vectors using a large language model to generate viewing requirements for IPTV users, filtering IPTV video content that matches these requirements from the IPTV video content library, and recommending such content to IPTV users. This includes: Obtain the loading instruction, and load the large language model from the model file using the loading instruction, and input the fused feature vector into the trained large language model; By processing and fusing feature vectors through a large language model, the viewing needs of IPTV users are generated. These viewing needs are then matched with video metadata in the IPTV video content library to obtain matching results. From these results, IPTV video content that meets the viewing needs is selected and recommended to IPTV users.
7. The recommended method according to claim 1, characterized in that, After fusing the IPTV user's viewing behavior feature vector and viewing preference information feature vector to generate a fused feature vector, processing the fused feature vector through a large language model to generate the IPTV user's viewing needs, filtering IPTV video content that meets the viewing needs from the IPTV video content library, and recommending IPTV video content that meets the viewing needs to the IPTV user, the recommendation method includes: Obtain recommended IPTV video content that matches your viewing needs, and store the recommended results in a preset storage area.
8. A recommendation device based on a retrieval-enhanced generative model, characterized in that, Applied to electronic devices, including: The acquisition module is used to acquire viewing behavior data of IPTV users. The viewing behavior data includes program identifier, program name, program type, viewing time and viewing duration. The encoding module is used to encode the program identifier, program content, program name, and program type respectively, and to obtain vectors for the program identifier, program name, and program type respectively. It also encodes the viewing time and viewing duration respectively, and to obtain vectors for the viewing time and viewing duration respectively. The splicing module is used to splice the vectors of program identifier, program name, program type, viewing time, and viewing duration to obtain the viewing behavior feature vector of IPTV users. The extraction module is used to input the viewing behavior feature vector of IPTV users into the retrieval enhancement generation model. Through the retrieval enhancement generation model, similarity retrieval is performed on the viewing behavior feature vector of IPTV users in the knowledge graph to obtain the similarity between each entity in the knowledge graph and IPTV users. The viewing preference information associated with the entity with the highest similarity is obtained, and feature extraction is performed on the viewing preference information to obtain the feature vector of the viewing preference information. The recommendation module is used to fuse the feature vectors of IPTV users' viewing behavior and viewing preference information to generate a fused feature vector. The fused feature vector is then processed by a large language model to generate the viewing needs of IPTV users. IPTV video content that meets the viewing needs is then selected from the IPTV video content library and recommended to IPTV users.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the recommended method as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the recommended method as described in any one of claims 1 to 7.