Intelligent advertisement material automatic optimization method, system and device and readable storage medium

Abstract

The invention relates to an intelligent advertisement material automatic optimization method, system and device and a readable storage medium, and the method comprises the steps: receiving the real-time putting effect number of a target advertisement video; analyzing the target advertisement video through a pre-trained multi-modal analysis model, and generating a commodity display sequence, copywriting content and background music features in the advertisement video; associating the delivery effect data to the analyzed commodity display sequence, copywriting content and background music features, and inputting the data into a dynamic optimization decision engine to generate a three-element optimization instruction; and executing an automatic modification operation on the target advertisement video according to the three-element optimization instruction, generating an optimized advertisement video and putting the optimized advertisement video so as to realize the purposes of real-time analysis, accurate optimization and efficient putting of the advertisement video content and improvement of the user attraction and conversion effect of the advertisement.

Description

Technical Field

[0001] The present invention relates to the field of advertising delivery technology, and in particular to a method, system, device and readable storage medium for automatic optimization of intelligent advertising materials. Background Art

[0002] The advertising landscape is undergoing a profound transformation, from static content design to dynamic strategy adjustments. Traditional ad optimization relies on manual experience and single-dimensional data, making it difficult to adapt to real-time changes in user behavior and complex scenarios. With the advancement of artificial intelligence and real-time data analysis capabilities, advertising systems are gradually incorporating key technologies such as product display logic, text semantic relevance, and audio emotion matching to precisely align advertising content with user needs. However, existing solutions still lack dynamic resource allocation, cross-factor collaborative optimization, and closed-loop feedback mechanisms, limiting the effectiveness of advertising. Summary of the Invention

[0003] The main purpose of the present invention is to provide a method, system, device and readable storage medium for automatic optimization of intelligent advertising materials, which can realize real-time analysis, precise optimization and efficient delivery of advertising video content through the collaboration of multimodal data analysis and dynamic optimization decision engine, so as to enhance the user appeal and conversion effect of advertising.

[0004] To achieve the above-mentioned purpose, the present invention provides a method for automatically optimizing intelligent advertising materials, comprising the following steps:

[0005] Receive real-time delivery effect data of the target advertising video, wherein the delivery effect data includes user behavior indicators and conversion rate indicators;

[0006] Analyzing the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video;

[0007] The delivery effect data is associated with the analyzed product display sequence, copy content, and background music features, and input into a dynamic optimization decision engine to generate three-element optimization instructions;

[0008] An automatic modification operation is performed on the target advertising video according to the three-element optimization instruction to generate an optimized advertising video and deliver it.

[0009] Furthermore, the step of receiving real-time delivery effect data of the target advertising video includes:

[0010] Obtaining real-time delivery effect data of the target advertising video, wherein the delivery effect data includes user behavior indicators and conversion rate indicators;

[0011] Among them, the user behavior indicators include the video playback completion rate, the distribution density of the product click heat map on the timeline, and the user jump-out time point. The conversion rate indicator is generated by tracking the complete behavior chain of the user jumping from the advertisement to the product details page.

[0012] Furthermore, the step of parsing the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video includes:

[0013] Inputting the target advertising video into a pre-trained multimodal analysis model;

[0014] Using the multimodal analysis model, the product display segments are segmented according to the video frame timestamps to generate a product display sequence with time sequence markers;

[0015] Identifying the dynamically superimposed text area in the target advertising video and extracting the copy content, including keywords and sentiment polarity values;

[0016] The background music track in the target advertisement video is separated, the beat frequency and emotion classification label are extracted, and background music features are generated.

[0017] Furthermore, the construction of the multimodal analysis model includes:

[0018] The multimodal analysis model consists of three sub-networks;

[0019] The visual temporal network detects product objects through a convolutional neural network and generates temporal tags based on inter-frame displacement correlation;

[0020] The text perception network uses optical character recognition to locate the superimposed text area and combines it with the semantic embedding model to output the sentiment polarity value;

[0021] The audio decoupling network uses spectral graph convolution to extract beat features and outputs emotion labels through a classifier.

[0022] Furthermore, the step of associating the delivery effect data with the parsed product display sequence, copy content, and background music features, and inputting the data into a dynamic optimization decision engine to generate three-element optimization instructions includes:

[0023] Aligning the user behavior indicators and conversion rate indicators in the delivery effect data with the product display sequence, copy content, and background music features on a timeline, and inputting them into a dynamic optimization decision engine;

[0024] Generate a priority index based on the slope change of the conversion rate indicator with the product display period, and output product order adjustment instructions;

[0025] Based on the matching relationship between the peak value of the click behavior heat map in the user behavior indicator and the time point of the copywriting appearance, the positioning coordinates are generated and the copywriting keyword replacement instructions are output;

[0026] Generate background music adaptation instructions based on the overlap between the dwell time distribution in the user behavior indicators and the background music emotion labels;

[0027] Summarize the product order adjustment instructions, copy keyword replacement instructions, and background music adaptation instructions to generate three-element optimization instructions.

[0028] Furthermore, the step of inputting the dynamic optimization decision engine to generate three-factor optimization instructions also includes:

[0029] During the optimized ad video delivery cycle, new user behavior indicators are monitored in real time;

[0030] Dynamically align new user behavior metrics with historical analysis elements, including product display sequence, copywriting content, and background music features, to generate incremental optimization instructions.

[0031] Merge incremental optimization instructions into three-element optimization instructions.

[0032] Furthermore, the step of automatically modifying the target advertising video according to the three-element optimization instruction, generating an optimized advertising video and delivering the optimized advertising video includes:

[0033] According to the product sequence adjustment instruction in the three-element optimization instruction, the video clip time sequence is reorganized according to the priority index;

[0034] Insert new copy at the positioning coordinates according to the copy keyword replacement instructions in the three-element optimization instructions;

[0035] According to the background music adaptation instruction in the three-element optimization instruction, a replacement segment with a matching beat frequency is matched from a preset music library to cover the original soundtrack;

[0036] Generate an optimized advertising video based on the modification results and execute delivery through the advertising delivery interface.

[0037] The present invention also provides an intelligent advertising material automatic optimization system, comprising:

[0038] The data acquisition unit receives real-time delivery effect data of the target advertising video;

[0039] A video analysis unit, configured to analyze the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video;

[0040] An instruction generation unit is used to associate the delivery effect data with the analyzed product display sequence, copy content, and background music features, and input them into a dynamic optimization decision engine to generate three-element optimization instructions;

[0041] The instruction execution unit is used to perform automatic modification operations on the target advertising video according to the three-element optimization instructions, generate an optimized advertising video and deliver it.

[0042] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned intelligent advertising material automatic optimization method when executing the computer program.

[0043] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above-mentioned intelligent advertising material automatic optimization method when executed by a processor.

[0044] The intelligent advertising material automatic optimization method, system, device and readable storage medium provided by the present invention have the following beneficial effects: the present invention realizes the dynamic reconstruction and precise delivery of advertising content through the synergistic effect of product display sequence analysis, copy semantic optimization and audio emotion adaptation. The present invention adjusts the product display order in real time based on user behavior data, dynamically optimizes copy keywords to enhance attractiveness, and matches background music emotions to extend user stay time. At the same time, it continuously tracks the delivery effect by using an incremental optimization mechanism, and continuously iterates the strategy through closed-loop feedback, breaking through the static mode of traditional advertising optimization, significantly improving the matching degree between advertising content and user needs, enhancing market adaptability, and providing an innovative solution for the dual improvement of advertising delivery efficiency and user conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 1 is a flow chart of a method for automatic optimization of intelligent advertising materials according to an embodiment of the present invention;

[0046] Figure 2 This is a structural block diagram of an intelligent advertising material automatic optimization system according to an embodiment of the present invention;

[0047] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0048] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0050] Reference Figure 1 , which is a flow chart of a method for automatically optimizing intelligent advertising materials proposed by the present invention, comprising the following steps:

[0051] S1, receiving real-time delivery effect data of a target advertising video, wherein the delivery effect data includes user behavior indicators and conversion rate indicators;

[0052] S2, parsing the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video;

[0053] S3, the delivery effect data is associated with the analyzed product display sequence, copy content and background music features, and input into the dynamic optimization decision engine to generate three-element optimization instructions.

[0054] S4, performing an automatic modification operation on the target advertising video according to the three-element optimization instruction, generating an optimized advertising video and delivering it.

[0055] In one embodiment, for step S1,

[0056] The steps of receiving real-time delivery effect data of a target advertising video include:

[0057] Obtaining real-time delivery effect data of the target advertising video, wherein the delivery effect data includes user behavior indicators and conversion rate indicators;

[0058] Among them, the user behavior indicators include the video playback completion rate, the distribution density of the product click heat map on the timeline, and the user jump-out time point. The conversion rate indicator is generated by tracking the complete behavior chain of the user jumping from the advertisement to the product details page.

[0059] In the specific implementation, the real-time collection of the target advertising video delivery effect data provides a dynamic decision-making basis for subsequent optimization. The process of this embodiment focuses on quantifying the correlation between user behavior and conversion effect, where the user behavior indicator captures the user interaction characteristics through multi-dimensional data: video playback completion rate (R completion ) is defined as the ratio of the number of users who have completed the viewing to the total number of playbacks, that is, R completion =(N end / N start )×100%, where N end N is the number of users who have completed the viewing; startThe number of users who start playing the ad directly reflects the attractiveness of the ad content; the time axis distribution density D of the product click heat map heatmap (t) = C(t) / T, where C(t) is the number of clicks within time interval t; T is the total duration of the ad. This density function can identify the peak period of user attention; the time point when the user jumps out (T exit ) By recording the timestamp of users terminating viewing early, combined with the bounce rate (R exit =(N exit / N start )×100%) to analyze the loss nodes of the advertisement. The conversion rate indicator (CR) is based on the behavior chain tracking technology and is defined as CR=(N conversion / N click )×100%, where N conversion N is the number of users who jumped from the ad to the product details page and completed the purchase / registration. click This is the total number of ad clicks. This indicator is tracked throughout the entire link through tracking technology and user path analysis.

[0060] In one embodiment, for step S2,

[0061] The step of parsing the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video includes:

[0062] Inputting the target advertising video into a pre-trained multimodal analysis model;

[0063] Using the multimodal analysis model, the product display segments are segmented according to the video frame timestamps to generate a product display sequence with time sequence markers;

[0064] Identifying the dynamically superimposed text area in the target advertising video and extracting the copy content, including keywords and sentiment polarity values;

[0065] The background music track in the target advertisement video is separated, the beat frequency and emotion classification label are extracted, and background music features are generated.

[0066] In the specific implementation, the target advertising video is structured and analyzed through a pre-trained multimodal analysis model, and the multi-dimensional information of the video (visual, text, audio) is converted into actionable optimization factors. The advertising video is used as input, and the visual timing network is used to analyze the video frame by frame: the spatiotemporal features of the product object are extracted through the convolutional neural network (CNN), and the inter-frame displacement association is combined (such as the optical flow method to calculate the motion vector of adjacent frames) to generate a product display sequence with time sequence tags. For example, if a product appears continuously in the 3rd to 5th seconds, the model will record its timestamp range (t start =3s,tend = 5s) and position coordinates (x, y, w, h), and based on the smoothness of the inter-frame displacement (such as mean square error where d i is the displacement of the i-th frame) to judge the coherence of the display. For dynamically superimposed text areas, the text perception network uses optical character recognition (OCR, such as the EAST algorithm) to locate the text box, and combines the semantic embedding model (such as BERT) to perform sentiment polarity analysis on the extracted keywords, and outputs the sentiment polarity value p∈[-1,1], where p>0 indicates positive emotions and p<0 indicates negative emotions. For example, if the words "limited time offer" appear in the advertisement, the model will set its sentiment polarity value to p=0.8 and mark it as a highly attractive keyword. In background music processing, the audio decoupling network generates a spectrum through short-time Fourier transform, and then uses a convolutional neural network to extract the beat frequency (f beat ) and emotional classification labels (such as "cheerful" and "nervous"). Ultimately, the model integrates product display sequences, copy content, and background music features into structured data, providing a multimodal benchmark for subsequent dynamic optimization decisions.

[0067] In one embodiment, the construction of the multimodal analysis model includes:

[0068] The multimodal analysis model consists of three sub-networks;

[0069] The visual temporal network detects product objects through a convolutional neural network and generates temporal tags based on inter-frame displacement correlation;

[0070] The text perception network uses optical character recognition to locate the superimposed text area and combines it with the semantic embedding model to output the sentiment polarity value;

[0071] The audio decoupling network uses spectral graph convolution to extract beat features and outputs emotion labels through a classifier.

[0072] Specifically, the construction of the multimodal analysis model adopts an architecture in which three sub-networks work together, corresponding to the feature extraction and fusion of visual, textual and audio modalities respectively. The core of the visual temporal network is the two-stream CNN (Two-Stream CNN), whose backbone network (such as ResNet-50) is responsible for extracting static product features, while the auxiliary network (such as LSTM) generates temporal tags through inter-frame displacement association to form a dynamic description of the product display sequence. For example, if a product in an advertisement is suddenly enlarged in the 5th second and accompanied by a highlight effect, the model will identify it as a key display node through the mutation of the inter-frame displacement (such as Δd>θ, where θ is a preset threshold) and mark it as a high-priority area. The text perception network combines OCR and semantic embedding models. First, the text area (such as x, y, w, h coordinates) is located through the EAST algorithm, and then the extracted text is input into the BERT model for sentiment analysis to calculate the sentiment polarity value. Where score(t i ) is the sentiment score of the i-th keyword (range [-1, 1]. For example, if high-frequency keywords such as "environmental protection" and "health" appear in the advertisement copy, the model will output p = 0.75 and mark it as a "positive sentiment" label. The audio decoupling network converts the original audio signal into a spectrogram through STFT, and then uses ResNet-18 to extract the beat frequency (f beat ) and emotion classification labels (such as "motivational" and "sad"). By mapping the three types of features to a unified semantic space (such as using contrastive learning to align visual-text-audio features), the model can achieve cross-modal associations. For example, it can identify the synergy between the "motivational" emotions of background music and the high-frequency actions of product displays (such as rapid camera switching), thereby providing more accurate decision-making basis for subsequent optimization. This divide-and-conquer sub-network design not only ensures the independence of each modal feature, but also realizes the fusion of global information through cross-modal alignment, significantly improving the accuracy and robustness of advertising analysis.

[0073] In one embodiment, for step S3,

[0074] The step of associating the delivery effect data with the parsed product display sequence, copy content, and background music features, and inputting the data into a dynamic optimization decision engine to generate three-element optimization instructions includes:

[0075] Aligning the user behavior indicators and conversion rate indicators in the delivery effect data with the product display sequence, copy content, and background music features on a timeline, and inputting them into a dynamic optimization decision engine;

[0076] Generate a priority index based on the slope change of the conversion rate indicator with the product display period, and output product order adjustment instructions;

[0077] Based on the matching relationship between the peak value of the click behavior heat map in the user behavior indicator and the time point of the copywriting appearance, the positioning coordinates are generated and the copywriting keyword replacement instructions are output;

[0078] Generate background music adaptation instructions based on the overlap between the dwell time distribution in the user behavior indicators and the background music emotion labels;

[0079] Summarize the product order adjustment instructions, copy keyword replacement instructions, and background music adaptation instructions to generate three-element optimization instructions.

[0080] In the specific implementation, the user behavior indicators (such as heat map density D heatmap (t)) and the timestamp t of the product display sequence item , the time when the copy appears t text , Background Music Emotion Labelaudio For example, if a product is displayed in the 4th to 6th seconds, the heat map peak D peak The time when the keyword "limited time offer" appears in the copywriting text =5s high overlap (matching degree Where δ = 1s is the time tolerance), the copywriting keyword replacement instruction is triggered, replacing "limited time offer" with a more urgent "last 24 hours". For product order adjustment, the system calculates the time derivative of the conversion rate CR(t) Generate priority index where w i is the product weight (e.g., high-priced product weight w = 1.5), thereby dynamically reorganizing the video clip sequence. In background music adaptation, calculate the user stay time distribution T stay (t) and background music emotion label E audio Correlation coefficient When ρ>0.7, the original music clip is retained, otherwise the beat frequency f is matched from the preset music library. beat Similar replacement segments (such as replacing 120bpm fast-paced music with a 90bpm soothing melody). Ultimately, the three-factor optimization instructions are sent to the modification module in real time through a distributed scheduling framework (such as Kafka Streams), ensuring that the optimization strategy takes effect quickly.

[0081] In one embodiment, the step of inputting the dynamic optimization decision engine to generate the three-factor optimization instructions further includes:

[0082] During the optimized ad video delivery cycle, new user behavior indicators are monitored in real time;

[0083] Dynamically align new user behavior metrics with historical analysis elements, including product display sequence, copywriting content, and background music features, to generate incremental optimization instructions.

[0084] Merge incremental optimization instructions into three-element optimization instructions.

[0085] Specifically, the incremental optimization mechanism of the dynamic optimization decision engine achieves continuous iteration through real-time data stream processing. During the delivery cycle of the optimized advertising video, new user behavior indicators (such as the newly added exit time point T' exit , Add click heat map D' heatmap ), and uses a sliding window algorithm (such as a rolling window with a length of 5 minutes) and historical analysis elements (product display sequence S item 、Copy content C text , Background music characteristics F audio) for dynamic alignment. For example, if the newly added heat map data shows that the click density in the 8th to 10th seconds drops by 30%, and the copywriting keyword in the historical record for this period is "Buy Now", the system will trigger an incremental optimization instruction and replace the keyword with "Click to Receive". The generation of incremental optimization instructions relies on online learning algorithms (such as FTRL optimizer), and its update formula is Where θ is the optimization parameter, gt is the gradient, G t is the cumulative squared value of historical gradients, and η is the learning rate. Ultimately, the incremental instructions are combined with the original three-factor optimization instructions through a priority queue (e.g., weighted sorting based on conversion rate improvement), forming a closed-loop optimization strategy that ensures the continuous adaptability of advertising creatives in a dynamic market environment.

[0086] In one embodiment, for step S4,

[0087] The step of automatically modifying the target advertising video according to the three-element optimization instruction, generating an optimized advertising video, and delivering the optimized advertising video comprises:

[0088] According to the product sequence adjustment instruction in the three-element optimization instruction, the video clip time sequence is reorganized according to the priority index;

[0089] Insert new copy at the positioning coordinates according to the copy keyword replacement instructions in the three-element optimization instructions;

[0090] According to the background music adaptation instruction in the three-element optimization instruction, a replacement segment with a matching beat frequency is matched from a preset music library to cover the original soundtrack;

[0091] Generate an optimized advertising video based on the modification results and execute delivery through the advertising delivery interface.

[0092] In the specific implementation, by performing automated modification operations on the target advertising video, an efficient, low-latency, high-fidelity advertising material generation pipeline is built to achieve a closed-loop conversion from optimization instructions to actual delivery. i Reorganize the original video clips and use a timestamp-based nonlinear editing algorithm (such as FFmpeg's concat demuxer) to splice the product display clips with time stamps in the original video into a new order to ensure a natural transition without screen tearing. For example, if the conversion rate slope of a product A is 0. If the value of the first video is higher than 0.05 of the product B, its priority will be raised to play before the product B. Then, according to the text keyword replacement instruction, the new text is inserted at the location coordinates (x, y, w, h) identified by OCR, and the image restoration technology (such as OpenCV's inpainting) is used to erase the old text background. Then, the new text is synthesized through font matching and rendering engine (such as FreeType), and the font size, color and animation effect are kept consistent with the original video style. For the background music adaptation instruction, the system retrieves the beat frequency f from the preset music library. beat The closest alternative clip is found (with an error range of ±2BPM), and audio mixing algorithms (such as crossfading) are used to seamlessly cover the original soundtrack to avoid abrupt switching. Finally, an ad delivery interface (such as the MetaAds API or the Ocean Engine API) is called to push the generated optimized ad video to the target platform and accelerate its delivery to the user's terminal device through the CDN.

[0093] Reference Figure 2 , is a structural block diagram of an intelligent advertising material automatic optimization system according to an embodiment of the present invention, comprising:

[0094] The data acquisition unit receives real-time delivery effect data of the target advertising video;

[0095] A video analysis unit, configured to analyze the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video;

[0096] An instruction generation unit is used to associate the delivery effect data with the analyzed product display sequence, copy content, and background music features, and input them into a dynamic optimization decision engine to generate three-element optimization instructions;

[0097] The instruction execution unit is used to perform automatic modification operations on the target advertising video according to the three-element optimization instructions, generate an optimized advertising video and deliver it.

[0098] For the specific implementation of each unit in the above device example, please refer to the above method embodiment, which will not be repeated here.

[0099] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0100] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0101] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0102] In summary, the present invention receives the real-time delivery effect data of the target advertising video; analyzes the target advertising video through a pre-trained multimodal analysis model to generate the product display sequence, copy content and background music features in the advertising video; associates the delivery effect data with the analyzed product display sequence, copy content and background music features, and inputs them into a dynamic optimization decision engine to generate a three-element optimization instruction; automatically modifies the target advertising video according to the three-element optimization instruction, generates an optimized advertising video and delivers it, so as to achieve real-time analysis, precise optimization and efficient delivery of advertising video content, and enhance the user appeal and conversion effect of the advertisement.

[0103] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Among them, any reference to memory, storage, database, or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAM bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0104] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0105] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for automatic optimization of intelligent advertising materials, characterized in that: The following steps are involved: Receive real-time delivery effect data of the target advertising video, wherein the delivery effect data includes user behavior indicators and conversion rate indicators; Analyzing the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video; The delivery effect data is associated with the analyzed product display sequence, copy content, and background music features, and input into a dynamic optimization decision engine to generate three-element optimization instructions; An automatic modification operation is performed on the target advertising video according to the three-element optimization instruction to generate an optimized advertising video and deliver it.

2. The method for automatic optimization of intelligent advertising materials according to claim 1, characterized in that: The step of receiving real-time delivery effect data of the target advertising video includes: Obtaining real-time delivery effect data of the target advertising video, wherein the delivery effect data includes user behavior indicators and conversion rate indicators; Among them, the user behavior indicators include the video playback completion rate, the distribution density of the product click heat map on the timeline, and the user jump-out time point. The conversion rate indicator is generated by tracking the complete behavior chain of the user jumping from the advertisement to the product details page.

3. The method for automatic optimization of intelligent advertising materials according to claim 1, characterized in that: The step of parsing the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video includes: Inputting the target advertising video into a pre-trained multimodal analysis model; Using the multimodal analysis model, the product display segments are segmented according to the video frame timestamps to generate a product display sequence with time sequence markers; Identifying the dynamically superimposed text area in the target advertising video and extracting the copy content, including keywords and sentiment polarity values; The background music track in the target advertisement video is separated, the beat frequency and emotion classification label are extracted, and background music features are generated.

4. The method for automatic optimization of intelligent advertising materials according to claim 3, characterized in that: The construction of the multimodal analysis model includes: The multimodal analysis model consists of three sub-networks; The visual temporal network detects product objects through a convolutional neural network and generates temporal tags based on inter-frame displacement correlation; The text perception network uses optical character recognition to locate the superimposed text area and combines it with the semantic embedding model to output the sentiment polarity value; The audio decoupling network uses spectral graph convolution to extract beat features and outputs emotion labels through a classifier.

5. The method for automatic optimization of intelligent advertising materials according to claim 1, characterized in that: The step of associating the delivery effect data with the parsed product display sequence, copy content, and background music features, and inputting the data into a dynamic optimization decision engine to generate three-element optimization instructions includes: Aligning the user behavior indicators and conversion rate indicators in the delivery effect data with the product display sequence, copy content, and background music features on a timeline, and inputting them into a dynamic optimization decision engine; Generate a priority index based on the slope change of the conversion rate indicator with the product display period, and output product order adjustment instructions; Based on the matching relationship between the peak value of the click behavior heat map in the user behavior indicator and the time point of the copywriting appearance, the positioning coordinates are generated and the copywriting keyword replacement instructions are output; Generate background music adaptation instructions based on the overlap between the dwell time distribution in the user behavior indicators and the background music emotion labels; Summarize the product order adjustment instructions, copy keyword replacement instructions, and background music adaptation instructions to generate three-element optimization instructions.

6. The method for automatic optimization of intelligent advertising materials according to claim 1, characterized in that: The step of inputting the dynamic optimization decision engine to generate three-element optimization instructions also includes: During the optimized ad video delivery cycle, new user behavior indicators are monitored in real time; Dynamically align new user behavior metrics with historical analysis elements, including product display sequence, copywriting content, and background music features, to generate incremental optimization instructions. Merge incremental optimization instructions into three-element optimization instructions.

7. The method for automatic optimization of intelligent advertising materials according to claim 1, characterized in that: The step of automatically modifying the target advertising video according to the three-element optimization instruction, generating an optimized advertising video, and delivering the optimized advertising video includes: According to the product sequence adjustment instruction in the three-element optimization instruction, the video clip time sequence is reorganized according to the priority index; Insert new copy at the positioning coordinates according to the copy keyword replacement instructions in the three-element optimization instructions; According to the background music adaptation instruction in the three-element optimization instruction, a replacement segment with a matching beat frequency is matched from a preset music library to cover the original soundtrack; Generate an optimized advertising video based on the modification results and execute delivery through the advertising delivery interface.

8. An intelligent advertising material automatic optimization system, characterized in that: include: The data acquisition unit receives real-time delivery effect data of the target advertising video; A video analysis unit, configured to analyze the target advertising video using a pre-trained multimodal analysis model to generate product display sequences, text content, and background music features in the advertising video; An instruction generation unit is used to associate the delivery effect data with the analyzed product display sequence, copy content, and background music features, and input them into a dynamic optimization decision engine to generate three-element optimization instructions; The instruction execution unit is used to perform automatic modification operations on the target advertising video according to the three-element optimization instructions, generate an optimized advertising video and deliver it.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the intelligent advertising material automatic optimization method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for automatic optimization of intelligent advertising materials according to any one of claims 1 to 7 are implemented.

Images