Method and system for automatically generating and pushing personalized content based on AIGC

Abstract

The invention discloses a personalized content automatic generation and pushing method and system based on AIGC, and mainly relates to the technical field of artificial intelligence content generation. Comprising the following steps: acquiring data information of historical watching records, browsing behaviors and interaction records of a user; according to the obtained data information, analyzing the data through a large language model, and generating a user portrait label and a personalized cue word; based on the generated user portrait label and the personalized cue word, selecting and generating a corresponding creation prompt; on the basis of the generated creation prompt, automatic generation of multi-modal content is completed through a plurality of generation models; and carrying out personalized pushing on the automatically generated multi-modal content, and carrying out feedback collection. The method has the beneficial effects that accurate content customization and pushing can be realized through a user portrait and content automatic generation technology.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence generated content, and specifically to a method and system for automatically generating and pushing personalized content based on AIGC. Background Art

[0002] With the rapid development of artificial intelligence and big data technologies, personalized recommendations based on user interests and behaviors have become an important development trend in the current content industry.

[0003] Existing recommendation systems (such as Netflix's collaborative filtering algorithm and YouTube's deep neural network recommendation) mainly rely on user behavior data (clicks, browsing) or static label systems (such as "action movies" and "comedy movies"), and lack deep semantic understanding of user interests (Reference: Korenetal, 2009).

[0004] Although multimodal generation technologies (such as DALL·E and Midjourney) can generate text, graphics or audio content, the subject consistency between the modalities is poor (for example, the generated music and video styles conflict), and they are not dynamically integrated with user portraits.

[0005] In view of the above problems, how to design an encrypted traffic classification method that takes into account both improving classification accuracy and utilizing the structural information of the traffic itself is an issue that needs to be urgently addressed in this field. Summary of the invention

[0006] The purpose of the present invention is to provide a personalized content automatic generation and push method and system based on AIGC, which combines the personalized content generation and push system of AIGC technology, and realizes accurate content customization and push through user portrait and content automatic generation technology.

[0007] In order to achieve the above object, the present invention is implemented through the following technical solutions: On the one hand, the present invention provides a method for automatically generating and pushing personalized content based on AIGC, comprising the following steps: S1: Obtain data information about the user's historical viewing records, browsing behaviors, and interaction records; S2: Analyze the data using a large language model based on the data information obtained in step S1 to generate user portrait labels and personalized prompt words; S3: Based on the user portrait tag and personalized prompt words generated in step S2, select and generate corresponding creation prompts; S4: Based on the creation prompt generated in step S3, the multimodal content is automatically generated through multiple generation models; S5: Personalize the push of automatically generated multimodal content and collect feedback.

[0008] Preferably, the data information of the user's historical viewing record is obtained including but not limited to: TV series, movies, videos, music, articles, and pictures watched by users; The data information of the user's browsing behavior obtained includes but is not limited to: User's search history and browsing history; The data information of obtaining the user's interaction records includes but is not limited to: User click, favorite, like, and comment behavior data.

[0009] Preferably, the step S2, generating a user portrait tag, is specifically: The Transformer-based multi-head attention mechanism is used to model user behavior sequences: ; in, is the query matrix, is the key-value matrix, is the vector dimension; By calculating the attention weights of different time steps in the behavior sequence, the long-term dependencies of user interests are dynamically extracted, and the final user portrait is expressed as: ; in, It is a three-layer fully connected network with GELU as the activation function, and outputs the probability distribution of user labels.

[0010] Preferably, the step S2, generating personalized prompt words, is specifically: Use the mapping function from label to prompt word and combine it with reinforcement learning to optimize the generation quality: ; in, is the semantic similarity between the prompt word and the user tag, Score the diversity of the prompt words, , is the weight hyperparameter.

[0011] Preferably, in step S3, corresponding creation prompts are selected and generated, specifically: prompts of text-to-image, text-to-sound, and image-to-video are generated according to the media type, and a cross-modal joint loss function is used by jointly training the text-to-image, text-to-sound, and image-to-video models: ; in, The image-text alignment loss calculated for the CLIP model, is the cycle consistency loss, , is the balance coefficient.

[0012] Preferably, the personalized push is specifically: Execute push strategy using contextual bandit algorithm with feedback: ; in, The type of content to be pushed. is the user context feature; The weight parameter for the type of content being pushed. is the noise term, and the parameters are updated through online learning , maximize user click-through rate and viewing time.

[0013] Preferably, the feedback collection is specifically: Collect user interaction feedback and behavior data in real time to further optimize user portraits and recommendation algorithms. The feedback-driven model incremental update mechanism is: The loss function of the feedback-driven model based on user feedback data is: ; in, For users to The actual feedback is: click is recorded as (1), and no click is recorded as (0); The click probability predicted by the model.

[0014] On the other hand, a push system based on the above-mentioned AIGC-based personalized content automatic generation and push method is also provided, including: Data collection module, used to obtain data information about users' historical viewing records, browsing behaviors and interaction records; The data processing module is used to: analyze the data through a large language model based on the acquired data information, generate user portrait tags and personalized prompt words, and select and generate corresponding creation prompts; The automatic generation module is used to: automatically generate multimodal content through multiple generation models based on the generated creation prompts; The personalized push and feedback module is used to: personalize the push of automatically generated multimodal content and collect feedback.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. Cross-modal content generation: The present invention can generate content in multiple modalities (such as text-to-picture, text-to-sound, picture-to-video, etc.) according to user needs. It has strong content creation capabilities and can meet the needs of different users for content formats.

[0016] 2. Accurate personalized recommendations: Accurate user portrait labels and prompt words generated based on the large language model can help the system better understand user interests and improve the relevance and accuracy of recommended content.

[0017] 3. Closed-loop feedback mechanism: By collecting user feedback, the system can continuously optimize user portraits and recommendation strategies, achieve self-iteration of content recommendations, and improve user satisfaction and platform stickiness.

[0018] 4. Efficient content generation and delivery: By automating the content generation and delivery process, manual intervention and creation time are greatly reduced, and content production efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the application equally.

[0021] In the present invention, terms such as "upper", "lower", "left", "right", "front", "back", "vertical", "horizontal", "side", "bottom", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings. They are relational words determined only for the convenience of describing the structural relationships of the various parts or elements of the present invention, and do not specifically refer to any part or element in the present invention and should not be understood as limitations on the present invention.

[0022] Example: like Figure 1 As shown, this embodiment provides a method for automatically generating and pushing personalized content based on AIGC, including the following steps: S1: Obtain data information about the user's historical viewing records, browsing behaviors, and interaction records; S2: Analyze the data using a large language model based on the data information obtained in step S1 to generate user portrait labels and personalized prompt words; S3: Based on the user portrait tag and personalized prompt words generated in step S2, select and generate corresponding creation prompts; S4: Based on the creation prompt generated in step S3, the multimodal content is automatically generated through multiple generation models; S5: Personalize the push of automatically generated multimodal content and collect feedback.

[0023] In this embodiment, it is assumed that user A is a science fiction fan, and his historical behavior data includes watching movies such as "Interstellar" and "Inception", frequently clicking on science fiction videos, and searching for keywords such as "black hole theory" and "time travel". The system needs to generate personalized content for him and push it.

[0024] According to step 1, get the historical behavior data of user A: Data source: User A’s browsing history, search history, and interactive behaviors (clicks, favorites, and comments) on film and television platforms; Example data: Viewing history: 10 science fiction movies (80%), 2 documentaries (20%); Search keywords: "quantum physics" "interstellar travel"; Interactive behavior: Liked science fiction videos 15 times, and commented with keywords "hardcore science fiction" and "visual effects"; Based on the above data information, preprocess it: Anonymization processing: user ID is desensitized and behavior data is aggregated into feature vectors; Time window: extract the behavior sequences of the last 30 days and sort them by timestamp.

[0025] Step S2 is specifically: (1) Model configuration: Large language model: Qwen2.5-72B-Instruct; Attention mechanism: 4-head attention, hidden layer dimension 512, sequence length 256; (2) User portrait generation: Input: User A’s behavior sequence encoding (dimension 512); The output tags include: Interest tags: "Science fiction enthusiast (confidence 0.92)" "Hardcore science (0.85)" Emotional labels: “High immersion preference (0.88)” and “Complex narrative tendency (0.78)”; Attention weight: long-term interest (science fiction) weight 0.75, short-term interest (documentary) weight 0.25; (3) Multimodal prompt word generation: Reinforcement learning strategy: PPO algorithm, reward function weight (correlation), (diversity); Output prompt words include: Vincent's image: "Future city skyline and suspended spaceship, cyberpunk style"; Vincent: "Tense and suspenseful electronic music, BPM=120, timbre: futuristic"; Pictured video: "Nebula rotating dynamic special effects, 4K resolution, 60fps frame rate".

[0026] Step S3 is specifically: (1) Stable Diffusion Model: Input prompt words: "Future city skyline and floating spaceship"; Parameters: sampling temperature 0.8, generation resolution 1024×1024, iteration number 50; Output: 4 candidate images, screened by CLIP (image-text similarity ≥ 0.85); (2) Vincent model (Jukedeck improved version): Input prompt: "Tense and suspenseful electronic music"; Parameters: Track length 30 seconds, select "Sci-fi Synthesizer" as the sound library, and reverb intensity 0.6; Output: 3 candidate audios, with user profile matching degree ≥ 90%; (3) Stable Video Diffusion: Input image: Select the Wensheng image result; Parameters: video length 10 seconds, motion blur intensity 0.5, key frame interval 0.2 seconds; Output: dynamic nebula rotation video, PSNR=29dB (meets the cycle consistency requirement).

[0027] Step S4 is specifically: (1) Push strategy: Contextual slot machine algorithm: action space = {video, music, pictures and text}, exploration rate ; Real-time features: User A’s current context (tag “science fiction enthusiast”, online time “evening”); Push content: Prioritize video (weight 0.6) > music (0.3) > pictures and text (0.1); (2) Push example: Video: "Future City Special Effects Short Film"; Music: "Cyberpunk electronic remix"; Picture and text: "Popular Science Article on Black Hole Theory"; (3) Feedback collection: The user clicks on the video (CTR=1) and watches for 8 minutes (exceeding the threshold of 5 minutes); Failure to click on the image or text triggers a negative feedback signal; (4) Model update: Online learning: Update user portrait weight (science fiction label confidence + 0.05); Loss function: Based on the click-through rate (actual 1 vs predicted 0.9), adjust the push strategy parameters (learning rate 0.01).

[0028] Step S5 is specifically: (1) Cross-modal consistency verification: Video and music alignment: CLIP similarity 0.82 (threshold 0.8); Cycle consistency loss: PSNR = 28.5dB (reached the target); (2) Dynamic update of user portrait: Added new tag: "Special Effects Visual Preference (0.78)"; Eliminate the tag: "Documentary interest (confidence dropped to 0.15)" (3) System performance indicators: Generation efficiency: Single content generation takes 25 minutes (traditional method takes 4 hours); User satisfaction: NPS score +45 (baseline value +20).

[0029] like Figure 2 As shown, this embodiment also provides a push system based on the above-mentioned AIGC-based personalized content automatic generation and push method, including: Data collection module, used to obtain data information about users' historical viewing records, browsing behaviors and interaction records; The data processing module is used to: analyze the data through a large language model based on the acquired data information, generate user portrait tags and personalized prompt words, and select and generate corresponding creation prompts; The automatic generation module is used to: automatically generate multimodal content through multiple generation models based on the generated creation prompts; The personalized push and feedback module is used to: personalize the push of automatically generated multimodal content and collect feedback.

[0030] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. The method for automatically generating and pushing personalized content based on AIGC is characterized by: The following steps are involved: S1: Obtain data information about the user's historical viewing records, browsing behaviors, and interaction records; S2: According to the data information obtained in step S1, the data is analyzed by a large language model to generate user portrait labels and personalized prompt words; S3: Based on the user portrait tag and personalized prompt words generated in step S2, select and generate corresponding creation prompts; S4: Based on the creation prompt generated in step S3, the multimodal content is automatically generated through multiple generation models; S5: Personalize the push of automatically generated multimodal content and collect feedback.

2. The method for automatically generating and pushing personalized content based on AIGC according to claim 1, characterized in that: The data information of obtaining the user's historical viewing records includes but is not limited to: TV series, movies, videos, music, articles, and pictures watched by users; The data information of the user's browsing behavior obtained includes but is not limited to: User's search history and browsing history; The data information of obtaining the user's interaction records includes but is not limited to: User click, favorite, like, and comment behavior data.

3. The method for automatically generating and pushing personalized content based on AIGC according to claim 1, characterized in that: The step S2, generating a user portrait tag, is specifically: The Transformer-based multi-head attention mechanism is used to model user behavior sequences: ； in, is the query matrix, is the key-value matrix, is the vector dimension; By calculating the attention weights of different time steps in the behavior sequence, the long-term dependencies of user interests are dynamically extracted, and the final user portrait is expressed as: ； in, It is a three-layer fully connected network with GELU as the activation function, and outputs the probability distribution of user labels.

4. The method for automatically generating and pushing personalized content based on AIGC according to claim 2, characterized in that: The step S2 is to generate personalized prompt words, specifically: Use the mapping function from label to prompt word and combine it with reinforcement learning to optimize the generation quality: ； in, is the semantic similarity between the prompt word and the user tag, Score the diversity of the prompt words, , is the weight hyperparameter.

5. The method for automatically generating and pushing personalized content based on AIGC according to claim 1, characterized in that: In step S3, corresponding creation prompts are selected and generated, specifically: prompts of text-to-image, text-to-sound, and image-to-video are generated according to the media type, and the cross-modal joint loss function is used by jointly training the text-to-image, text-to-sound, and image-to-video models: ； in, The image-text alignment loss calculated for the CLIP model, is the cycle consistency loss, , is the balance coefficient.

6. The method for automatically generating and pushing personalized content based on AIGC according to claim 1, characterized in that: The personalized push is specifically as follows: Execute push strategy using contextual bandit algorithm with feedback: ； in, The type of content to be pushed. is the user context feature; The weight parameter for the type of content being pushed. is the noise term, and the parameters are updated through online learning , maximize user click-through rate and viewing time.

7. The method for automatically generating and pushing personalized content based on AIGC according to claim 6, characterized in that: The feedback collection is specifically: Collect user interaction feedback and behavior data in real time to further optimize user portraits and recommendation algorithms. The feedback-driven model incremental update mechanism is: The loss function of the feedback-driven model based on user feedback data is: ； in, For users to The actual feedback is: click is recorded as (1), and no click is recorded as (0); The click probability predicted by the model.

8. A push system based on the AIGC-based personalized content automatic generation and push method as claimed in claim 1, characterized in that: include: Data collection module, used to obtain data information about users' historical viewing records, browsing behaviors and interaction records; The data processing module is used to: analyze the data through a large language model based on the acquired data information, generate user portrait tags and personalized prompt words, and select and generate corresponding creation prompts; An automatic generation module, used to: automatically generate multimodal content through multiple generation models based on the generated creation prompts; The personalized push and feedback module is used to: personalize the push of automatically generated multimodal content and collect feedback.

Images