An artificial intelligence advertisement generation method and device based on user portraits, an electronic device, and a storage medium

Abstract

The embodiment of the application discloses a kind of artificial intelligence advertisement generation method, device, electronic equipment and storage medium based on user portrait, it is related to artificial intelligence advertisement generation technical field, method includes: obtaining user behavior information and device state information, user behavior information includes historical behavior data and real-time behavior data;Device type and network state are determined according to device state information;According to user behavior information, determine the attention state information of user;According to device type, network state and attention state information, generate multi-dimensional user portrait;According to user portrait, determine the length type of advertisement, and the length type of advertisement includes: ultra-short type, shorter type and complete type;Using advertisement editing model, generate advertisement according to the length type of advertisement and user portrait.The application solves the problem of insufficient flexibility and low accuracy in traditional advertisement generation method.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence advertising generation technology, and in particular to an artificial intelligence advertising generation method and apparatus based on user profiles. Background Technology

[0002] With the widespread adoption of the internet and mobile devices, digital advertising has become a crucial tool for brand promotion and user attraction. Traditional ad generation methods often rely on static rules or simple user classifications, making it difficult to adapt to dynamic changes in user behavior and personalized needs. For example, in existing technologies, ad length and content are often fixed, unable to be flexibly adjusted based on user attention levels, network environment, or device type, resulting in poor user experience or low ad effectiveness.

[0003] In recent years, advancements in artificial intelligence (AI) technology have brought new opportunities for ad generation. By analyzing user behavior data and device status, user profiles can be built and ad delivery optimized. However, existing methods still have the following shortcomings: First, the lack of dynamic assessment of user attention states leads to a mismatch between ad rhythm and user needs; second, ad length selection is too simplistic, failing to adjust based on user tolerance and real-time feedback; and third, ad creative editing lacks intelligence, making it difficult to accurately adapt to different scenarios. Therefore, a comprehensive and intelligent ad generation method is urgently needed to improve ad accuracy and user acceptance. Summary of the Invention

[0004] This application provides an AI-based advertising generation method based on user profiles, which enables accurate advertising generation.

[0005] This application provides the following solution:

[0006] According to a first aspect, an AI-powered advertising generation method based on user profiles is provided, characterized in that the method includes: acquiring user behavior information and device status information, wherein the user behavior information includes historical behavior data and real-time behavior data; determining device type and network status based on the device status information; determining user attention status information based on the user behavior information; generating a multi-dimensional user profile based on the device type, network status, and attention status information; determining an advertising length type based on the user profile, wherein the advertising length type includes: ultra-short, short, and full; and generating an advertising using an advertising clipping model based on the advertising length type and the user profile.

[0007] According to one achievable method in this application embodiment, the real-time behavior data includes: current page dwell time and interaction frequency; determining the user's attention state information based on the user behavior information includes: judging the attention state as high attention, low attention, or distraction state based on the current page dwell time and interaction frequency; and correcting the attention state determination result by comparing the current behavior with the user's historical behavior patterns in conjunction with the historical behavior data.

[0008] According to one achievable method in this application embodiment, determining the ad length type based on the user profile includes: using a pre-trained ad length determination model to determine the ad length type based on the user profile; wherein, the pre-training includes: acquiring a training dataset, the training dataset including: multi-dimensional user profile samples and their corresponding ad length type labels and ad viewing behavior data; training the ad length determination model using a supervised learning method, with the user profile as input; the ad length determination model outputting the probability distribution of each length type, and determining the ad length type according to a preset threshold or the maximum probability; during the training process, a time dimension feature is added, the time dimension feature including: the current time period and usage scenario information; the training objective includes: minimizing the difference between the ad length type and the ad length type label.

[0009] According to one achievable method in an embodiment of this application, determining the ad length type based on the user profile includes: for users whose attention state information indicates a low attention state or low network speed, determining the ad length type as ultra-short; for users whose attention state information indicates a medium attention state or medium network speed, determining the ad length type as relatively short; for users whose attention state information indicates a high attention state or high network speed, determining the ad length type as full; determining the user's historical ad tolerance, real-time feedback, and / or the content currently viewed by the user based on the user behavior data; and dynamically adjusting the ad length type based on the user's historical ad tolerance, real-time feedback, and / or the content currently viewed by the user.

[0010] According to one achievable method in this application embodiment, dynamically adjusting the ad length type based on the user's historical ad tolerance and the real-time feedback includes: collecting the user's past ad viewing records, the ad viewing records including: the complete viewing duration of each ad, the skip time point, and / or the skip rate; calculating the user's ad tolerance threshold, wherein the tolerance threshold is determined by statistically analyzing the user's average viewing duration or the critical duration for skipping ads; classifying users into low tolerance, medium tolerance, and high tolerance types based on the tolerance threshold; matching the corresponding ad length according to the user classification to obtain a first length value; during ad playback, monitoring the user's interactive behavior in real time to obtain real-time feedback, the interactive behavior including: skipping ads, performing a mute operation, fast-forwarding, and / or switching applications; if the real-time feedback shows that the user's skip rate increases or tolerance decreases, then shortening the first length value; if the real-time feedback shows that the user watches the ad completely and has a high frequency of interactive behavior, then extending the first length value; and determining the ad length type based on the first length value.

[0011] According to one achievable method in this application embodiment, the historical behavior data includes video clips, image styles, and text content reflecting the user's historical preferences; the step of generating an advertisement using an advertising editing model based on the advertisement length type and the user profile includes: selecting materials matching the user profile from an advertising material library based on the historical behavior data; extracting key images from the materials using computer vision technology and adjusting the editing rhythm and visual effects in conjunction with the user's attention state, wherein the editing rhythm includes: the frequency of image switching and information density; optimizing the advertising copy and voice-over using natural language processing technology; and generating personalized advertising content that conforms to the user's preferences through generative adversarial networks or text-to-video technology.

[0012] According to one achievable method in an embodiment of this application, generating an advertisement based on the advertisement length type and the user profile using an advertisement editing model includes: obtaining weather condition information; classifying the weather conditions based on the weather condition information to obtain weather types, wherein the weather types include: suitable for going out and unsuitable for going out; the advertisement editing model modifies the advertisement length type according to the weather type, and generates an advertisement based on the advertisement length type and the user profile.

[0013] According to a second aspect, an AI-powered advertising generation device based on user profiles is provided, characterized in that the device comprises: an information acquisition unit configured to acquire user behavior information and device status information, wherein the user behavior information includes historical behavior data and real-time behavior data; an information analysis unit configured to determine device type and network status based on the device status information; an attention status determination unit configured to determine user attention status information based on the user behavior information; a user profile generation unit configured to generate a multi-dimensional user profile based on the device type, network status, and attention status information; a length type determination unit configured to determine the advertising length type based on the user profile, wherein the advertising length type includes: ultra-short, short, and full; and an advertising generation unit configured to generate an advertising based on the advertising length type and the user profile using an advertising clipping model.

[0014] According to a third aspect, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0015] According to a fourth aspect, an electronic device is provided, comprising: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any one of the first aspects.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0017] This application provides an AI-based advertising generation method based on user behavior. By acquiring user behavior information and device status information, and combining attention status, device type, and network status, a multi-dimensional user profile is generated. Personalized ads are then dynamically generated using an ad clipping model, thereby solving the problems of insufficient flexibility and low accuracy in traditional ad generation methods.

[0018] Of course, any product implementing this application does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0020] Figure 1This is a system architecture diagram applicable to the embodiments of this application;

[0021] Figure 2 This is a flowchart of the AI-based ad generation method based on user profiles provided in the embodiments of this application;

[0022] Figure 3 A structural block diagram of an AI-powered advertising generation device based on user profiles provided in an embodiment of this application;

[0023] Figure 4 A schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0025] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0026] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0027] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0028] Existing ad generation methods mostly rely on static rules or simple user classification, making it difficult to adapt to dynamic changes in user behavior and personalized needs. Therefore, this application provides a new approach. To facilitate understanding of this application, the system architecture on which it is based is first described. Figure 1 An exemplary system architecture that can be applied to embodiments of this application is shown, such as Figure 1As shown, the system architecture may include: user devices and an AI-powered ad generation device based on user profiles located on the server side.

[0029] User devices can include, but are not limited to, smart mobile terminals, smart home devices, wearable devices, and PCs (Personal Computers). Smart mobile devices can include mobile phones, tablets, laptops, PDAs (Personal Digital Assistants), and connected cars. Smart home devices can include smart TVs, smart refrigerators, and so on. Wearable devices can include smartwatches, smart glasses, virtual reality devices, augmented reality devices, and mixed reality devices (i.e., devices that support both virtual and augmented reality).

[0030] AI-powered ad generation devices based on user profiles can be configured as standalone servers, within server clusters, or on cloud servers. Cloud servers, also known as cloud computing servers or cloud hosts, are a hosting product within the cloud computing service ecosystem, designed to address the management difficulties and weak service scalability inherent in traditional physical hosts and Virtual Private Servers (VPS) services. Besides... Figure 1 In addition to the architecture shown, the AI-powered ad generation device based on user profiles can also be set up on a computer terminal with strong computing power.

[0031] As one feasible approach, an AI-powered ad generation device generates ads based on user behavior and device status information, returns the ads to the user's device via the network, and then the user's device displays the ads to the user. It should be understood that... Figure 1 The user devices and AI-powered ad generation devices shown are merely illustrative. Depending on the implementation requirements, any number of user devices and AI-powered ad generation devices can be included.

[0032] Figure 2 This is a flowchart of an AI-powered ad generation method based on user profiles provided in an embodiment of this application. The method can be... Figure 1 The AI-powered ad generation device based on user profiles in the system shown is executed. For example... Figure 2 As shown, the method may include the following steps:

[0033] Step 201: Obtain user behavior information and device status information. User behavior information includes historical behavior data and real-time behavior data.

[0034] Step 202: Determine the device type and network status based on the device status information.

[0035] Step 204: Generate a multi-dimensional user profile based on device type, network status, and attention status information.

[0036] Step 205: Determine the ad length type based on the user profile. Ad length types include: ultra-short, short, and full.

[0037] Step 206: Use the ad clipping model to generate ads based on ad length type and user profile.

[0038] As can be seen from the above process, this application provides an AI-based advertising generation method based on user behavior. By acquiring user behavior information and device status information, and combining attention status, device type and network status, a multi-dimensional user profile is generated. Personalized advertisements are then dynamically generated using an ad clipping model, thereby solving the problems of insufficient flexibility and low accuracy in traditional advertising generation methods.

[0039] First, the above step 201, namely "obtaining user behavior information and device status information, wherein the user behavior information includes historical behavior data and real-time behavior data", will be described in detail with reference to the embodiments.

[0040] This application aims to provide foundational support for subsequent user profiling and ad generation through multi-dimensional data collection. Specifically, "user behavior information" refers to various interaction data generated by users when using devices or applications, divided into two subsets: "historical behavior data" and "real-time behavior data." "Historical behavior data" reflects user behavior patterns over a past period, such as the total duration of ad viewing in the past 30 days (e.g., an average of 15 seconds per viewing), click-through rate (e.g., 5 ad clicks per day), ad skipping frequency (e.g., 70% of ads skipped within 5 seconds), and preferred content types (e.g., a preference for watching sports videos). This data is typically obtained through long-term accumulation via client logs or cloud databases and is used to analyze users' long-term habits and tolerance. "Real-time behavior data" captures users' current instantaneous behavior, such as the time spent on the current page (e.g., 3 seconds on a news page), scrolling speed (e.g., 2 cm per second), interaction frequency (e.g., 2 clicks per minute), and usage scenario (e.g., inferred from time and location as a morning commute). Meanwhile, "device status information" includes the device's hardware and network status, such as device model, operating system version, network type (Wi-Fi or 4G), and bandwidth speed (20Mbps).

[0041] By acquiring this information, the system can gain a comprehensive understanding of users' historical preferences, current status, and device environment, providing accurate data for subsequent attention state assessment and ad editing.

[0042] The following describes step 202, namely "determining the device type and network status based on device status information," in detail with reference to an embodiment.

[0043] "Device status information" refers to the technical parameters and operational data collected from the user's device, while "device type" and "network status" are two core dimensions extracted from this information. Determining the "device type" involves the device's hardware characteristics, such as model, screen size, and operating system version. This information is typically obtained through device APIs or client probes to assess device performance and usage scenarios. For example, a high-end smartphone may support high-definition video playback, while a low-end device may be more suitable for lightweight content. "Network status" reflects the device's current connection quality, including network type (e.g., Wi-Fi, 4G, 5G), real-time bandwidth (e.g., 20Mbps or 1Mbps), latency (e.g., 50ms), and packet loss rate (e.g., 2%). This data can be collected in real-time using network monitoring tools to evaluate the feasibility of ad loading and user experience. For example, suppose user B uses a mid-range Android phone; the device status information displays its model, screen size, and operating system; network monitoring indicates it uses a 4G network with a bandwidth of 5Mbps and a latency of 80ms. Based on this data, the "device type" can be determined as a "mid-range smartphone," and the "network status" as a "medium-speed network."

[0044] The following describes step 203, namely "determining the user's attention state information based on user behavior information," in detail with reference to an embodiment.

[0045] This invention analyzes user behavior data to infer their current level of focus, thereby providing a basis for the length and editing rhythm of advertisements. Attention state represents the user's level of focus and can be classified as "high attention," "low attention," and "distracted state," or it can be given in the form of a rating.

[0046] Attention state is determined based on user behavior information. This can be based solely on real-time behavioral data, or a combination of historical and real-time data analysis, further optimized by intelligent algorithms. For example, the system might use a pre-trained machine learning model, taking behavioral data as input and outputting an attention state classification. Training data comes from labeled samples (such as user feedback or experimental observations). The model corrects its initial judgment by comparing current behavior with historical patterns. Taking user E as an example, historical data shows a tendency towards high attention (frequently watching complete ads), but real-time data indicates a current dwell time of only 3 seconds and frequent swiping. After comprehensive analysis, the model might adjust the classification to "distracted state." Furthermore, if the device supports sensor data (such as camera-detected eye movements or heart rate changes), the results can be further validated; for example, focused eye movements indicate focus, while scattered eye movements indicate distraction.

[0047] For example, suppose user A watches videos on a tablet at home. Historical data shows they watch ads for an average of 20 seconds with a low skip rate; real-time data shows a 15-second dwell time on the current page, an interaction frequency of 4 times per minute, and a usage scenario of "evening leisure." Based on the real-time data, the user is initially classified as "high attention state," and this is confirmed by historical tolerance data, ultimately outputting "high attention." Conversely, if user B uses their phone at work, with a real-time dwell time of only 2 seconds, frequent app switching, and a historical skip rate as high as 80%, the system classifies them as "distracted state." This method, through multi-dimensional data analysis, ensures the accuracy and practicality of attention state determination, laying the foundation for subsequent ad generation.

[0048] As an implementable approach, real-time behavioral data includes the current page dwell time and interaction frequency; determining the user's attention state information based on the user behavioral information includes: judging the attention state as high attention, low attention, or distraction based on the current page dwell time and interaction frequency; optimizing the attention state judgment result by combining historical behavioral data, and correcting the attention state by comparing the current behavior with the user's historical behavior patterns.

[0049] Specifically, this embodiment determines the attention state based on the current page dwell time and interaction frequency. "Current page dwell time" refers to the length of time a user stays on a certain page or application interface, usually measured in seconds. For example, a user stays on a news page for 5 seconds or on a short video page for 2 seconds. "Interaction frequency" reflects the number of times a user interacts with the interface per unit of time, such as 3 clicks or 2 swipes per minute. These indicators are directly related to the user's level of focus.

[0050] Based on the aforementioned real-time data, a preliminary judgment rule is provided for "determining whether the attention state is high attention, low attention, or distraction based on the current page dwell time and the interaction frequency." Specifically, a long dwell time and high interaction frequency usually correspond to a "high attention state," indicating that the user is willing to invest time and effort; a short dwell time and low interaction frequency indicate a "low attention state," showing that the user has limited interest; if the interaction frequency is irregular (such as fluctuating) or accompanied by rapid switching behavior, it may be a "distracted state," meaning that the user's attention is scattered. Taking user C as an example, their current page dwell time is 15 seconds, and their interaction frequency is 4 times per minute, which the system initially judges as a "high attention state"; while user D's dwell time is only 2 seconds, with no interaction, and is judged as a "low attention state"; user E's dwell time is 3 seconds, but the interaction frequency fluctuates (quickly swiping after clicking), and is judged as a "distracted state."

[0051] However, relying solely on real-time data can be inaccurate due to the variability of user behavior. Therefore, the approach of "combining historical behavioral data to optimize attention state judgment results and correcting attention state by comparing current behavior with historical user behavior patterns" introduces a historical dimension to improve reliability. Historical behavioral data can include past ad viewing duration, skip rate, and interaction habits. For example, user F's average ad viewing time over the past 30 days is 20 seconds, with a skip rate of only 20%, indicating high tolerance and attentional tendency. The correction process adjusts by comparing current behavior with historical patterns. For instance, user F's current dwell time is only 3 seconds, with low interaction frequency, initially judged as "low attention state." However, their historical data shows a high attentional tendency, so the system may consider the current behavior a temporary anomaly and correct it to "moderate attention state." This method uses long-term patterns to correct for instantaneous fluctuations and avoids misjudgment.

[0052] This step can be implemented in several ways. For example, based on predefined rules, statistical indicators of current behavior data can be compared with those of historical behavior data to adjust the attention state. Specifically, key statistical indicators of historical behavior, such as average page dwell time, average interaction frequency, ad completion rate, and skip rate, can be calculated. Current behavior data can be compared with historical indicators, and thresholds can be set. The attention state can be adjusted based on the magnitude of the deviation. For example, if the current dwell time is much lower than the historical average, but the history shows a high attention tendency, "low attention" might be corrected to "medium attention." Alternatively, a supervised learning model can be trained, taking historical and current behavior data as input and outputting the corrected attention state. Specifically, a labeled dataset can be collected, containing historical behavior samples (such as dwell time and interaction frequency over the past 30 days), current behavior samples, and their corresponding attention state labels (such as labels from user feedback or experiments). A classification model, such as a support vector machine, random forest, or deep neural network, can be used, with input features including historical and current dwell time and interaction frequency, and the output being a "high / low / distracted" state. The model can be trained, optimizing the cross-entropy loss function to enable the model to learn the comparative relationship between historical and current behavior.

[0053] Furthermore, this application can also determine a user's attention state by combining page jump time. Jump time, as part of user behavior information, can indirectly reflect the user's attention level. Generally speaking, a shorter jump time may indicate that the user is quickly browsing content with low attention or is in a distracted state; while a longer jump time may mean that the user is staying on a page and reading or operating in depth, with higher attention. The specific determination needs to consider the statistical distribution of jump time and the context.

[0054] In implementation, attention status can be determined based on jump time through the following steps: First, collect the user's real-time jump time sequence, for example, by recording the timestamp of each page switch in the application or website through front-end tracking, and calculating the interval between adjacent jumps. Second, set rules for judging attention status. For example, if the average jump time is less than a certain threshold (e.g., 5 seconds), it can be initially judged as a "low attention state," because rapid jumps suggest that the user is not deeply engaged with the content; if the jump time is in a medium range (e.g., 5-15 seconds), it may be a "medium attention state"; if the jump time is long (e.g., more than 15 seconds), it tends to be a "high attention state." In addition, the volatility of jump time also needs to be considered: if the jump time interval is irregular (e.g., 2 seconds, 10 seconds, 3 seconds), it may indicate a "distraction state," indicating that the user's attention is not focused. These rules can be adjusted according to the actual application scenario.

[0055] Combining historical behavioral data can further refine the judgment results. Historical behavioral data may contain past user jump time patterns. For example, a user's average jump time is 20 seconds, and the jumps are relatively stable (low standard deviation), indicating that they are accustomed to browsing pages for a longer period of time. If the current jump time deviates significantly from the historical pattern (e.g., suddenly drops to 3 seconds), it may be due to temporary distraction or environmental changes. The system can make a preliminary judgment by comparison. For example, user A's historical data shows an average jump time of 25 seconds, and the current jump time is 4 seconds, initially judged as "low attention state." However, considering their historical tendency, the system may correct it to "moderate attention state." This comparison method can be achieved through statistical analysis (such as calculating the percentage of deviation) or machine learning models (such as classifiers).

[0056] The following describes step 204, namely "generating a multi-dimensional user profile based on the device type, network status, and attention status information," in detail with reference to an embodiment.

[0057] This invention constructs a comprehensive user profile model by integrating device-related hardware characteristics, network environment, and user behavior. This multi-dimensional user profile, through the fusion of multi-source data, reflects the user's overall usage habits, environmental context, and attentional tendencies, providing a precise basis for subsequent ad length selection and content generation. The following will explain this feature from a technical implementation perspective and illustrate it with specific examples.

[0058] The process of generating multi-dimensional user profiles involves integrating the three dimensions (device type, network status, and attention status) into a comprehensive feature vector or label set. Specifically, this can be achieved through data concatenation or feature fusion techniques. For example, the system quantifies device type, network status, and attention status into feature values ​​(e.g., device type: high-end = 1, mid-range = 0.5, low-end = 0; network status: high-speed = 1, mid-speed = 0.5, low-speed = 0; attention status: high = 1, low = 0.5, distracted = 0), and then concatenates them into a three-dimensional vector (e.g., [1, 1, 1] represents "high-end device, high-speed network, high attention"). Alternatively, machine learning methods (such as clustering or embedding models) can be used to map these dimensions into a higher-level profile label. For example, K-means clustering can categorize users into groups such as "focused high-end users" and "fast-browsing mid-range users."

[0059] Furthermore, the multi-dimensional user profile in this application may also include other dimensions of information, such as user preference information. User preference dimensions reflect a user's interest in content type, style, or theme, such as whether they prefer entertainment videos, technology news, or shopping promotions. Preference features can be extracted by analyzing users' historical behavioral data (such as browsing history, search keywords, and clicked ad types).

[0060] Multi-dimensional user profiles can also include time and scenario information, including capturing the specific time and environmental context of user behavior, such as morning commute, lunch break, or evening leisure. This dimension can reflect the differences in user habits and needs at different times. Time and scenario are inferred through device clock and location data. Specific implementations include: (1) recording the timestamp of the behavior (e.g., 8:00 AM or 8:00 PM); (2) combining geographical location information (e.g., GPS location as a subway station or home) to infer the scenario; (3) performing pattern analysis on time and scenario data, for example, user C uses their phone on the subway from 8:00 AM to 9:00 AM every day, which is determined to be a "commuting scenario"; (4) quantifying the results into features (e.g., "morning commute = 1, evening leisure = 0") and adding them to the profile. For example, user C's profile can be expanded to "mid-range device, medium-speed network, low attention, commuting scenario", which is suitable for generating short-term, high-impact advertisements.

[0061] Multi-dimensional user profiles can also include basic attributes such as age, gender, and occupation, supplementing personal information that behavioral data cannot directly reflect, thus improving the comprehensiveness of the profile. This can be obtained through user registration information, third-party data, or behavioral inference. Explicit data can be obtained from user accounts (e.g., gender: male, age: 30 years old); if no direct data is available, it can be inferred from behavioral patterns. For example, if user D frequently browses maternity and baby products, it can be inferred as "female, 25-35 years old." Machine learning models (such as logistic regression) can be used to train an inferrer to predict gender and age group based on historical browsing categories (e.g., fashion, games); the results are then encoded into profile dimensions. For example, user D's profile could be "high-end devices, high-speed network, high attention span, female, 25-35 years old," making them suitable for pushing maternity and baby related advertisements.

[0062] The following describes step 205, namely "determining the ad length type based on the user profile, wherein the ad length type includes: ultra-short, short, and full," in detail with reference to the embodiments.

[0063] This step establishes a mapping relationship between dimensions such as device type, network status, and attention status in the user profile and ad length type (ultra-short, short, and full), ensuring that the ad duration can meet the user's current acceptance capacity while maximizing information delivery efficiency and user experience.

[0064] Ad length types are used to represent ads of different lengths. This application includes ad length types such as ultra-short, shorter, and full, with ad length increasing in that order. Ultra-short ads typically last 3-15 seconds, suitable for quickly conveying core information; shorter ads last 15-30 seconds, balancing information content and user tolerance; and full ads last 30 seconds or more, suitable for detailed narratives or brand displays. The process of determining ad length is based on user profile analysis, assessing user attention levels, network conditions, and device performance to determine their acceptance of different ad lengths. For example, users with low attention spans may find long ads difficult to accept, while high-speed networks and high-performance devices support longer loading and playback times. This step can be achieved using rule mapping or machine learning models, transforming profile features into length type decisions.

[0065] In determining the ad type, the system sets clear judgment criteria based on various dimensions of the user profile. For example, if the user profile indicates "low attention span" or "slow network," then ultra-short ads are selected because their short duration quickly attracts attention and adapts to weak network environments; if it indicates "medium attention span" or "medium network," then shorter ads are selected to provide a moderate amount of information; if it indicates "high attention span" and "high network," then full-length ads are selected to deliver complete content during the user's focused attention time. Device type is used as a supplementary dimension: high-performance devices (such as high-end smartphones) support the complex content of full-length ads, while low-end devices tend to use ultra-short ads to reduce loading burden.

[0066] As an implementable approach, determining the ad length type based on the user profile includes: determining the ad length type as ultra-short for users with low attention or low network speed; determining the ad length type as relatively short for users with medium attention or medium network speed; and determining the ad length type as full for users with high attention or high network speed. The user's historical ad tolerance, real-time feedback, and / or currently viewed content are determined based on the user behavior data. The ad length type is then dynamically adjusted based on the user's historical ad tolerance, real-time feedback, and / or currently viewed content.

[0067] Specifically, this implementation uses a dynamic adjustment mechanism to modify the ad length type initially selected based on static user profile characteristics (such as attention state and network state). The user profile provides an initial judgment; for example, users with low attention tend to prefer ultra-short ads, but actual user acceptance may vary due to historical habits, current reactions, or changes in interests. "Historical ad tolerance" reflects a user's long-term acceptance of ad length, such as whether they are accustomed to watching long ads; "real-time feedback" captures the user's immediate behavior during ad playback, such as whether they skip or interact; and "currently viewed content" reveals the user's immediate interests, which may influence their tolerance for ad length. Through these three factors, the system can dynamically adjust the length type, avoiding a decline in user experience or poor ad performance due to overly rigid initial selection. For example, if user tolerance is high, even with low attention, they may accept longer ads; if the real-time skip rate is high, the ad length needs to be shortened.

[0068] The adjustment method based on historical ad tolerance is as follows: First, collect historical user data, such as the viewing duration of each ad over the past 30 days (average 15 seconds), skip time (usually skipped after 10 seconds), and completion rate (60%). Then, calculate the tolerance threshold, for example, by statistically analyzing the average viewing duration or the skip threshold, to categorize tolerance (low, medium, high). Finally, adjust the ad length type based on the tolerance. For example, user A's profile is "low attention, slow network," initially selecting an ultra-short ad (5 seconds). However, their historical data shows an average viewing time of 20 seconds, indicating medium tolerance, so the system adjusts to a shorter ad (15 seconds). This adjustment is achieved through rule mapping, such as "if tolerance is higher than expected for attention state, extend the ad length by one level," ensuring the ad duration better matches user habits.

[0069] The adjustment method based on real-time feedback is as follows: First, monitor user interaction behavior in real time, such as whether they skip the ad (skip after 5 seconds), mute, fast forward, and / or switch apps; then, analyze feedback trends, such as an increase in skip rate (from 20% to 50%) or an increase in interaction frequency (2 clicks per minute); finally, adjust the length based on feedback. For example, if user B's profile is "medium attention, medium network speed," a shorter ad (15 seconds) is initially selected, but the user skips at 5 seconds, and real-time feedback shows decreased tolerance, the system shortens it to an ultra-short ad (10 seconds). The adjustment logic can be implemented using conditional rules (such as "skip rate > 50%, shorten by one level") or machine learning models (such as reinforcement learning, optimizing length with completion as a reward). This approach ensures that ad length responds instantly to changes in user status.

[0070] The adjustment method based on the user's current browsing content is as follows: First, collect the current page content, for example, through NLP analysis of keywords ("technology news", "promotional page") or page type (video, article); second, assess the content's impact on attention, for example, video content may increase tolerance, while fast-paced social media browsing may decrease tolerance; finally, adjust the length and type. For example, user C's profile is "high attention, high-speed network," initially selecting a full-length video (30 seconds), but currently browsing a fast-scrolling short video page, the system infers that their tolerance has decreased, adjusting to a shorter length (20 seconds). Adjustments can be achieved through predefined rules (such as "short content pages shorten by one level") or model prediction (input content features, output length offset).

[0071] In practical applications, the user's historical ad tolerance, real-time feedback, and currently viewed content can be used individually or in combination to form a comprehensive adjustment strategy. Specific implementations include: (1) setting priorities, such as prioritizing real-time feedback over tolerance, and tolerance over viewed content; (2) weighted fusion, such as tolerability weighted at 0.4, feedback at 0.4, and content at 0.2, to calculate the adjusted length value; and (3) dynamic iteration, recording the effect (e.g., completion rate) after each adjustment to optimize the next decision. For example, user D's profile is "low attention, low network speed," initially set to ultra-short (5 seconds), with moderate historical tolerance (15 seconds), no skipped real-time feedback, and currently viewing a technology article; the overall adjustment is set to shorter (15 seconds). This method, through multi-factor collaboration, ensures that the length type both fits the initial profile and adapts to dynamic changes.

[0072] As a feasible method, this application can dynamically adjust the ad length type based on the user's historical ad tolerance and real-time feedback. Specifically, it involves: collecting the user's past ad viewing records, including the complete viewing duration of each ad, the skip time, and / or the skip rate; calculating the user's ad tolerance threshold, wherein the tolerance threshold is determined by statistically analyzing the user's average viewing duration or the critical duration for skipping ads; classifying users into low tolerance, medium tolerance, and high tolerance types based on the tolerance threshold; matching the corresponding ad length according to the user classification to obtain a first length value; monitoring the user's interactive behavior in real time during ad playback, including whether to skip the ad, perform a mute operation, fast-forwarding, and / or switching applications; shortening the first length value if real-time feedback shows an increased skip rate or decreased tolerance; extending the first length value if real-time feedback shows that the user watched the ad completely and had a high frequency of interactive behavior; and determining the ad length type based on the first length value.

[0073] In determining ad types, machine learning models, such as pre-trained ad length determination models, can also be used. As an implementable approach, a pre-trained ad length determination model can be used to determine the ad length type based on the user profile. The pre-training includes: acquiring a training dataset, which includes multi-dimensional user profile samples and their corresponding ad length type labels and ad viewing behavior data; training the ad length determination model using supervised learning methods, with the user profile as input and the ad length type as output; incorporating time-dimensional features during training, including the current time period and usage scenario information; the ad length determination model outputs the probability distribution of each length type and selects the ad length type based on a preset threshold or the maximum probability; the training objective includes minimizing the difference between the ad length type and the ad length type label.

[0074] Specifically, the pre-training process involves acquiring a training dataset containing user profile samples, ad length labels, and viewing behavior data. Using user profiles as input and ad length type as output, the trained model outputs a probability distribution for each length type and selects the final type based on a preset threshold or the maximum probability. The training objective is to minimize the difference between the predicted results and the actual labels, ensuring model accuracy. In particular, incorporating time-dimensional features (such as the current time period and usage scenario information) enhances the model's context-awareness, making ad length selection more closely aligned with the user's actual state.

[0075] This application introduces dynamic information about time and context to enhance the ad length determination model's ability to understand and predict user behavior. The time dimension feature places user behavior within a specific spatiotemporal context, avoiding the problem of static models ignoring changes in user state over time. For example, the same user's attention level and ad acceptance may differ drastically between morning commutes and evening leisure time. By incorporating these features into the training process, the model can learn the impact of time and context on ad length preferences, thereby generating ads that better meet user needs. Specifically, the "current time period" refers to the specific time range in which user behavior occurs, typically divided into hours or finer-grained units, such as morning (6:00-9:00), noon (11:00-14:00), and evening (18:00-22:00). This feature can be directly obtained from the device's timestamp data and input into the model in numerical form (e.g., one-hot encoding: morning = [1,0,0]). The introduction of time periods is based on the periodicity of user behavior patterns. For example, research shows that users tend to browse quickly during morning commutes, making them suitable for ultra-short ads; while users may be more focused during evening leisure time, accepting full-length ads. During training, the model learns the optimal length for different time periods by analyzing the relationship between time periods and ad length labels in the dataset. For example, if morning samples in the dataset mostly correspond to very short labels, the model will tend to predict shorter durations in the morning.

[0076] "Usage scenario information" further enriches the semantics of the time dimension, referring to the specific environment or activity context in which user behavior occurs, such as commuting, resting at home, or working. This feature is typically inferred by combining timestamps with geolocation data (GPS) or behavioral patterns. For example, a user quickly navigating pages near a subway station at 8:00 AM daily could be inferred as a "commuting scenario"; a user spending a long time on a fixed Wi-Fi connection at 8:00 PM might indicate "relaxing at home." Scenario information can be represented by classification labels (e.g., commuting = 1, leisure = 0) or embedding vectors and added to the training data. The correlation between scenario and ad length lies in the fact that users' attention allocation and tolerance differ in different scenarios. For example, commuting scenarios may require short and fast ads, while leisure scenarios are more suitable for longer narrative ads. By learning these patterns, the model enhances its adaptability to scenario changes.

[0077] In terms of technical implementation, the specific method for incorporating time-dimensional features is to use "current time period" and "usage scenario information" as additional input features, concatenating them with other dimensions of the user profile (such as device type, network status, and attention status) to form a complete feature vector. For example, user A's profile vector might be [Device: 1, Network: 1, Attention: 0.5, Time Period: Morning = 1, Scenario: Commuting = 1]. During training, this extended vector is input into a supervised learning model (such as a deep neural network, DNN), and the weights are optimized through backpropagation, enabling the model to capture the impact of the time dimension on ad length selection. If the dataset shows that morning commuters tend to choose ultra-short ads, the model will increase the prediction probability of ultra-short ads under similar conditions. This approach ensures that the model not only focuses on static user attributes but also adapts to dynamic scenarios.

[0078] The following describes step 206, namely "generating an advertisement based on the advertisement length type and user profile using an advertisement clipping model", in detail with reference to the embodiments.

[0079] An ad editing model is an AI-based system that typically integrates computer vision, natural language processing (NLP), and video editing techniques. It extracts content from a resource library and edits it according to a specified length and user characteristics. Its workflow includes three main steps: First, it filters content that matches user preferences based on user profiles, such as selecting detailed product displays related to the interests of users with "high attention spans" or choosing lightweight content for users with "low-speed internet"; second, it determines the editing framework based on the ad's length type, for example, ultra-short ads (5 seconds) need to highly condense core information, while full-length ads (30 seconds) can retain more narrative elements; finally, it adjusts the editing rhythm and content layout through algorithms to generate the final ad. This process ensures that the ad is not only of appropriate length but also maximizes user engagement and message delivery.

[0080] In terms of technical implementation, the ad editing model first utilizes computer vision technology to analyze video content in a media library. For example, it uses convolutional neural networks or object detection models to identify key frames in the video, including brand logos, product close-ups, or emotionally charged scenes, and scores each frame to assess its importance. Then, the model adjusts the editing rhythm based on the user's attention state: for "high attention states," the scene transitions are slower (e.g., once every 5 seconds) to preserve details; for "low attention states," the transitions are faster (e.g., once every 2 seconds) to highlight key points; and for "distracted states," dynamic effects (e.g., flashing) are added to attract attention. Next, the model trims the content according to length type; for example, ultra-short ads retain only 3-5 high-scoring frames, while full-length ads stitch together more frames and optimize narrative coherence. Finally, NLP technology can optimize ad copy and voice-over to ensure that information is clearly conveyed within the specified duration.

[0081] Generative technologies can also be integrated into ad editing models to further enhance personalization. For example, through Generative Adversarial Networks (GANs) or Text-to-Video (TTV) technology, models can generate entirely new content based on user profiles, rather than relying solely on existing materials. For instance, if a user profile indicates a preference for "technology content," a GAN can generate footage of a virtual spokesperson explaining a tech product, while a TTV can convert the text "Latest smartphone launch" into a dynamic video. These technologies, combined with traditional editing, make ad content more tailored to user needs. For example, for users with "high-end devices and high attention spans," the model might generate a full 30-second ad containing GAN-synthesized tech scenes; for users with "low-end devices and low attention spans," a 5-second ultra-short ad could be edited, showcasing only a close-up of the product.

[0082] As one feasible approach, the historical behavioral data in this application includes video clips, image styles, and text content reflecting users' historical preferences. The ad editing model, based on ad length type and the user profile, generates ads by: selecting ad materials matching the user profile from an ad material library based on the historical behavioral data; extracting key frames from the materials using computer vision technology and adjusting the editing rhythm and visual effects in conjunction with the user's attention state, wherein the editing rhythm includes: frame switching frequency and information density; optimizing ad copy and voice-over using natural language processing technology; and generating personalized ad content that conforms to user preferences through generative adversarial networks or text-to-video technology.

[0083] Specifically, this application adaptively controls the presentation rhythm and visual effects of advertisements based on the user's attention level, ensuring that the advertisement content matches the user's acceptance capacity. Attention state directly affects the user's ability to process information: in a high-attention state, users can accept a slower pace and higher information density; in a low-attention state, users need quick switching and concise information to maintain interest; in a distracted state, they need high-frequency switching and strong visual stimulation to re-attract attention. "Screen switching frequency" determines the editing speed; for example, switching every 5 seconds is a slow pace, while switching every 1 second is a fast pace. "Information density" controls the complexity of the content; for example, low density only displays the brand logo, while high density includes product close-ups, usage scenarios, and other multiple information. By adjusting these two factors, the advertisement editing model can generate advertisement versions adapted to the user's attention state.

[0084] In implementation, the ad editing model first uses attention state information to set the editing strategy. The specific steps include: (1) Analyzing attention state: Obtaining attention state (such as high, low, distracted) from user profiles, which is determined based on historical and real-time behavioral data; (2) Adjusting screen switching frequency: Pre-setting rhythm templates according to attention state, for example, setting a slow rhythm (3-5 seconds / time) for high attention, a fast rhythm (1-2 seconds / time) for low attention, and an extremely fast rhythm (0.5-1 second / time) for distracted state; (3) Adjusting information density: Selecting the complexity of screen content in combination with attention state, for example, retaining multi-layered information for high attention, and simplifying to core elements for low attention; (4) Optimizing screen effects: Adding visual effects (such as fade-in / fade-out, flashing) according to rhythm and density to enhance user experience. Computer vision technology is used to extract key images (such as product close-ups, logos) and splice them according to the adjusted rhythm to finally generate an ad.

[0085] As another feasible approach, this application can also generate advertisements by incorporating weather information. Weather conditions affect users' lifestyles and behavioral patterns, thereby indirectly influencing their acceptance of advertisement duration. Specifically, weather information is obtained, and the weather conditions are categorized based on this information to obtain weather types, including: suitable for going out and unsuitable for going out. The advertisement editing model modifies the advertisement length type according to the weather type and generates advertisements based on the user profile.

[0086] In implementation, firstly, weather information is acquired and categorized. The latitude and longitude of the user's location are obtained in real-time through the device's location services (such as GPS), combined with weather data obtained from a weather API, such as temperature (25°C), rainfall (10mm / h), and wind speed (5m / s). Then, weather types are categorized according to predefined rules: if there is no rainfall, the temperature is suitable (e.g., 15-30°C), and the wind speed is low, it is classified as "suitable for going out"; if there is rainfall, extreme temperatures (e.g., <0°C or >35°C), and strong winds, it is classified as "unsuitable for going out". Weather parameter thresholds can be preset (e.g., rainfall >5mm / h is unsuitable for going out), or a classifier can be trained using machine learning (e.g., decision trees). Secondly, the ad length is modified according to the weather type using adjustment rules or machine learning models. For example, "unsuitable for going out" ads tend to be longer (e.g., from shorter to full length) because users may prefer to stay indoors; "suitable for going out" ads tend to be shorter (e.g., from full length to shorter length) to accommodate users' potential outdoor activity preferences. Finally, the ad editing model edits the footage based on the adjusted length and user profile to generate an ad.

[0087] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0088] According to another embodiment, an artificial intelligence advertising generation device based on user profiles is provided. Figure 3 A schematic block diagram of an AI-powered ad generation apparatus based on user profiles according to one embodiment is shown, such as Figure 3 As shown, the device 300 includes:

[0089] The information acquisition unit 301 is configured to acquire user behavior information and device status information, wherein the user behavior information includes historical behavior data and real-time behavior data.

[0090] The information analysis unit 302 is configured to determine the device type and network status based on the device status information.

[0091] Attention state determination unit 303 is configured to determine the user's attention state information based on the user behavior information.

[0092] User profile generation unit 304 is configured to generate multi-dimensional user profiles based on the device type, network status, and attention status information.

[0093] The length type determination unit 305 is configured to determine the advertisement length type based on the user profile, wherein the advertisement length type includes: ultra-short, short, and full.

[0094] The ad generation unit 306 is configured to generate an ad based on the ad length type and the user profile using an ad clipping model.

[0095] As one possible implementation method, the real-time behavior data includes: current page dwell time and interaction frequency; when the attention state determination unit 303 determines the user's attention state information based on the user behavior information, it can be configured to: determine the attention state as high attention, low attention, or distraction based on the current page dwell time and the interaction frequency; and correct the attention state determination result by comparing the current behavior with the user's historical behavior pattern in conjunction with the historical behavior data.

[0096] As one possible implementation, the length type determination unit 305 can be configured to determine the ad length type based on the user profile by: using a pre-trained ad length determination model to determine the ad length type based on the user profile; wherein, the pre-training includes: acquiring a training dataset, which includes: multi-dimensional user profile samples and their corresponding ad length type labels and ad viewing behavior data; training the ad length determination model using a supervised learning method, with the user profile as input; the ad length determination model outputs the probability distribution of each length type, and determines the ad length type based on a preset threshold or the maximum probability; during the training process, a time dimension feature is added, which includes: the current time period and usage scenario information; the training objective includes: minimizing the difference between the ad length type and the ad length type label.

[0097] As one possible implementation, the length type determination unit 305 can be configured to determine the ad length type based on the user profile as follows: for users whose attention state information indicates a low attention state or low network speed, determine the ad length type as ultra-short; for users whose attention state information indicates a medium attention state or medium network speed, determine the ad length type as relatively short; for users whose attention state information indicates a high attention state or high network speed, determine the ad length type as full; determine the user's historical ad tolerance, real-time feedback, and / or the content currently viewed by the user based on the user behavior data; and dynamically adjust the ad length type based on the user's historical ad tolerance, real-time feedback, and / or the content currently viewed by the user.

[0098] As one possible implementation, the length type determination unit 305, when dynamically adjusting the ad length type based on the user's historical ad tolerance and the real-time feedback, can be configured to: collect the user's past ad viewing records, including: the complete viewing duration of each ad, the skip time point, and / or the skip rate; calculate the user's ad tolerance threshold, wherein the tolerance threshold is determined by statistically analyzing the user's average viewing duration or the critical duration for skipping ads; classify the user into low tolerance, medium tolerance, and high tolerance types based on the tolerance threshold; match the corresponding ad length according to the user classification to obtain a first length value; during ad playback, monitor the user's interactive behavior in real time to obtain real-time feedback, including: whether to skip the ad, perform a mute operation, fast-forward behavior, and / or switch applications; if the real-time feedback shows that the user's skip rate increases or tolerance decreases, shorten the first length value; if the real-time feedback shows that the user watches the ad completely and has a high frequency of interactive behavior, extend the first length value; and determine the ad length type based on the first length value.

[0099] As one possible approach, historical behavioral data includes video clips, image styles, and text content reflecting users' historical preferences. When generating ads using an ad editing model based on the ad length type and the user profile, the ad generation unit 306 can be configured to: filter ad materials matching the user profile from an ad material library based on the historical behavioral data; extract key images from the ad materials using computer vision technology and adjust the editing rhythm and visual effects in conjunction with the user's attention state, wherein the editing rhythm includes: screen switching frequency and information density; optimize ad copy and voice-over using natural language processing technology; and generate personalized ad content that matches user preferences through generative adversarial networks or text-to-video technology.

[0100] As one possible implementation method, when the ad generation unit 306 generates an ad using an ad editing model based on the ad length type and the user profile, it can be configured to: obtain weather information, classify the weather conditions based on the weather information to obtain a weather type, wherein the weather type includes: suitable for going out and unsuitable for going out; the ad editing model modifies the ad length type based on the weather type, and generates an ad based on the ad length type and the user profile.

[0101] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0102] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0103] In addition, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0104] And an electronic device, comprising:

[0105] One or more processors; and

[0106] A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any of the foregoing method embodiments.

[0107] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0108] in, Figure 4An exemplary architecture of an electronic device is shown, which may include a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, and a memory 420. The processor 410, video display adapter 411, disk drive 412, input / output interface 413, network interface 414, and memory 420 can communicate with each other via a communication bus 430.

[0109] The processor 410 can be implemented using a general-purpose CPU, microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and implement the technical solution provided in this application.

[0110] The memory 420 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 420 can store the operating system 421 for controlling the operation of the electronic device 400, and the basic input / output system (BIOS) 422 for controlling the low-level operations of the electronic device 400. Additionally, it can store a web browser 423, a data storage management system 424, and an AI-powered advertising generation device 425 based on user profiles, etc. The aforementioned AI-powered advertising generation device 425 based on user profiles can be the application program that specifically implements the aforementioned steps in this embodiment. In summary, when implementing the technical solution provided in this application through software or firmware, the relevant program code is stored in the memory 420 and executed by the processor 410.

[0111] Input / output interface 413 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0112] Network interface 414 is used to connect a communication module (not shown in the figure) to enable communication and interaction between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0113] Bus 430 includes a pathway for transmitting information between various components of the device, such as processor 410, video display adapter 411, disk drive 412, input / output interface 413, network interface 414, and memory 420.

[0114] It should be noted that although the above-described device only shows the processor 410, video display adapter 411, disk drive 412, input / output interface 413, network interface 414, memory 420, bus 430, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the solution of this application, and does not necessarily include all the components shown in the figures.

[0115] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer program product. This computer program product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0116] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A user profile-based artificial intelligence advertisement generation method, characterized by, The method comprises: acquiring user behavior information and device state information, the user behavior information including historical behavior data and real-time behavior data; determining device type and network state according to the device state information; determining user attention state information according to the user behavior information, the user behavior information including average page jump time and jump time interval; if the average page jump time is less than a first threshold value, the user is determined to be in a low attention state; if the average page jump time is greater than or equal to the first threshold value and less than or equal to a second threshold value, the user is determined to be in a medium attention state; if the average page jump time is greater than the second threshold value, the user is determined to be in a high attention state; if the jump time interval is irregular, the user is determined to be in a distraction state; if the user behavior information does not match historical jump time patterns, the attention state information is corrected according to the historical jump time patterns; generating a multi-dimensional user portrait according to the device type, network state and attention state information; determining an advertisement length type according to the user portrait, the advertisement length type including: ultra-short type, shorter type and complete type; determining user historical advertisement tolerance, real-time feedback and user current browsing content according to the user behavior data; dynamically adjusting the advertisement length type according to the user historical advertisement tolerance, real-time feedback and user current browsing content; wherein dynamic adjustment according to the real-time feedback is prior to dynamic adjustment according to the advertisement tolerance, and dynamic adjustment according to the advertisement tolerance is prior to dynamic adjustment according to the user current browsing content; wherein dynamically adjusting the advertisement length type according to the user current browsing content comprises: collecting current page content; evaluating the influence of the page content on attention, and adjusting the advertisement length type according to a predefined rule; the predefined rule includes: if the page content is video content, the advertisement length type is adjusted to a longer level of advertisement length type; if the page content is social media content, the advertisement length type is adjusted to a shorter level of advertisement length type; generating an advertisement according to the advertisement length type and the user portrait using an advertisement editing model; dynamically adjusting the advertisement length type according to the user historical advertisement tolerance and the real-time feedback comprises: collecting user past advertisement viewing records, the advertisement viewing records including: complete viewing duration of each advertisement, skip time point and / or skip rate; calculating user advertisement tolerance threshold, wherein the tolerance threshold is determined by statistics of user average viewing duration or critical duration of skipping advertisements; classifying users into low tolerance type, medium tolerance type and high tolerance type according to the tolerance threshold; obtaining a first length value by matching the user classification with the corresponding advertisement length; real-time monitoring user interactive behavior to obtain real-time feedback during the advertisement playing process, the interactive behavior including: whether to skip the advertisement, execute mute operation, fast forward behavior and / or switch application program; shortening the first length value if the real-time feedback shows that the user skip rate increases or the tolerance decreases; if the real-time feedback shows that the user has watched the advertisement completely and the frequency of interaction behavior is high, the first length value is extended; determine the advertisement length type according to the first length value.

2. The method of claim 1, wherein, the advertisement length type is determined according to the user portrait, including: an advertisement length determination model obtained by pre-training is used to determine the advertisement length type according to the user portrait; the pre-training includes: obtaining a training data set, the training data set including: multi-dimensional user portrait samples and their corresponding advertisement length type labels and advertisement watching behavior data; using a supervised learning method to train the advertisement length determination model, taking the user portrait as input; the advertisement length determination model outputs the probability distribution of each length type, and determines the advertisement length type according to a preset threshold or the maximum probability; in the training process, time dimension features are added, including: current time period and use scene information; the training target of the pre-training includes: minimizing the difference between the advertisement length type and the advertisement length type label.

3. The method of claim 1, wherein, the advertisement length type is determined according to the user portrait, including: for the attention state information being a low attention state or a low-speed network user, determining the advertisement length type to be a super short type; for the attention state information being a medium attention state or a medium-speed network user, determining the advertisement length type to be a short type; for the attention state information being a high attention state or a high-speed network user, determining the advertisement length type to be a complete type.

4. The method of claim 1, wherein, the historical behavior data includes user historical preferences for video clips, image styles and script content; the advertisement is generated according to the advertisement length type and the user portrait by using the advertisement editing model, including: according to the historical behavior data, filtering materials matching the user portrait from an advertisement material library; using computer vision technology to extract key pictures in the materials, and adjusting the editing rhythm and picture effect in combination with the attention state, the editing rhythm including: picture switching frequency and information density; using natural language processing technology to optimize advertisement scripts and dubbing; generating personalized advertisement content conforming to user preferences through generative adversarial networks or text-to-video technology.

5. The method of claim 1, wherein, the advertisement is generated according to the advertisement length type and the user portrait by using the advertisement editing model, including: obtaining weather condition information, classifying weather conditions according to the weather condition information to obtain weather types, the weather types including: suitable for going out type and not suitable for going out type; the advertisement editing model modifies the advertisement length type according to the weather type, and generates an advertisement according to the advertisement length type and the user portrait. 6.A user profile-based artificial intelligence advertisement generation apparatus, characterized by, the device includes: an information acquisition unit configured to acquire user behavior information and device state information, the user behavior information including historical behavior data and real-time behavior data; an information analysis unit configured to determine device type and network state according to the device state information; The attention state determining unit is configured to determine the attention state information of the user according to the user behavior information, wherein the user behavior information includes the average page jump time and the jump time interval; if the average page jump time is less than a first threshold value, the user is determined to be in a low attention state; if the average page jump time is greater than or equal to the first threshold value and less than or equal to a second threshold value, the user is determined to be in a medium attention state; if the average page jump time is greater than the second threshold value, the user is determined to be in a high attention state; if the jump time interval is irregular, the user is determined to be in a distraction state; if the user behavior information does not conform to a historical jump time mode, the attention state information is corrected according to the historical jump time mode; The user portrait generating unit is configured to generate a multi-dimensional user portrait according to the device type, the network state and the attention state information; The length type determining unit is configured to determine the advertisement length type according to the user portrait, wherein the advertisement length type includes an ultra-short type, a shorter type and a complete type; determine the historical advertisement tolerance of the user, the real-time feedback and the content currently browsed by the user according to the user behavior data; dynamically adjust the advertisement length type according to the historical advertisement tolerance of the user, the real-time feedback and the content currently browsed by the user; wherein the dynamic adjustment according to the real-time feedback is prior to the dynamic adjustment according to the advertisement tolerance, and the dynamic adjustment according to the advertisement tolerance is prior to the dynamic adjustment according to the content currently browsed by the user; wherein the dynamic adjustment of the advertisement length type according to the content currently browsed by the user includes: collecting the current page content; evaluating the influence of the page content on attention, and adjusting the advertisement length type according to a predefined rule; the predefined rule includes: if the page content is video content, the advertisement length type is adjusted to a longer level of advertisement length type; if the page content is social media content, the advertisement length type is adjusted to a shorter level of advertisement length type; The advertisement generating unit is configured to generate an advertisement according to the advertisement length type and the user portrait by using an advertisement clip model; The dynamic adjustment of the advertisement length type according to the historical advertisement tolerance of the user and the real-time feedback includes: Collecting the past advertisement viewing records of the user, wherein the advertisement viewing records include the complete viewing duration of each advertisement, the skipping time point and / or the skipping rate; Calculating the advertisement tolerance threshold of the user, wherein the tolerance threshold is determined by counting the average viewing duration of the user or the critical duration of skipping the advertisement; Classifying the user into a low-tolerance type, a medium-tolerance type and a high-tolerance type according to the tolerance threshold; Matching the corresponding advertisement length according to the user classification to obtain a first length value; Real-time monitoring the interactive behavior of the user to obtain real-time feedback during the advertisement playing process, wherein the interactive behavior includes whether to skip the advertisement, execute a mute operation, fast forward behavior and / or switch the application program; If the real-time feedback shows that the skipping rate of the user is increased or the tolerance is decreased, the first length value is shortened. if real-time feedback shows that the user watches the advertisement completely and the frequency of interaction behavior is high, then the first length value is extended; determining the advertisement length type according to the first length value.

7. An electronic device, comprising: comprise: one or more processors; and a memory associated with the one or more processors, the memory for storing program instructions that, when read and executed by the one or more processors, perform the steps of the method of any one of claims 1 to 5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method of any one of claims 1 to 5.

Images