IPTV new content recommendation method, device and equipment and storage medium
By processing multimodal information of new IPTV content using a large language model, content summaries and audience profiles are generated. Combined with a fusion model and a matching degree model, the content is recommended to preferred users, solving the problem of low efficiency in recommending new IPTV content and achieving efficient and reliable recommendation results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-25
- Publication Date
- 2026-04-07
AI Technical Summary
The current process of recommending new IPTV content is cumbersome, resulting in low efficiency. It mainly relies on manual methods, which consume a lot of human resources and time.
The system employs a large language model to process multimodal information of new IPTV content, generating content summaries, fine-grained tags, and audience profiles. It then uses a fusion model and retrieval components to obtain similar content and combines interest, collaboration, and attribute matching models to recommend content to preferred users.
It reduces the time required to recommend new IPTV content, improves recommendation efficiency, and enhances the reliability and credibility of recommendations. Users can quickly discover new content, increasing its exposure and utilization.
Smart Images
Figure CN121808035A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of artificial intelligence technology and IPTV technology, and in particular to a method, apparatus, device and storage medium for recommending new IPTV content. Background Technology
[0002] IPTV is an abbreviation for Interactive Personal Television. As an emerging product that integrates internet technology with traditional television functions, IPTV breaks the limitations of traditional one-way television transmission. At the same time, new IPTV content is constantly emerging, enriching users' viewing resources and enhancing their viewing experience.
[0003] However, the current process for recommending new IPTV content is cumbersome, hindering the improvement of recommendation efficiency. This is because existing technologies primarily rely on manual recommendations, which consume significant human and time resources, increasing the recommendation time and thus reducing efficiency. Summary of the Invention
[0004] This application provides a method, apparatus, device, and storage medium for recommending new IPTV content, in order to solve the technical problem that the existing IPTV content recommendation process is cumbersome and not conducive to improving the efficiency of IPTV content recommendation.
[0005] In a first aspect, embodiments of this application provide an IPTV new content recommendation method, applied to electronic devices, the IPTV new content recommendation method comprising: Acquire multimodal information about new IPTV content, which refers to the audiovisual content added by the IPTV platform within the target period; By processing the multimodal information of new IPTV content and the prompts for the summary generation task using a large language model, a content summary of the new IPTV content is obtained. By processing the multimodal information of new IPTV content and the prompts for the tagging task using a large language model, multiple fine-grained tags of the new IPTV content are obtained. By processing the multimodal information of new IPTV content and the prompts for the audience analysis task using a large language model, an audience profile of the new IPTV content is obtained. By processing the multimodal information of new IPTV content and the prompts for the query statement generation task using a large language model, a similar content query statement of the new IPTV content is obtained. A fusion model is used to fuse the embedding vectors of the content summary, the embedding vectors of multiple fine-grained tags, and the embedding vectors of the audience profile to obtain the fusion vector of the new IPTV content. A normalization model is used to normalize the fusion vector of the new IPTV content to obtain the semantic vector of the new IPTV content. The semantic vector of the new IPTV content and the similar content query statement are transmitted to the retrieval interface of the vector database through the retrieval component. The semantic vector of the new IPTV content is retrieved by using the similar content query statement through the retrieval interface to obtain the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than the preset value is selected as the similar content of the new IPTV content. The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context is then reasoned through a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database based on the group tags. The interest matching degree model is used to obtain the interest matching degree of each candidate user. The collaborative matching degree model is used to generate the collaborative matching degree of each candidate user. The attribute matching degree model is used to obtain the attribute matching degree of each candidate user. Based on the interest matching degree, collaborative matching degree, attribute matching degree and comprehensive model of each candidate user, the comprehensive matching degree of each candidate user is generated. Candidate users with a comprehensive matching degree higher than the preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminal corresponding to the preferred users.
[0006] In one possible implementation of the first aspect, descriptive information of similar content and user behavior data of similar content are concatenated into a context. A large language model is used to reason about the context to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are retrieved from the user database based on the recommendation strategy, including: The descriptive information of similar content is composed of similar content identifiers, similar content titles, similar content tags, and similar content click-through rates. The user behavior data of similar content is composed of similar content viewer identifiers, similar content click counts, similar content viewing durations, and similar content completion rates. The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context is then used to infer the reasoning of the context through a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database according to the recommendation strategy.
[0007] In one possible implementation of the first aspect, the fusion model is as follows: Vfinal = 0.5 × V1 + 0.3 ×V2 + 0.2 ×V3; Vfinal is the fusion vector of the new IPTV content; V1 is the embedding vector of the content summary of the new IPTV content; V2 is the embedding vector of multiple fine-grained tags of the new IPTV content; V3 is the embedding vector of the audience profile of the new IPTV content. The normalization model is: ; Vnorm is the semantic vector of the new IPTV content, Vfinal is the fusion vector of the new IPTV content, and ||Vfinal||2 is the Euclidean norm of the fusion vector of the new IPTV content.
[0008] In one possible implementation of the first aspect, the interest matching model is as follows: ; F i This represents the interest matching degree corresponding to the i-th candidate user; the lower the interest matching degree corresponding to the i-th candidate user, the weaker the overlap between the multiple interest tags of the i-th candidate user and the multiple fine-grained tags of the new IPTV content; the higher the interest matching degree corresponding to the i-th candidate user, the stronger the overlap between the multiple interest tags of the i-th candidate user and the multiple fine-grained tags of the new IPTV content. The higher the degree of overlap between a user's multiple interest tags and multiple fine-grained tags of new IPTV content; A i The number of tags representing the intersection between multiple interest tags of the i-th candidate user and multiple fine-grained tags of new IPTV content; L i This represents the number of tags that intersect between multiple interest tags of the i-th candidate user and multiple fine-grained tags of new IPTV content.
[0009] In one possible implementation of the first aspect, the collaborative matching degree model is as follows: ; J i This represents the collaborative matching degree corresponding to the i-th candidate user; the lower the collaborative matching degree corresponding to the i-th candidate user, the weaker the preference of the i-th candidate user for new IPTV content; the higher the attribute matching degree corresponding to the i-th candidate user, the stronger the preference of the i-th candidate user for new IPTV content. B i This indicates the number of similar content viewed by each candidate user.
[0010] In one possible implementation of the first aspect, the attribute matching degree model is as follows: The attribute matching degree model is as follows: ; Ei This represents the attribute matching degree corresponding to the i-th candidate user. The lower the attribute matching degree corresponding to the i-th candidate user, the weaker the fit between the i-th candidate user and the new IPTV content in terms of age and gender. The higher the attribute matching degree corresponding to the i-th candidate user, the stronger the fit between the i-th candidate user and the new IPTV content in terms of age and gender. C i C represents the age matching degree corresponding to the i-th candidate user; when the age of the i-th candidate user is within the target age range, C i C is 1; when the age of the i-th candidate user is not within the target age range, C i =0; D i C represents the gender matching degree corresponding to the i-th candidate user; when the gender of the i-th candidate user is the target gender, C i C is 1; when the gender of the i-th candidate user is not the target gender, C i It is 0.
[0011] In one possible implementation of the first aspect, the synthesis model is as follows: ; L i This represents the overall matching degree corresponding to the i-th candidate user. The lower the overall matching degree corresponding to the i-th candidate user, the weaker the overall matching effect of the i-th candidate user in terms of interest matching degree, collaboration matching degree, and attribute matching degree. The higher the overall matching degree corresponding to the i-th candidate user, the stronger the overall matching effect of the i-th candidate user in terms of interest matching degree, collaboration matching degree, and attribute matching degree. F i This represents the interest matching degree corresponding to the i-th candidate user; J i This represents the collaborative matching degree corresponding to the i-th candidate user; E i This represents the attribute matching degree corresponding to the i-th candidate user; W1 represents the first weighting coefficient; W2 represents the second weighting coefficient; W3 represents the third weighting coefficient, and the sum of W1, W2, and W3 is 1; W1, W2, and W3 are all greater than zero.
[0012] Secondly, embodiments of this application provide an IPTV new content recommendation device, applied to electronic devices, including: The first acquisition module is used to acquire multimodal information of new IPTV content, which refers to the audiovisual content added by the IPTV platform within the target period. The processing module is used to process the multimodal information of new IPTV content and the prompt words for the summary generation task through a large language model to obtain a content summary of the new IPTV content; to process the multimodal information of new IPTV content and the prompt words for the tagging task through a large language model to obtain multiple fine-grained tags for the new IPTV content; to process the multimodal information of new IPTV content and the prompt words for the audience analysis task through a large language model to obtain an audience profile of the new IPTV content; and to process the multimodal information of new IPTV content and the prompt words for the query statement generation task through a large language model to obtain a similar content query statement for the new IPTV content. The retrieval module uses a fusion model to fuse the embedding vectors of the content summary, multiple fine-grained tags, and audience profiles to obtain the fusion vector of the new IPTV content. A normalization model is then used to normalize this fusion vector to obtain the semantic vector of the new IPTV content. The retrieval component transmits the semantic vector of the new IPTV content and similar content query statements to the retrieval interface of the vector database. The retrieval interface uses the similar content query statements to retrieve the semantic vector of the new IPTV content, obtaining the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than a preset value is selected as similar content to the new IPTV content. The second acquisition module is used to concatenate the descriptive information of similar content and the user behavior data of similar content into a context, reason about the context through a large language model, generate a recommendation strategy, obtain group tags from the recommendation strategy, and obtain multiple candidate users from the user database based on the group tags. The recommendation module is used to obtain the interest matching degree of each candidate user through the interest matching degree model, the collaborative matching degree model, and the attribute matching degree model. Based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model of each candidate user, a comprehensive matching degree is generated for each candidate user. Candidate users with a comprehensive matching degree higher than the preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminal corresponding to the preferred 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 IPTV new content recommendation 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 IPTV new content recommendation 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 execute the IPTV new content recommendation method described in the first aspect above.
[0016] The beneficial effects of this application's embodiments are twofold. Firstly, through the interest matching degree model, the interest matching degree corresponding to each candidate user is obtained; through the collaborative matching degree model, the collaborative matching degree corresponding to each candidate user is generated; and through the attribute matching degree model, the attribute matching degree corresponding to each candidate user is obtained. Based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model corresponding to each candidate user, a comprehensive matching degree corresponding to each candidate user is generated. Candidate users with a comprehensive matching degree higher than a preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminals corresponding to these preferred users. Since there is no need for manual recommendation of new IPTV content, this reduces I... The timing of PTV new content recommendations helps improve the efficiency of IPTV new content recommendations. On the other hand, the lower the overall matching degree of a candidate user, the weaker the overall matching effect of the candidate user in terms of interest matching, collaboration matching, and attribute matching. The higher the overall matching degree of a candidate user, the stronger the overall matching effect of the candidate user in terms of interest matching, collaboration matching, and attribute matching. Selecting candidate users with an overall matching degree higher than the preset matching degree as preferred users means that the preferred users have met the expectations in terms of interest matching, collaboration matching, and attribute matching. Recommending IPTV new content to the IPTV terminals corresponding to the preferred users helps improve the reliability of recommending IPTV new content. 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 This application scenario diagram illustrates the IPTV new content recommendation method provided in the embodiments of this application. Figure 2 This is a flowchart illustrating the IPTV new content recommendation method provided in an embodiment of this application; Figure 3A flowchart illustrating the implementation of S205 provided in this application embodiment; Figure 4 A schematic block diagram of an IPTV new content recommendation device provided in an embodiment 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] The IPTV new content recommendation method provided in this application embodiment can be applied to electronic devices such as servers, mobile phones, tablets, wearable devices, in-vehicle devices, laptops, and netbooks. This application embodiment does not impose any restrictions on the specific type of electronic device.
[0021] Please see Figure 1 , Figure 1 The application scenario diagram of the IPTV new content recommendation method provided in the embodiments of this application is described in detail below: Electronic devices access the content management system to retrieve new IPTV content.
[0022] In this embodiment of the application, the electronic device accesses the content management system and obtains new IPTV content from the content management system, which can reduce the acquisition time of new IPTV content and improve the acquisition efficiency of new IPTV content.
[0023] Please see Figure 2 , Figure 2 This is a flowchart illustrating the IPTV new content recommendation method provided in this application embodiment, which can be applied to electronic devices.
[0024] like Figure 2 As shown in the embodiments of this application, the IPTV new content recommendation method includes the following steps, which are detailed below: S201, Obtain multimodal information of new IPTV content. New IPTV content refers to the audiovisual content added by the IPTV platform within the target period. This includes acquiring multimodal information about new IPTV content. New IPTV content refers to the audiovisual content added by the IPTV platform within the target period, including: Access the content management system, retrieve new IPTV content from the content management system, process the new IPTV content through the preprocessing component to obtain the multimodal information of the new IPTV content, and obtain the summary generation task, tag annotation task, audience analysis task, and query statement generation task of the new IPTV content from the task information. The new IPTV content is the audiovisual content added by the IPTV platform within the target period.
[0025] The multimodal information of new IPTV content includes video metadata, keyframe images, and text information.
[0026] The video metadata for new IPTV content includes the title, duration, production year, region, language, resolution, and bitrate.
[0027] The process of extracting keyframe images for new IPTV content is as follows: A dual-mode keyframe image extraction strategy is employed. The first mode uses fixed-time sampling, extracting one frame every 30 seconds to ensure comprehensive coverage of new IPTV content. The second mode is intelligent sampling based on scene change detection, using a histogram differencing algorithm to calculate the color distribution differences between adjacent frames. When the difference exceeds a threshold, a scene change is identified, and that frame is extracted as the keyframe image for the new IPTV content. After merging and deduplicating the extraction results from both modes, 200 to 300 keyframe images can typically be extracted from each 2-hour movie. The extracted keyframe images are uniformly saved in JPEG format, with a resolution adjusted to 1280 pixels horizontally and 720 pixels vertically.
[0028] The text information for new IPTV content is transcribed using a speech recognition engine. First, the audio stream is separated from the video file and converted to WAV format with a 16kHz sampling rate. Then, it is segmented into 10-second audio clips and fed into the speech recognition engine to obtain the speech recognition results. These results include the speech-recognized text of the new IPTV content, a timestamp, and a confidence score. The speech-recognized text of the new IPTV content, along with the video title, description, and subtitles, is then integrated to obtain the full text information for the new IPTV content.
[0029] Among them, new IPTV content refers to audiovisual content added by the IPTV platform within the target period. The target period is set to one of the following: 24 hours, 12 hours, 6 hours, 3 hours, or 1 hour.
[0030] S202: By processing the multimodal information of new IPTV content and the prompt words for the summary generation task using a large language model, a content summary of the new IPTV content is obtained; by processing the multimodal information of new IPTV content and the prompt words for the tagging task using a large language model, multiple fine-grained tags of the new IPTV content are obtained; by processing the multimodal information of new IPTV content and the prompt words for the audience analysis task using a large language model, an audience profile of the new IPTV content is obtained; by processing the multimodal information of new IPTV content and the prompt words for the query statement generation task using a large language model, a similar content query statement of the new IPTV content is obtained. Specifically, the process involves: processing multimodal information of new IPTV content using a large language model and using prompts for the summary generation task to obtain a content summary of the new IPTV content; processing multimodal information of new IPTV content using a large language model and using prompts for the tagging task to obtain multiple fine-grained tags for the new IPTV content; processing multimodal information of new IPTV content using a large language model and using prompts for the audience analysis task to obtain an audience profile of the new IPTV content; and processing multimodal information of new IPTV content using a large language model and using prompts for the query statement generation task to obtain similar content query statements for the new IPTV content, including: Obtain prompts for the summary generation task, the tagging task, the audience analysis task, and the query generation task; input the multimodal information of IPTV new content, prompts for the summary generation task, prompts for the tagging task, prompts for the audience analysis task, and prompts for the query generation task into the large language model. By processing the multimodal information of new IPTV content using a large language model and using prompts for the summary generation task, a content summary of the new IPTV content is obtained. By processing the multimodal information of new IPTV content using a large language model and using prompts for the tagging task, multiple fine-grained tags for the new IPTV content are obtained. By processing the multimodal information of new IPTV content using a large language model and using prompts for the audience analysis task, an audience profile of the new IPTV content is obtained. By processing the multimodal information of new IPTV content using a large language model and using prompts for the query statement generation task, similar content query statements for the new IPTV content are obtained.
[0031] For ease of explanation, the following example is provided: For example, the prompt words for the abstract generation task are: Based on the video information below, please generate a summary of no more than 200 words, highlighting the core plot, theme, and emotional tone. Requirements: 1. Use objective descriptions and avoid subjective evaluations; 2. Emphasize the core conflicts and turning points; 3. Summarize the overall emotional atmosphere; 4. If it is a series, explain the position of this episode within the overall story.
[0032] For example: the prompt words for the labeling task are: Based on the following summary, please extract 10-15 fine-grained tags, including the following dimensions: 1. Genre, such as action, romance, suspense, science fiction; 2. Emotional tone, such as joy, sadness, tension, or warmth; 3. Rhythm style, such as fast rhythm, slow rhythm, and a balanced pace; 4. Suitable for groups such as teenagers, middle-aged people, the elderly, and families; 5. Core themes, such as friendship, growth, revenge, and redemption; 6. Scene style, such as urban, rural, ancient, or futuristic; 7. Production quality, such as high-quality production, low-budget, blockbuster. Output format: Tag1, Tag2, TagN, each tag separated by a vertical bar. Do not number or explain. N is a positive integer.
[0033] For example: the prompts for the audience analysis task are: Based on the following characteristics—summary and tags—analyze the characteristics of the most suitable target audience for this content. Including: 1. Age groups, such as children aged 0-12, teenagers aged 13-18, young adults aged 19-35, middle-aged adults aged 36-55, and elderly adults aged 56 and above; 2. Gender bias, such as male preference, female preference, gender balance, or no obvious preference; 3. Interests and preferences, such as sports, music, travel, and history; 4. Viewing scenarios, such as personal enjoyment, family viewing, friends gathering, or study reference; multiple selections are allowed. 5. Educational background preference, such as junior high school or above, high school or above, university or above; 6. Motivations for watching, such as entertainment and relaxation, emotional resonance, knowledge learning, and curiosity and exploration.
[0034] For example: the prompt for the query statement to generate the task is: Based on the abstract, please generate 3-5 semantic queries for retrieving similar content. Each query should describe the characteristics of the content from a different perspective.
[0035] Requirements: 1. The first query should focus on the core plot and theme; 2. The second query focuses on the emotional tone and atmosphere; 3. The third query focuses on the relationships between characters and their characteristics. 4. The fourth query statement focuses on visual style and production characteristics; 5. The fifth query focuses on the target audience and viewing experience.
[0036] Output format: Each query statement should be 20-50 characters long, separated by newline characters.
[0037] This application adopts a privately deployed large language model, where all data and computation are completed in the operator's local environment, avoiding the risk of leakage of user data and content data, and complying with the operator's data security specifications.
[0038] S203 employs a fusion model to fuse the embedding vectors of the content summary, multiple fine-grained tags, and audience profiles to obtain the fusion vector of the new IPTV content. A normalization model is then used to normalize the fusion vector of the new IPTV content to obtain its semantic vector. The semantic vector of the new IPTV content and the similar content query statement are transmitted to the retrieval interface of the vector database through the retrieval component. The semantic vector of the new IPTV content is retrieved using the similar content query statement through the retrieval interface to obtain the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than a preset value is selected as the similar content of the new IPTV content. The retrieval component is based on RAG technology. RAG stands for Retrieval Augmented Generation.
[0039] The historical content refers to audiovisual content that was uploaded to the IPTV platform one day prior.
[0040] The fusion model is as follows: Vfinal = 0.5 × V1 + 0.3 ×V2 + 0.2 ×V3; Vfinal is the fusion vector of the new IPTV content; V1 is the embedding vector of the content summary of the new IPTV content; V2 is the embedding vector of multiple fine-grained tags of the new IPTV content; V3 is the embedding vector of the audience profile of the new IPTV content. The normalization model is: ; Vnorm is the semantic vector of the new IPTV content, Vfinal is the fusion vector of the new IPTV content, and ||Vfinal||2 is the Euclidean norm of the fusion vector of the new IPTV content.
[0041] S204: Concatenate the descriptive information of similar content and the user behavior data of similar content into a context, reason about the context through a large language model, generate a recommendation strategy, obtain group tags from the recommendation strategy, and obtain multiple candidate users from the user database based on the group tags; Specifically, descriptive information of similar content and user behavior data of similar content are concatenated into a context. A large language model is used to reason about this context, generating a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are retrieved from the user database based on the recommendation strategy, including: The descriptive information of similar content is composed of similar content identifiers, similar content titles, similar content tags, and similar content click-through rates. The user behavior data of similar content is composed of similar content viewer identifiers, similar content click counts, similar content viewing durations, and similar content completion rates. The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context is then used to infer the reasoning of the context through a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database according to the recommendation strategy.
[0042] The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context and prompt words of the recommendation task are processed by a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database based on the group tags.
[0043] Among them, group tags are tags for user groups.
[0044] For example, the group tag is the tag for users who watch for more than 2 hours a day on average; For example, a group tag is a tag for users who watch for more than 3 hours a day on average.
[0045] For example, a group tag is a tag for users who watch for more than 5 hours a day on average.
[0046] S205: Using the interest matching degree model, obtain the interest matching degree corresponding to each candidate user; using the collaborative matching degree model, generate the collaborative matching degree corresponding to each candidate user; using the attribute matching degree model, obtain the attribute matching degree corresponding to each candidate user; based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model corresponding to each candidate user, generate the comprehensive matching degree corresponding to each candidate user; select candidate users whose comprehensive matching degree is higher than the preset matching degree as preferred users, and recommend new IPTV content to the IPTV terminal corresponding to the preferred users.
[0047] Among them, the IPTV terminal corresponding to the preferred user is the hardware device directly operated by the preferred user when using IPTV service. As the core interaction medium between the preferred user and the IPTV platform, the IPTV terminal undertakes multiple tasks such as signal reception, decoding and presentation, and function control.
[0048] The overall matching score of candidate users is used to evaluate their comprehensive matching effect in terms of interest matching, collaboration matching, and attribute matching. A lower overall matching score indicates a weaker overall matching effect across these three aspects; conversely, a higher overall matching score indicates a stronger overall matching effect. Selecting candidate users with an overall matching score higher than the preset matching score as preferred users signifies that these preferred users have met expectations in all three aspects. Recommending new IPTV content to the IPTV terminals corresponding to these preferred users helps them quickly discover new IPTV content, reducing the time cost of active searching and thus improving the efficiency of acquiring new IPTV content. Furthermore, it increases the exposure and click-through rate of new IPTV content, thereby improving its utilization rate.
[0049] The interest matching model is as follows: ; F i This represents the interest matching degree corresponding to the i-th candidate user; the lower the interest matching degree corresponding to the i-th candidate user, the weaker the overlap between the multiple interest tags of the i-th candidate user and the multiple fine-grained tags of the new IPTV content; the higher the interest matching degree corresponding to the i-th candidate user, the stronger the overlap between the multiple interest tags of the i-th candidate user and the multiple fine-grained tags of the new IPTV content. The higher the degree of overlap between a user's multiple interest tags and multiple fine-grained tags of new IPTV content; A i The number of tags representing the intersection between multiple interest tags of the i-th candidate user and multiple fine-grained tags of new IPTV content; L i This represents the number of tags that intersect between multiple interest tags of the i-th candidate user and multiple fine-grained tags of new IPTV content.
[0050] The collaborative matching degree model is as follows: ; J i This represents the collaborative matching degree corresponding to the i-th candidate user; the lower the collaborative matching degree corresponding to the i-th candidate user, the weaker the preference of the i-th candidate user for new IPTV content; the higher the attribute matching degree corresponding to the i-th candidate user, the stronger the preference of the i-th candidate user for new IPTV content. B i This indicates the number of similar content viewed by each candidate user.
[0051] The upper limit of the collaborative matching degree corresponding to the i-th candidate user is 1, which means that the i-th candidate user has viewed 5 or more similar content. The collaborative matching degree corresponding to the i-th candidate user is close to 1, which means that the i-th candidate user has been exposed to similar content for a long time. It is not a random click, but a real need verified by multiple viewing behaviors.
[0052] The recommendation strategy includes target age group and target gender.
[0053] The attribute matching degree model is as follows: ; E i This represents the attribute matching degree corresponding to the i-th candidate user. The lower the attribute matching degree corresponding to the i-th candidate user, the weaker the fit between the i-th candidate user and the new IPTV content in terms of age and gender. The higher the attribute matching degree corresponding to the i-th candidate user, the stronger the fit between the i-th candidate user and the new IPTV content in terms of age and gender. C i C represents the age matching degree corresponding to the i-th candidate user; when the age of the i-th candidate user is within the target age range, C i C is 1; when the age of the i-th candidate user is not within the target age range, C i =0; D i C represents the gender matching degree corresponding to the i-th candidate user; when the gender of the i-th candidate user is the target gender, C i C is 1; when the gender of the i-th candidate user is not the target gender, C i It is 0.
[0054] The integrated model is as follows: ; L i This represents the overall matching degree corresponding to the i-th candidate user. The lower the overall matching degree corresponding to the i-th candidate user, the weaker the overall matching effect of the i-th candidate user in terms of interest matching degree, collaboration matching degree, and attribute matching degree. The higher the overall matching degree corresponding to the i-th candidate user, the stronger the overall matching effect of the i-th candidate user in terms of interest matching degree, collaboration matching degree, and attribute matching degree. F i This represents the interest matching degree corresponding to the i-th candidate user; J i This represents the collaborative matching degree corresponding to the i-th candidate user; E i This represents the attribute matching degree corresponding to the i-th candidate user; W1 represents the first weighting coefficient; W2 represents the second weighting coefficient; W3 represents the third weighting coefficient, and the sum of W1, W2, and W3 is 1; W1, W2, and W3 are all greater than zero.
[0055] For ease of explanation, the following example is provided: For example, W1, W2, and W3 are 0.5, 0.3, and 0.2 respectively; the interest matching degree, collaboration matching degree, and attribute matching degree corresponding to the i-th candidate user are respectively... The values are 0.68, 0.6, and 1. At this time, L i = 0.5 × 0.68 + 0.3 × 0.6 + 0.2 × 1 = 0.34 + 0.18 + 0.2 = 0.72. A higher overall matching degree indicates a better overall matching effect for the i-th candidate user in terms of interest matching, collaboration matching, and attribute matching, and a greater probability of successfully recommending new IPTV content to the i-th candidate user.
[0056] The beneficial effects of this application's embodiments are twofold. Firstly, through the interest matching degree model, the interest matching degree corresponding to each candidate user is obtained; through the collaborative matching degree model, the collaborative matching degree corresponding to each candidate user is generated; and through the attribute matching degree model, the attribute matching degree corresponding to each candidate user is obtained. Based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model corresponding to each candidate user, a comprehensive matching degree corresponding to each candidate user is generated. Candidate users with a comprehensive matching degree higher than a preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminals corresponding to these preferred users. Since there is no need for manual recommendation of new IPTV content, this reduces I... The timing of PTV new content recommendations helps improve the efficiency of IPTV new content recommendations. On the other hand, the lower the overall matching degree of a candidate user, the weaker the overall matching effect of the candidate user in terms of interest matching, collaboration matching, and attribute matching. The higher the overall matching degree of a candidate user, the stronger the overall matching effect of the candidate user in terms of interest matching, collaboration matching, and attribute matching. Selecting candidate users with an overall matching degree higher than the preset matching degree as preferred users means that the preferred users have met the expectations in terms of interest matching, collaboration matching, and attribute matching. Recommending IPTV new content to the IPTV terminals corresponding to the preferred users helps improve the reliability of recommending IPTV new content.
[0057] Please see Figure 3 , Figure 3 The implementation flowchart of S205 provided in the embodiments of this application is described in detail below: S301. The interest matching degree of each candidate user is obtained through the interest matching degree model, the collaborative matching degree of each candidate user is generated through the collaborative matching degree model, and the attribute matching degree of each candidate user is obtained through the attribute matching degree model. S302, based on the interest matching degree, collaborative matching degree, attribute matching degree and comprehensive model of each candidate user, generate the comprehensive matching degree of each candidate user, select the candidate user whose comprehensive matching degree is higher than the preset matching degree as the preferred user, obtain the current time, and recommend new IPTV content to the IPTV terminal corresponding to the preferred user when the current time is the push time.
[0058] In this embodiment, candidate users with a comprehensive matching degree higher than a preset matching degree are selected as preferred users. The current time is obtained, and when the current time is the push time, new IPTV content is recommended to the IPTV terminal corresponding to the preferred user, which helps to improve the recommendation quality of new IPTV content.
[0059] For the IPTV new content recommendation method described in the above embodiments, please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic block diagram of an IPTV new content recommendation device provided in an embodiment of this application. Figure 4 The IPTV new content recommendation 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 IPTV new content recommendation device 400 shown will be described in detail. The IPTV new content recommendation device 400 may include a first acquisition module 401, a processing module 402, a retrieval module 403, a second acquisition module 404, and a recommendation module 405.
[0060] The first acquisition module 401 is used to acquire multimodal information of new IPTV content, which refers to the audiovisual content added by the IPTV platform within the target period. The processing module 402 is used to process the multimodal information of new IPTV content and the prompt words for the summary generation task through a large language model to obtain a content summary of the new IPTV content; to process the multimodal information of new IPTV content and the prompt words for the tagging task through a large language model to obtain multiple fine-grained tags for the new IPTV content; to process the multimodal information of new IPTV content and the prompt words for the audience analysis task through a large language model to obtain an audience profile of the new IPTV content; and to process the multimodal information of new IPTV content and the prompt words for the query statement generation task through a large language model to obtain a similar content query statement for the new IPTV content. The retrieval module 403 is used to fuse the embedding vector of the content summary, the embedding vector of multiple fine-grained tags, and the embedding vector of the audience profile using a fusion model to obtain the fusion vector of the new IPTV content. It then uses a normalization model to normalize the fusion vector of the new IPTV content to obtain the semantic vector of the new IPTV content. The semantic vector of the new IPTV content and the similar content query statement are transmitted to the retrieval interface of the vector database through the retrieval component. The semantic vector of the new IPTV content is retrieved using the similar content query statement through the retrieval interface to obtain the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than a preset value is selected as the similar content of the new IPTV content. The second acquisition module 404 is used to concatenate the description information of similar content and the user behavior data of similar content into a context, reason about the context through a large language model, generate a recommendation strategy, obtain group tags from the recommendation strategy, and obtain multiple candidate users from the user database based on the group tags. The recommendation module 405 is used to obtain the interest matching degree corresponding to each candidate user through the interest matching degree model, generate the collaborative matching degree corresponding to each candidate user through the collaborative matching degree model, obtain the attribute matching degree corresponding to each candidate user through the attribute matching degree model, generate the comprehensive matching degree corresponding to each candidate user based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model, select candidate users whose comprehensive matching degree is higher than the preset matching degree as preferred users, and recommend new IPTV content to the IPTV terminal corresponding to the preferred users.
[0061] 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.
[0062] The beneficial effects of this application's embodiments are twofold. Firstly, through the interest matching degree model, the interest matching degree corresponding to each candidate user is obtained; through the collaborative matching degree model, the collaborative matching degree corresponding to each candidate user is generated; and through the attribute matching degree model, the attribute matching degree corresponding to each candidate user is obtained. Based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model corresponding to each candidate user, a comprehensive matching degree corresponding to each candidate user is generated. Candidate users with a comprehensive matching degree higher than a preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminals corresponding to these preferred users. Since there is no need for manual recommendation of new IPTV content, this reduces I... The timing of PTV new content recommendations helps improve the efficiency of IPTV new content recommendations. On the other hand, the lower the overall matching degree of a candidate user, the weaker the overall matching effect of the candidate user in terms of interest matching, collaboration matching, and attribute matching. The higher the overall matching degree of a candidate user, the stronger the overall matching effect of the candidate user in terms of interest matching, collaboration matching, and attribute matching. Selecting candidate users with an overall matching degree higher than the preset matching degree as preferred users means that the preferred users have met the expectations in terms of interest matching, collaboration matching, and attribute matching. Recommending IPTV new content to the IPTV terminals corresponding to the preferred users helps improve the reliability of recommending IPTV new content.
[0063] 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.
[0064] 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.
[0065] 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.
[0066] 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 multimodal information about new IPTV content, which refers to the audiovisual content added by the IPTV platform within the target period; By processing the multimodal information of new IPTV content and the prompts for the summary generation task using a large language model, a content summary of the new IPTV content is obtained. By processing the multimodal information of new IPTV content and the prompts for the tagging task using a large language model, multiple fine-grained tags of the new IPTV content are obtained. By processing the multimodal information of new IPTV content and the prompts for the audience analysis task using a large language model, an audience profile of the new IPTV content is obtained. By processing the multimodal information of new IPTV content and the prompts for the query statement generation task using a large language model, a similar content query statement of the new IPTV content is obtained. A fusion model is used to fuse the embedding vectors of the content summary, the embedding vectors of multiple fine-grained tags, and the embedding vectors of the audience profile to obtain the fusion vector of the new IPTV content. A normalization model is used to normalize the fusion vector of the new IPTV content to obtain the semantic vector of the new IPTV content. The semantic vector of the new IPTV content and the similar content query statement are transmitted to the retrieval interface of the vector database through the retrieval component. The semantic vector of the new IPTV content is retrieved by using the similar content query statement through the retrieval interface to obtain the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than the preset value is selected as the similar content of the new IPTV content. The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context is then reasoned through a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database based on the group tags. The interest matching degree model is used to obtain the interest matching degree of each candidate user. The collaborative matching degree model is used to generate the collaborative matching degree of each candidate user. The attribute matching degree model is used to obtain the attribute matching degree of each candidate user. Based on the interest matching degree, collaborative matching degree, attribute matching degree and comprehensive model of each candidate user, the comprehensive matching degree of each candidate user is generated. Candidate users with a comprehensive matching degree higher than the preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminal corresponding to the preferred users.
[0067] 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. Furthermore, the memory 21 may include both internal and external storage units 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.
[0068] 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.
[0069] 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.
[0070] The computer-readable storage medium stores program code that can be called by a processor to execute the IPTV new content recommendation method described in the above method embodiments.
[0071] The program code includes the code for any step in the IPTV new content recommendation method described in the above method embodiments.
[0072] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.
[0073] Since the computer program stored in the computer-readable storage medium can execute any of the IPTV new content recommendation methods provided in the embodiments of this application, the computer-readable storage medium can achieve the beneficial effects that any of the IPTV new content recommendation methods provided in the embodiments of this application can achieve, as detailed in the preceding embodiments, and will not be repeated here.
[0074] This application provides a computer program product that, when run on an electronic device, causes the electronic device to execute the aforementioned IPTV new content recommendation method.
[0075] 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.
[0076] 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 method for recommending new IPTV content, characterized in that, The IPTV new content recommendation method, applied to electronic devices, includes: Acquire multimodal information about new IPTV content, which refers to the audiovisual content added by the IPTV platform within the target period; By processing the multimodal information of new IPTV content and the prompts for the summary generation task using a large language model, a content summary of the new IPTV content is obtained. By processing the multimodal information of new IPTV content and the prompts for the tagging task using a large language model, multiple fine-grained tags of the new IPTV content are obtained. By processing the multimodal information of new IPTV content and the prompts for the audience analysis task using a large language model, an audience profile of the new IPTV content is obtained. By processing the multimodal information of new IPTV content and the prompts for the query statement generation task using a large language model, a similar content query statement of the new IPTV content is obtained. A fusion model is used to fuse the embedding vectors of the content summary, the embedding vectors of multiple fine-grained tags, and the embedding vectors of the audience profile to obtain the fusion vector of the new IPTV content. A normalization model is used to normalize the fusion vector of the new IPTV content to obtain the semantic vector of the new IPTV content. The semantic vector of the new IPTV content and the similar content query statement are transmitted to the retrieval interface of the vector database through the retrieval component. The semantic vector of the new IPTV content is retrieved by using the similar content query statement through the retrieval interface to obtain the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than the preset value is selected as the similar content of the new IPTV content. The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context is then reasoned through a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database based on the group tags. The interest matching degree model is used to obtain the interest matching degree of each candidate user. The collaborative matching degree model is used to generate the collaborative matching degree of each candidate user. The attribute matching degree model is used to obtain the attribute matching degree of each candidate user. Based on the interest matching degree, collaborative matching degree, attribute matching degree and comprehensive model of each candidate user, the comprehensive matching degree of each candidate user is generated. Candidate users with a comprehensive matching degree higher than the preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminal corresponding to the preferred users.
2. The IPTV new content recommendation method according to claim 1, characterized in that, Descriptive information of similar content and user behavior data of similar content are concatenated into a context. A large language model is used to reason about this context, generating a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are retrieved from the user database based on the recommendation strategy, including: The descriptive information of similar content is composed of similar content identifiers, similar content titles, similar content tags, and similar content click-through rates. The user behavior data of similar content is composed of similar content viewer identifiers, similar content click counts, similar content viewing durations, and similar content completion rates. The descriptive information of similar content and user behavior data of similar content are concatenated into a context. The context is then used to infer the reasoning of the context through a large language model to generate a recommendation strategy. Group tags are obtained from the recommendation strategy, and multiple candidate users are obtained from the user database according to the recommendation strategy.
3. The IPTV new content recommendation method according to claim 1, characterized in that, The fusion model is as follows: Vfinal = 0.5 × V1 + 0.3 ×V2 + 0.2 ×V3; Vfinal is the fusion vector of the new IPTV content; V1 is the embedding vector of the content summary of the new IPTV content; V2 is the embedding vector of multiple fine-grained tags of the new IPTV content; V3 is the embedding vector of the audience profile of the new IPTV content. The normalization model is: ； Vnorm is the semantic vector of the new IPTV content, Vfinal is the fusion vector of the new IPTV content, and ||Vfinal||2 is the Euclidean norm of the fusion vector of the new IPTV content.
4. The IPTV new content recommendation method according to claim 1, characterized in that, Interest matching The model is: ； F i This represents the interest matching degree corresponding to the i-th candidate user; the lower the interest matching degree corresponding to the i-th candidate user, the weaker the overlap between the multiple interest tags of the i-th candidate user and the multiple fine-grained tags of the new IPTV content; the higher the interest matching degree corresponding to the i-th candidate user, the stronger the overlap between the multiple interest tags of the i-th candidate user and the multiple fine-grained tags of the new IPTV content. The higher the degree of overlap between a user's multiple interest tags and multiple fine-grained tags of new IPTV content; A i The number of tags representing the intersection between multiple interest tags of the i-th candidate user and multiple fine-grained tags of new IPTV content; L i This represents the number of tags that intersect between multiple interest tags of the i-th candidate user and multiple fine-grained tags of new IPTV content.
5. The IPTV new content recommendation method according to claim 1, characterized in that, Collaborative matching degree The model is: ； J i This represents the collaborative matching degree corresponding to the i-th candidate user; the lower the collaborative matching degree corresponding to the i-th candidate user, the weaker the preference of the i-th candidate user for new IPTV content. The higher the attribute matching degree of the i-th candidate user, the stronger the i-th candidate user's preference for new IPTV content; B i This indicates the number of similar content viewed by each candidate user.
6. The IPTV new content recommendation method according to claim 1, characterized in that, The attribute matching degree model is as follows: ； E i This represents the attribute matching degree corresponding to the i-th candidate user. The lower the attribute matching degree corresponding to the i-th candidate user, the weaker the fit between the i-th candidate user and the new IPTV content in terms of age and gender. The higher the attribute matching degree corresponding to the i-th candidate user, the stronger the fit between the i-th candidate user and the new IPTV content in terms of age and gender. C i This represents the age matching degree corresponding to the i-th candidate user; When the age of the i-th candidate user falls within the target age range, C i =1; When the age of the i-th candidate user is not within the target age range, C i =0; D i C represents the gender matching degree corresponding to the i-th candidate user; when the gender of the i-th candidate user is the target gender, C i C is 1; when the gender of the i-th candidate user is not the target gender, C i It is 0.
7. The IPTV new content recommendation method according to claim 1, characterized in that, The comprehensive model is as follows: ； L i This represents the overall matching degree corresponding to the i-th candidate user. The lower the overall matching degree corresponding to the i-th candidate user, the weaker the overall matching effect of the i-th candidate user in terms of interest matching degree, collaboration matching degree, and attribute matching degree. The higher the overall matching degree corresponding to the i-th candidate user, the stronger the overall matching effect of the i-th candidate user in terms of interest matching degree, collaboration matching degree, and attribute matching degree. F i This represents the interest matching degree corresponding to the i-th candidate user; J i This represents the collaborative matching degree corresponding to the i-th candidate user; E i This represents the attribute matching degree corresponding to the i-th candidate user; W1 represents the first weighting coefficient; W2 represents the second weighting coefficient; W3 represents the third weighting coefficient, and the sum of W1, W2, and W3 is 1; W1, W2, and W3 are all greater than zero.
8. An IPTV new content recommendation device, characterized in that, Applied to electronic devices, including: The first acquisition module is used to acquire multimodal information of new IPTV content, which refers to the audiovisual content added by the IPTV platform within the target period. The processing module is used to process the multimodal information of new IPTV content and the prompt words for the summary generation task through a large language model to obtain a content summary of the new IPTV content; to process the multimodal information of new IPTV content and the prompt words for the tagging task through a large language model to obtain multiple fine-grained tags for the new IPTV content; to process the multimodal information of new IPTV content and the prompt words for the audience analysis task through a large language model to obtain an audience profile of the new IPTV content; and to process the multimodal information of new IPTV content and the prompt words for the query statement generation task through a large language model to obtain a similar content query statement for the new IPTV content. The retrieval module uses a fusion model to fuse the embedding vectors of the content summary, multiple fine-grained tags, and audience profiles to obtain the fusion vector of the new IPTV content. A normalization model is then used to normalize this fusion vector to obtain the semantic vector of the new IPTV content. The retrieval component transmits the semantic vector of the new IPTV content and similar content query statements to the retrieval interface of the vector database. The retrieval interface uses the similar content query statements to retrieve the semantic vector of the new IPTV content, obtaining the cosine similarity between the new IPTV content and each historical content. Historical content with a cosine similarity higher than a preset value is selected as similar content to the new IPTV content. The second acquisition module is used to concatenate the descriptive information of similar content and the user behavior data of similar content into a context, reason about the context through a large language model, generate a recommendation strategy, obtain group tags from the recommendation strategy, and obtain multiple candidate users from the user database based on the group tags. The recommendation module is used to obtain the interest matching degree of each candidate user through the interest matching degree model, the collaborative matching degree model, and the attribute matching degree model. Based on the interest matching degree, collaborative matching degree, attribute matching degree, and comprehensive model of each candidate user, a comprehensive matching degree is generated for each candidate user. Candidate users with a comprehensive matching degree higher than the preset matching degree are selected as preferred users, and new IPTV content is recommended to the IPTV terminal corresponding to the preferred 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 IPTV new content recommendation 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 the processor, it implements the IPTV new content recommendation method as described in any one of claims 1 to 7.