Cloud video advertisement intelligent putting management method and system based on information identification

Abstract

A cloud video advertisement intelligent putting management method and system based on information identification relates to the technical field of advertisement putting, and comprises the following steps: obtaining user feature information, obtaining user sight focus data and behavior data based on data preprocessing according to the user feature information, obtaining cloud video advertisement information, and obtaining cloud video advertisement information based on information identification. And obtaining cloud video advertisement attribute information. According to the method, real-time requirements and scenarized intentions of the user are analyzed in real time through the eye movement track and the user behaviors and the behavior semantic knowledge graph constructed based on the multi-source data, the second fixation area of the user is obtained on the basis of the first fixation area of the user, and the user preference tag is extracted from the second fixation area of the user to reveal the deep preference of the user, so that the user experience is improved. The weight of the first preference tag is adjusted through the second preference tag and the third preference tag, so that the system can quickly adapt to the short-term preference change and fashion trend, and the flexibility and timeliness of advertisement putting are enhanced.

Description

Technical Field

[0001] This invention relates to the field of advertising delivery technology, specifically to a cloud video advertising intelligent delivery management method and system based on information recognition. Background Technology

[0002] Cloud video advertising is not only an essential industry need to improve advertising conversion efficiency and optimize the allocation of digital marketing resources by accurately matching user needs, but also a key path to balance commercial value and user experience through technological innovation and promote the transformation of advertising from "traffic delivery" to "value delivery". At the same time, it provides practical support for the integrated application of technologies in multiple fields and the standardized development of the digital ecosystem.

[0003] Current mainstream cloud video advertising systems have significant technical limitations: on the one hand, traditional recommendation methods based on user profiles and basic behavioral data struggle to deeply understand users' real-time needs and contextual intentions during viewing; on the other hand, existing systems generally lack the ability to quickly respond to short-term shifts in user preferences, resulting in a low match between cloud video ad delivery and users' immediate needs, severely impacting the effectiveness of ad delivery. Specifically, these limitations manifest as: (1) a superficial understanding of needs, failing to extract deep semantic information from video content and user interactions; (2) a lagging response mechanism, making it difficult to capture and adapt to dynamic changes in user attention in a timely manner; and (3) insufficient fusion of multimodal data, with biometric data such as eye-tracking trajectories and behaviors not being fully utilized. These problems lead to low ad conversion rates, poor user experience, and loss of commercial value. Summary of the Invention

[0004] To address the aforementioned technical issues, this paper provides a cloud video advertising intelligent delivery management method and system based on information recognition. This technical solution solves the problem that the recommendation methods based on user profiles and basic behavioral data mentioned in the background technology are unable to deeply understand the real-time needs and contextual intentions of users during the viewing process. Existing systems generally lack the ability to quickly respond to short-term shifts in user preferences, resulting in a low degree of matching between cloud video advertising push and users' immediate needs, which seriously affects the delivery effect.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A cloud video advertising intelligent delivery management method based on information recognition includes: Obtain user characteristic information, which is the eye movement trajectory and behavioral information recorded when the user watches cloud video advertisements; Based on user characteristic information and data preprocessing, user gaze focus data and behavioral data are obtained; Obtain cloud video ad information; based on information recognition, obtain cloud video ad attribute information. Based on gaze focus data and cloud video ad attribute information, and using eye tracking and target detection, key elements of user preferences in the cloud video ad attribute information are obtained. Based on key element information of user preferences and attribute information of cloud video ads, and using a deep learning model, the initial cloud video ads are obtained. The initial cloud video ads are the most matching ads in the cloud video ad pool obtained by the deep learning model by matching cloud video ad attribute information and key element information of user preferences. Acquire eye movement and behavioral information of users when they watch the initial cloud video ads, and update user gaze focus data and behavioral data; Acquire multi-source data and construct a behavioral semantic knowledge graph based on the multi-source data, wherein the multi-source data includes eye-tracking trajectories, behavioral data, item attributes, and behavioral semantic information; Preferably, the step of obtaining key element information of user preferences in the cloud video ad attribute information based on gaze focus data and cloud video ad attribute information, using eye tracking and target detection, specifically includes: Based on cloud video advertising information, a cloud video advertising image set is obtained through image preprocessing, wherein the image preprocessing includes video segmentation and image noise reduction. Based on the cloud video ad image set and information recognition, cloud video ad attribute information is obtained. The cloud video ad attribute information includes the item name information contained in the cloud video ad, the bounding box coordinate information of the area containing the item, and the brand information. Based on user characteristic information and data preprocessing, user gaze focus data is obtained. The user gaze focus data includes the gaze duration and frequency of the object area and the vertical distance between the upper and lower eyelids. Based on user gaze focus data, obtain the user preference index; Based on user preference index and cloud video ad attribute information, and using eye tracking and target detection, key element information of user preferences is obtained. The key element information includes a first key element label and a second key element label.

[0006] Preferably, obtaining the user preference index based on user gaze focus data specifically includes: Based on user gaze focus data, obtain the object gaze density; Specifically, the object gaze density is as follows: ; In the formula, Let i be the gaze density of the i-th item. Let be the duration of gaze on the i-th item. Total watch time for cloud video ads This represents the number of times the i-th item is viewed for more than 1 second. The total number of times all items were looked at for more than 1 second; Based on the vertical distance between the upper and lower eyelids, obtain the user's initial interocular distance and the average interocular distance of the object being gazed upon; The user preference index is obtained based on the user's initial interocular distance, the average interocular distance of the object being gazed at, and the gaze density. Specifically, the user preference index is: ; In the formula, Let be the user preference index for the i-th item, controlling for sensitivity to interocular distance. is the ratio of the average interocular distance for the i-th object being gazed at to the user's initial interocular distance.

[0007] Preferably, the step of obtaining key element information of user preferences based on user preference index and cloud video ad attribute information, using eye tracking and object detection, specifically includes: Obtain the bounding box coordinates of the top five items in the user preference index, and use the area within the bounding box as the first area of ​​user gaze. Based on the user's first gaze area, obtain the first key user preference tag, which is the item name information and brand information within the user's first gaze area; Based on the user's gaze focus data in the first region, the user's deep preference region is obtained based on hierarchical clustering and contour extraction. The user's deep preference region is the region obtained by dividing the user's gaze focus into different clusters based on hierarchical clustering and then extracting the contour of each cluster to generate a closed boundary. Based on the user's deep preference region and the item gaze density, the user gaze second region is obtained. The user gaze second region is the top five regions with the highest item gaze density within the user's deep preference region. Based on the user's gaze over the second region, and using object detection, a second key user preference label is obtained. The second key user preference label is the information about the names of items in the second area of ​​the user's gaze obtained from object detection.

[0008] Preferably, the cloud video advertising pool specifically includes: Based on user gaze focus data and behavioral data, and using a hybrid expert model, user preference scores are obtained. Based on user preference ratings, core user preference tags are obtained using feature aggregation methods. Based on the cloud video ad attribute information and user core preference tags, and using multi-dimensional similarity fusion, a comprehensive matching score for the cloud video ad is obtained. If the comprehensive matching score is greater than a preset threshold, the cloud video ad is added to the cloud video ad pool. Based on user preference scores, ads with the lowest user preference scores in the cloud video ad pool are eliminated; Obtain historical user gaze focus data, and based on the historical user gaze focus data, obtain the first preference label, which is the item name information and brand information corresponding to the item with the largest fluctuation in the number of gazes; Based on user preference scores and core user preference tags, a second preference tag is obtained using a prediction model; Based on big data analysis, keywords of current trending events are obtained as third-preference tags; The frequency fluctuation range is used as the weight of the first preference label; Based on third-party preference tags, obtain keywords for trending events; Based on key element information of user preferences and keywords of trending events, obtain the frequency of corresponding keywords of trending events; The ratio of the number of times keywords related to trending events appear to the number of times trending events occur is used as the heat value correction coefficient. Based on the key element information of user preferences and the second preference label, obtain the intersection label value, where the intersection label value is the total number of item name information and brand information that are the same as the key element information of user preferences in the second preference label; The calorific value correction factor is adjusted by the ratio of the intersecting label values ​​to the total number of item name and product information contained in the second preference label; Based on the calorific value correction coefficient, the weight of the first preference label is adjusted to obtain the weight of the first preference label; Update the cloud video ad pool based on the first preference tag, the second preference tag, and the third preference tag.

[0009] Preferably, adjusting the delivery of cloud video ads based on the behavioral semantic knowledge graph and a deep learning model specifically includes: Based on user preference ratings and a user clustering algorithm, similar user groups are obtained. These similar user groups are those whose user preference ratings for the initially delivered cloud video ads differ from the user ratings of the users in the initial ad campaign by less than 0.1. Based on similar user groups and using a preference advertising mining method, a high-value cloud video ad set is obtained. The high-value cloud video ad set is the set of cloud video ads corresponding to user preference scores of similar user groups for the same cloud video ad that are higher than a preset threshold. Based on the updated user gaze focus data, and using eye tracking, the user's real-time gaze intersection point is obtained; Based on the updated behavioral data, a real-time interaction intent vector is obtained using a behavior encoding network. Based on the user's real-time gaze intersection and real-time interaction intent vector, real-time preference tags are obtained using a behavioral semantic knowledge graph. Based on real-time preference tags and a hierarchical filtering method, cloud video ads that meet the filtering criteria in the cloud video ad pool are added to the high-value cloud video ad set. The filtering criteria are item name information and brand information. Based on real-time preference tags and core user preference tags, cloud video ads are retrieved from high-value cloud video ad sets using a deep learning model.

[0010] Furthermore, a cloud video advertising intelligent delivery management system based on information recognition is proposed to implement the construction method described above, including: The main control module is used to obtain core preference tags of a group based on user gaze focus data and behavior data; obtain a high-value cloud video ad set based on preference ad mining methods according to similar user groups; obtain the real-time gaze intersection of the user based on eye tracking based on user gaze focus data; obtain the real-time interaction intent vector based on behavior encoding network based on behavior data; obtain the real-time preference tag based on behavior semantic knowledge graph based on the real-time gaze intersection and real-time interaction intent vector; and obtain the cloud video ads to be delivered from the high-value cloud video ad set based on the real-time preference tag and the user's core preference tag and a deep learning model. The information processing module is used to obtain a cloud video advertisement image set based on image preprocessing according to cloud video advertisement information; obtain cloud video advertisement attribute information based on information recognition according to the cloud video advertisement image set; obtain user gaze focus data based on data preprocessing according to user feature information; obtain user preference index based on user gaze focus data; and obtain key element information of user preference based on eye tracking and target detection according to user preference index and cloud video advertisement attribute information. The cloud video ad pool module is used to obtain a comprehensive matching score for new cloud video ads based on user gaze focus data and behavioral data. If the comprehensive matching score is greater than a preset threshold, the cloud video ad information is added to the cloud video ad pool. Based on user preference scores, the ad with the lowest user preference score in the cloud video ad pool is eliminated. Based on historical user gaze focus data, a first preference tag is obtained. Based on user preference scores and core user preference tags, a second preference tag is obtained. Based on big data analysis, keywords of current hot events are obtained as a third preference tag. Based on the second and third preference tags, the weight of the first preference tag is adjusted. Based on the first, second, and third preference tags, the cloud video ad pool is updated.

[0011] Optionally, the main control module specifically includes: High-value advertising units are used to obtain core preference tags of a group based on user gaze focus data and behavioral data, and to obtain a set of high-value cloud video ads based on similar user groups and preference ad mining methods. The real-time preference unit is used to obtain the user's real-time gaze intersection based on eye tracking data, obtain the real-time interaction intent vector based on behavior data and behavior encoding network, and obtain the real-time preference label based on the user's real-time gaze intersection and real-time interaction intent vector and behavior semantic knowledge graph. An advertising delivery unit is used to obtain cloud video ads to be delivered from a high-value cloud video ad set based on a deep learning model, according to real-time preference tags and user core preference tags.

[0012] Optionally, the information processing module specifically includes: The information acquisition unit is used to acquire a cloud video advertisement image set based on image preprocessing according to cloud video advertisement information, acquire cloud video advertisement attribute information based on information recognition according to the cloud video advertisement image set, and acquire user gaze focus data based on user feature information and data preprocessing. The information processing unit is used to obtain a user preference index based on user gaze focus data, and to obtain key element information of user preference based on eye tracking and target detection, using the user preference index and cloud video advertising attribute information.

[0013] Optionally, the cloud video advertising pool module specifically includes: The update unit is used to obtain the comprehensive matching score of the new cloud video advertisement based on the user's gaze focus data and behavior data. If the comprehensive matching score is greater than a preset threshold, the cloud video advertisement information is added to the cloud video advertisement pool. An elimination unit is used to eliminate the advertisement with the lowest user preference score in the cloud video advertisement pool based on the user preference score. The preference unit is used to obtain a first preference tag based on historical user gaze focus data, obtain a second preference tag based on user preference rating and user core preference tag, obtain keywords of current hot events as a third preference tag based on big data analysis, and adjust the weight of the first preference tag based on the second preference tag and the third preference tag.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a cloud video advertising intelligent delivery management method and system based on information recognition. By analyzing eye movement trajectories and user behavior, and constructing a behavioral semantic knowledge graph based on multi-source data, the system can analyze users' real-time needs and contextual intentions in real time. Based on the user's first gaze area, the system can obtain the user's second gaze area and extract user preference tags from the second gaze area to reveal the user's deeper preferences. By adjusting the weight of the first preference tag through the second and third preference tags, the system can quickly adapt to short-term preference changes and popular trends, enhancing the flexibility and timeliness of advertising delivery. Attached Figure Description

[0015] Figure 1 This is a flowchart of a cloud video advertising intelligent delivery management method based on information recognition proposed in this invention; Figure 2 This is a flowchart illustrating the process of obtaining key elements of user preferences in the cloud video advertising attribute information of the present invention. Figure 3 This is a flowchart illustrating the cloud video advertising delivery process of the present invention. Figure 4 This is a structural block diagram of a cloud video advertising intelligent delivery management system based on information recognition proposed in this invention. Detailed Implementation

[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0017] Reference Figure 1 - Figure 3 As shown in the figure, an embodiment of the present invention provides a cloud video advertising intelligent delivery management method based on information recognition, comprising: Obtain user characteristic information, which is the eye movement trajectory and behavioral information recorded when the user watches cloud video advertisements; Based on user characteristic information and data preprocessing, user gaze focus data and behavioral data are obtained; Obtain cloud video ad information; based on information recognition, obtain cloud video ad attribute information. Based on gaze focus data and cloud video ad attribute information, and using eye tracking and target detection, key elements of user preferences in the cloud video ad attribute information are obtained. Specifically, based on gaze focus data and cloud video ad attribute information, and using eye tracking and object detection, key elements of user preferences within the cloud video ad attribute information are obtained, including: Based on cloud video advertising information, a cloud video advertising image set is obtained through image preprocessing, wherein the image preprocessing includes video segmentation and image noise reduction. Based on the cloud video ad image set and information recognition, cloud video ad attribute information is obtained. The cloud video ad attribute information includes the item name information contained in the cloud video ad, the bounding box coordinate information of the area containing the item, and the brand information. Based on user characteristic information and data preprocessing, user gaze focus data is obtained. The user gaze focus data includes the gaze duration and frequency of the object area and the vertical distance between the upper and lower eyelids. Based on user gaze focus data, obtain the user preference index; Based on user preference index and cloud video ad attribute information, and using eye tracking and target detection, key element information of user preferences is obtained, which includes a first key element label and a second key element label. Specifically, based on user gaze focus data, a user preference index is obtained, which includes: Based on user gaze focus data, obtain the object gaze density; Specifically, the object gaze density is as follows: ; In the formula, Let i be the gaze density of the i-th item. Let be the duration of gaze on the i-th item. Total watch time for cloud video ads This represents the number of times the i-th item is viewed for more than 1 second. The total number of times all items were looked at for more than 1 second; Based on the vertical distance between the upper and lower eyelids, obtain the user's initial interocular distance and the average interocular distance of the object being gazed upon; The user preference index is obtained based on the user's initial interocular distance, the average interocular distance of the object being gazed at, and the gaze density. Specifically, the user preference index is: ; In the formula, Let be the user preference index for the i-th item, controlling for sensitivity to interocular distance. is the ratio of the average interocular distance for the i-th object being gazed at to the user's initial interocular distance.

[0018] Specifically, based on user preference index and cloud video ad attribute information, and using eye tracking and object detection, key elements of user preferences are obtained, including: Obtain the bounding box coordinates of the top five items in the user preference index, and use the area within the bounding box as the first area of ​​user gaze. Based on the user's first gaze area, obtain the first key user preference tag, which is the item name information and brand information within the user's first gaze area; Based on the user's gaze focus data in the first region, the user's deep preference region is obtained based on hierarchical clustering and contour extraction. The user's deep preference region is the region obtained by dividing the user's gaze focus into different clusters based on hierarchical clustering and then extracting the contour of each cluster to generate a closed boundary. Based on the user's deep preference region and the item gaze density, the user gaze second region is obtained. The user gaze second region is the top five regions with the highest item gaze density within the user's deep preference region. Based on the user's gaze over the second region, and using object detection, a second key user preference label is obtained. The second key user preference label is the information on the names of items within the second area of ​​user gaze obtained from object detection. Specifically, if the object detection obtains the name information of the item in the second area that the user is looking at, it will be used as the second key user preference label; if the object detection does not obtain the name information of the item in the second area that the user is looking at, it will not be used as the second key user preference label. In this solution, the object gaze density is obtained through user gaze focus data. The user preference index is obtained by using the user's initial interpupillary distance, the average interpupillary distance of the gazed objects, and the gaze density. Based on the user preference index, the area within the bounding box corresponding to the top five items with the highest user preference index is taken as the first user gaze area. Based on the user gaze focus data and the first user gaze area, the user's deep preference area is obtained. The deep preference area is then filtered to obtain the second user gaze area. Based on the data information contained in the first and second user gaze areas, the accuracy and personalization of ad matching are improved, providing strong support for achieving intelligent and efficient cloud video ad delivery.

[0019] Understandably, focusing on the first area of ​​user gaze—the top five items users are most interested in from cloud video ads—can improve ad matching accuracy. However, in practical applications, relying solely on the first area of ​​user gaze may overlook important preferences, resulting in less comprehensive and refined ad recommendations. For example, the first area of ​​user gaze might include the location of a car, but the user might be more interested in the car's tires, rearview mirrors, and taillights. In this case, relying solely on the first area of ​​user gaze cannot capture the user's deeper preferences. On the other hand, the second area of ​​user gaze, obtained through the user's gaze focus, can reveal the user's deeper and potential preferences, dynamically adapting to user changes and providing more personalized and accurate ad recommendations. Combining the information from the first and second areas of user gaze allows for a more comprehensive understanding of user needs, improving the effectiveness and relevance of ad delivery.

[0020] Based on key element information of user preferences and attribute information of cloud video ads, and using a deep learning model, the initial cloud video ads are obtained. The initial cloud video ads are the most matching ads in the cloud video ad pool obtained by the deep learning model by matching cloud video ad attribute information and key element information of user preferences. Specifically, the cloud video ad pool includes: Based on user gaze focus data and behavioral data, and using a hybrid expert model, user preference scores are obtained. Based on user preference ratings, core user preference tags are obtained using feature aggregation methods. Based on the cloud video ad attribute information and user core preference tags, and using multi-dimensional similarity fusion, a comprehensive matching score for the new cloud video ad is obtained. If the comprehensive matching score is greater than a preset threshold, the cloud video ad information is added to the cloud video ad pool. The preset threshold is 0.75; Based on user preference scores, ads with the lowest user preference scores in the cloud video ad pool are eliminated; Obtain historical user gaze focus data, and based on the historical user gaze focus data, obtain the first preference label, which is the item name information and brand information corresponding to the item with the largest fluctuation in the number of gazes; Based on user preference scores and core user preference tags, a second preference tag is obtained using a prediction model; Based on big data analysis, keywords of current trending events are obtained as third-preference tags; Each 10-minute time window is used as a weight for the first preference label, with the frequency fluctuation amplitude as the weight. The weight of the first preference label is as follows: ; In the formula, The weight of the first preference label, For users in the first The time window for the first Number of times an item is viewed. It is a function with maximum value. It is a minimum value function; Based on third-party preference tags, obtain keywords for trending events; Based on key element information of user preferences and keywords of trending events, obtain the frequency of corresponding keywords of trending events; The ratio of the number of times keywords related to trending events appear to the number of times trending events occur is used as the heat value correction coefficient. The calorific value correction coefficient is specifically as follows: ; In the formula, The number of trending event keywords is obtained based on key element information of user preferences and trending event keywords. This represents the total number of times trending events occurred, as determined by big data analysis. Based on the key element information of user preferences and the second preference label, obtain the intersection label value, where the intersection label value is the total number of item name information and brand information that are the same as the key element information of user preferences in the second preference label; The calorific value correction factor is adjusted by the ratio of the intersecting label values ​​to the total number of item name and product information contained in the second preference label; Specifically, the adjustment to the calorific value correction coefficient is as follows: ; In the formula, It is the ratio of the intersecting label values ​​to the total number of item name and product description information contained in the second preference label; Based on the calorific value correction coefficient, the weight of the first preference label is adjusted to obtain the weight of the first preference label; Specifically, the weight of the first preference label is adjusted based on the calorific value correction coefficient as follows: ; In the formula, if If it is greater than 0.8, then... Reassign the value to 0.8, if If less than 0.5, then Reassign the value to 0.5; Update the cloud video ad pool based on the first preference tag, the second preference tag, and the third preference tag; Specifically, the cloud video ad pool is updated based on the first preference tag, the second preference tag, and the third preference tag, including: Based on the first preference tag and cloud video ad attributes, the first preference matching score of cloud video ads is obtained through multi-dimensional similarity fusion. Based on the second preference tag and cloud video ad attributes, the second preference matching score of cloud video ads is obtained through multi-dimensional similarity fusion. Based on the third preference tag and cloud video ad attributes, the third preference matching score of cloud video ads is obtained by multi-dimensional similarity fusion. Obtain the preference matching score based on the first preference matching score, the second preference matching score, and the third preference matching score; Specifically, the preference matching score is as follows: ; In the formula, The score is the first preference match score. The score is the second preference matching score. The score is matched to the third preference. like If the value is greater than 0.75, then this cloud video ad will be added to the cloud video ad pool; Acquire eye movement and behavioral information of users when they watch the initial cloud video ads, and update user gaze focus data and behavioral data; In this solution, user preference scores are obtained through user gaze focus data and behavioral data. Based on a comparison of these scores with a set threshold, the ads in the cloud video ad pool undergo a first-stage update. A first preference tag is obtained based on historical user gaze focus data. A second preference tag is obtained based on the user preference score and core user preference tags. A third preference tag is obtained based on big data analysis. The frequency fluctuation range is used as the weight of the first preference tag. The ratio of the number of times trending event keywords appear to the number of times trending events occur in the second preference tag is used as a heat value correction coefficient. Intersecting tag values ​​are obtained based on key element information of user preferences and the second preference tag. The heat value correction coefficient is adjusted by the ratio of the intersecting tag value to the total number of item names and product descriptions contained in the second preference tag. The weight of the first preference tag is adjusted based on the adjusted heat value correction coefficient. Finally, the ad pool undergoes a second-stage update based on the adjusted first preference tag weight, achieving accurate matching and efficient management of ad content.

[0021] Understandably, using first, second, and third preference tags can quickly match and cover multi-dimensional preferences. However, in practice, this method suffers from static weight allocation leading to biases and a lack of real-time adjustment mechanisms. Adjusting the weight of the first preference tag solely based on the third preference tag can result in over-adjustment. For example, if a user is interested in a car that happens to be a current hot topic, the weight of the first preference tag adjusted accordingly will be smaller than it actually is. However, using second and third preference tags will increase the weight of the first preference tag in such situations to make it closer to reality. Furthermore, adjusting the weight of the first preference tag based on the second and third preference tags can better adapt to short-term fluctuations in user preferences and changes in popular trends, thus optimizing ad pool management.

[0022] Acquire multi-source data and construct a behavioral semantic knowledge graph based on the multi-source data, wherein the multi-source data includes eye-tracking trajectories, behavioral data, item attributes, and behavioral semantic information; Based on the behavioral semantic knowledge graph and a deep learning model, the delivery of cloud video ads is adjusted.

[0023] Specifically, based on the behavioral semantic knowledge graph and a deep learning model, the delivery of cloud video ads is adjusted, including: Based on user preference ratings and a user clustering algorithm, similar user groups are obtained. These similar user groups are those whose user preference ratings for the initially delivered cloud video ads differ from the user ratings of the users in the initial ad campaign by less than 0.1. Based on similar user groups and using a preference advertising mining method, a high-value cloud video ad set is obtained. The high-value cloud video ad set is the set of cloud video ads corresponding to user preference scores of similar user groups for the same cloud video ad that are higher than a preset threshold. Based on the updated user gaze focus data, and using eye tracking, the user's real-time gaze intersection point is obtained; Based on the updated behavioral data, a real-time interaction intent vector is obtained using a behavior encoding network. Based on the user's real-time gaze intersection and real-time interaction intent vector, real-time preference tags are obtained using a behavioral semantic knowledge graph. Based on real-time preference tags and a hierarchical filtering method, cloud video ads that meet the filtering criteria in the cloud video ad pool are added to the high-value cloud video ad set. The filtering criteria are item name information and brand information. Based on real-time preference tags and core user preference tags, and using a deep learning model, cloud video ads are retrieved from high-value cloud video ad sets for delivery. This solution introduces a user clustering algorithm to identify similar user groups, combines this with a preference-based advertising mining method to construct a high-value ad set, and further utilizes eye-tracking and behavioral encoding networks to obtain real-time user gaze intersections and interaction intent vectors. Real-time preference tags are then generated based on a behavioral semantic knowledge graph, enabling dynamic capture of user preferences. On this basis, a hierarchical filtering method is used to update the high-value ad set, and combined with core user preference tags, a deep learning model is used to accurately select the most suitable ads from the ad set for delivery. This method effectively integrates users' long-term preferences and short-term behavioral intents, improving the real-time performance, accuracy, and personalization of ad matching, thereby significantly enhancing ad delivery effectiveness and user experience.

[0024] Reference Figure 4 As shown, further, combining the above-mentioned intelligent cloud video advertising delivery management method based on information recognition, a cloud video advertising intelligent delivery management system based on information recognition is proposed, including: The main control module is used to obtain user preference scores based on user gaze focus data and behavioral data using a hybrid expert model; obtain similar user groups based on user preference scores using a user clustering algorithm; obtain core preference tags for these groups based on the preference scores of similar user groups using a feature aggregation method; obtain a set of high-value cloud video ads based on similar user groups using a preference ad mining method; obtain real-time gaze intersections based on user gaze focus data using eye tracking; obtain real-time interaction intent vectors based on behavioral encoding networks using behavioral encoding networks; obtain real-time preference tags based on behavioral semantic knowledge graphs using real-time gaze intersections and real-time interaction intent vectors; and obtain cloud video ads to be delivered from the set of high-value cloud video ads based on real-time preference tags and core user preference tags using a deep learning model. The information processing module is used to obtain a cloud video advertisement image set based on image preprocessing according to cloud video advertisement information; obtain cloud video advertisement attribute information based on information recognition according to the cloud video advertisement image set; obtain user gaze focus data based on data preprocessing according to user feature information; obtain user preference index based on user gaze focus data; and obtain key element information of user preference based on eye tracking and target detection according to user preference index and cloud video advertisement attribute information. The cloud video ad pool module is used to obtain user preference scores based on user gaze focus data and behavioral data using a hybrid expert model. Based on the user preference scores, it obtains core user preference tags using a feature aggregation method. Based on cloud video ad attribute information and core user preference tags, it obtains a comprehensive matching score for new cloud video ads using multi-dimensional similarity fusion. If the comprehensive matching score is greater than a preset threshold, the cloud video ad information is added to the cloud video ad pool. Based on the user preference scores, the ad with the lowest user preference scores in the cloud video ad pool is eliminated. Historical user gaze focus data is obtained. Based on the historical user gaze focus data, a first preference tag is obtained. Based on the user preference scores and core user preference tags, a second preference tag is obtained using a prediction model. Based on big data analysis, it obtains... The keywords of current trending events are taken as the third preference tag, and the frequency fluctuation range is used as the weight of the first preference tag. Based on the third preference tag, the keywords of trending events are obtained. According to the key element information of user preferences and the keywords of trending events, the frequency of the keywords of trending events is obtained. The ratio of the frequency of keywords of trending events to the frequency of occurrence of trending events is used as the heat value correction coefficient. According to the key element information of user preferences and the second preference tag, the intersection tag value is obtained. The heat value correction coefficient is adjusted by the ratio of the intersection tag value to the total number of item name information and product name information contained in the second preference tag. Based on the adjusted heat value correction coefficient, the weight of the first preference tag is adjusted, and the weight of the first preference tag is obtained. The cloud video ad pool is updated according to the first preference tag, the second preference tag and the third preference tag.

[0025] The main control module specifically includes: The high-value advertising unit is used to obtain user preference scores based on user gaze focus data and behavioral data, using a hybrid expert model; to obtain similar user groups based on user preference scores and user clustering algorithms; to obtain core preference tags of the groups based on the preference scores of the similar user groups and feature aggregation methods; and to obtain a set of high-value cloud video ads based on similar user groups and preference ad mining methods. The real-time preference unit is used to obtain the user's real-time gaze intersection based on eye tracking data, obtain the real-time interaction intent vector based on behavior data and behavior encoding network, and obtain the real-time preference label based on the user's real-time gaze intersection and real-time interaction intent vector and behavior semantic knowledge graph. An advertising delivery unit is used to obtain cloud video ads to be delivered from a high-value cloud video ad set based on a deep learning model, according to real-time preference tags and user core preference tags.

[0026] The information processing module specifically includes: The information acquisition unit is used to acquire a cloud video advertisement image set based on image preprocessing according to cloud video advertisement information, acquire cloud video advertisement attribute information based on information recognition according to the cloud video advertisement image set, and acquire user gaze focus data based on user feature information and data preprocessing. The information processing unit is used to obtain a user preference index based on user gaze focus data, and to obtain key element information of user preference based on eye tracking and target detection, using the user preference index and cloud video advertising attribute information.

[0027] The cloud video ad pool module specifically includes: The update unit is used to obtain a user preference score based on a hybrid expert model based on user gaze focus data and behavior data; obtain a user core preference tag based on a feature aggregation method based on the user preference score; obtain a comprehensive matching score of a new cloud video advertisement based on cloud video advertisement attribute information and user core preference tags based on multi-dimensional similarity fusion; and add the cloud video advertisement information to the cloud video advertisement pool if the comprehensive matching score is greater than a preset threshold. An elimination unit is used to eliminate the advertisement with the lowest user preference score in the cloud video advertisement pool based on the user preference score. The preference unit is used to acquire historical user gaze focus data, acquire a first preference tag based on the historical user gaze focus data, acquire a second preference tag based on user preference scores and core user preference tags, and based on a prediction model, acquire keywords of current hot events as a third preference tag based on big data analysis, use the frequency fluctuation range as the weight of the first preference tag, acquire hot event keywords based on the third preference tag, acquire the frequency of hot event keywords based on key element information of user preferences and hot event keywords, use the ratio of the frequency of hot event keywords to the frequency of hot event occurrences as a heat value correction coefficient, acquire intersection tag values ​​based on key element information of user preferences and the second preference tag, adjust the heat value correction coefficient based on the ratio of the intersection tag values ​​to the total number of item name information and product description information contained in the second preference tag, and adjust the weight of the first preference tag based on the adjusted heat value correction coefficient to acquire the weight of the first preference tag.

[0028] In summary, the advantages of this invention are as follows: by using eye-tracking trajectories and user behavior, and constructing a behavioral semantic knowledge graph based on multi-source data, the system can analyze users' real-time needs and contextualized intentions in real time. Based on the user's first gaze region, the system can obtain the user's second gaze region, extract user preference tags from the second gaze region to reveal the user's deeper preferences, and adjust the weight of the first preference tag through the second and third preference tags. This allows the system to quickly adapt to short-term changes in user preferences and popular trends, enhancing the flexibility and timeliness of advertising.

[0029] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A cloud video advertising intelligent delivery management method based on information recognition, characterized in that, include: Obtain user characteristic information, which is the eye movement trajectory and behavioral information recorded when the user watches cloud video advertisements; Based on user characteristic information and data preprocessing, user gaze focus data and behavioral data are obtained; Obtain cloud video ad information; based on information recognition, obtain cloud video ad attribute information. Based on gaze focus data and cloud video ad attribute information, and using eye tracking and target detection, key elements of user preferences in the cloud video ad attribute information are obtained. Based on key element information of user preferences and attribute information of cloud video ads, and using a deep learning model, the initial cloud video ads are obtained. The initial cloud video ads are the most matching ads in the cloud video ad pool obtained by the deep learning model by matching cloud video ad attribute information and key element information of user preferences. Acquire eye movement and behavioral information of users when they watch the initial cloud video ads, and update user gaze focus data and behavioral data; Acquire multi-source data and construct a behavioral semantic knowledge graph based on the multi-source data, wherein the multi-source data includes eye-tracking trajectories, behavioral data, item attributes, and behavioral semantic information; Based on the behavioral semantic knowledge graph and a deep learning model, the delivery of cloud video ads is adjusted.

2. The intelligent cloud video advertising delivery management method based on information recognition according to claim 1, characterized in that, The step of obtaining key elements of user preferences from the cloud video ad attribute information based on gaze focus data and cloud video ad attribute information, using eye tracking and object detection, specifically includes: Based on cloud video advertising information, a cloud video advertising image set is obtained through image preprocessing, wherein the image preprocessing includes video segmentation and image noise reduction. Based on the cloud video ad image set and information recognition, cloud video ad attribute information is obtained. The cloud video ad attribute information includes the item name information contained in the cloud video ad, the bounding box coordinate information of the area containing the item, and the brand information. Based on user characteristic information and data preprocessing, user gaze focus data is obtained. The user gaze focus data includes the gaze duration and frequency of the object area and the vertical distance between the upper and lower eyelids. Based on user gaze focus data, obtain the user preference index; Based on user preference index and cloud video ad attribute information, and using eye tracking and target detection, key element information of user preferences is obtained. The key element information includes a first key element label and a second key element label.

3. The intelligent cloud video advertising delivery management method based on information recognition according to claim 2, characterized in that, The process of obtaining a user preference index based on user gaze focus data specifically includes: Based on user gaze focus data, obtain the object gaze density; Specifically, the object gaze density is as follows: ； In the formula, Let i be the gaze density of the i-th item. Let be the duration of gaze on the i-th item. Total watch time for cloud video ads This represents the number of times the i-th item is viewed for more than 1 second. The total number of times all items were looked at for more than 1 second; Based on the vertical distance between the upper and lower eyelids, obtain the user's initial interocular distance and the average interocular distance of the object being gazed upon; The user preference index is obtained based on the user's initial interocular distance, the average interocular distance of the object being gazed at, and the gaze density. Specifically, the user preference index is: ； In the formula, Let be the user preference index for the i-th item, controlling for sensitivity to interocular distance. is the ratio of the average interocular distance for the i-th object being gazed at to the user's initial interocular distance.

4. The intelligent cloud video advertising delivery management method based on information recognition according to claim 1, characterized in that, The process of obtaining key elements of user preferences based on user preference index and cloud video ad attribute information, using eye tracking and object detection, specifically includes: Obtain the bounding box coordinates of the top five items in the user preference index, and use the area within the bounding box as the first area of ​​user gaze. Based on the user's first gaze area, obtain the first key user preference tag, which is the item name information and brand information within the user's first gaze area; Based on the user's gaze focus data in the first region, the user's deep preference region is obtained based on hierarchical clustering and contour extraction. The user's deep preference region is the region obtained by dividing the user's gaze focus into different clusters based on hierarchical clustering and then extracting the contour of each cluster to generate a closed boundary. Based on the user's deep preference region and the item gaze density, the user gaze second region is obtained. The user gaze second region is the top five regions with the highest item gaze density within the user's deep preference region. Based on the user's gaze over the second region, and using object detection, a second key user preference label is obtained. The second key user preference label is the information about the names of items in the second area of ​​the user's gaze obtained from object detection.

5. The intelligent cloud video advertising delivery management method based on information recognition according to claim 1, characterized in that, The cloud video advertising pool specifically includes: Based on user gaze focus data and behavioral data, and using a hybrid expert model, user preference scores are obtained. Based on user preference ratings, core user preference tags are obtained using feature aggregation methods. Based on the cloud video ad attribute information and user core preference tags, and using multi-dimensional similarity fusion, a comprehensive matching score for the cloud video ad is obtained. If the comprehensive matching score is greater than a preset threshold, the cloud video ad is added to the cloud video ad pool. Based on user preference scores, ads with the lowest user preference scores in the cloud video ad pool are eliminated; Obtain historical user gaze focus data, and based on the historical user gaze focus data, obtain the first preference label, which is the item name information and brand information corresponding to the item with the largest fluctuation in the number of gazes; Based on user preference scores and core user preference tags, a second preference tag is obtained using a prediction model; Based on big data analysis, keywords of current trending events are obtained as third-preference tags; The frequency fluctuation range is used as the weight of the first preference label; Based on third-party preference tags, obtain keywords for trending events; Based on key element information of user preferences and keywords of trending events, obtain the frequency of corresponding keywords of trending events; The ratio of the number of times keywords related to trending events appear to the number of times trending events occur is used as the heat value correction coefficient. Based on the key element information of user preferences and the second preference label, obtain the intersection label value, where the intersection label value is the total number of item name information and brand information that are the same as the key element information of user preferences in the second preference label; The calorific value correction factor is adjusted by the ratio of the intersecting label values ​​to the total number of item name and product information contained in the second preference label; The weight of the first preference label is adjusted based on the calorific value correction coefficient.

6. The intelligent delivery management method for cloud video advertising based on information recognition according to claim 4, characterized in that, The adjustment of cloud video ad delivery based on behavioral semantic knowledge graphs and deep learning models specifically includes: Based on user preference ratings and a user clustering algorithm, similar user groups are obtained. These similar user groups are those whose user preference ratings for the initially delivered cloud video ads differ from the user ratings of the users in the initial ad campaign by less than 0.

1. Based on similar user groups and using a preference advertising mining method, a high-value cloud video ad set is obtained. The high-value cloud video ad set is the set of cloud video ads corresponding to user preference scores of similar user groups for the same cloud video ad that are higher than a preset threshold. Based on the updated user gaze focus data, and using eye tracking, the user's real-time gaze intersection point is obtained; Based on the updated behavioral data, a real-time interaction intent vector is obtained using a behavior encoding network. Based on the user's real-time gaze intersection and real-time interaction intent vector, real-time preference tags are obtained using a behavioral semantic knowledge graph. Based on real-time preference tags and a hierarchical filtering method, cloud video ads that meet the filtering criteria in the cloud video ad pool are added to the high-value cloud video ad set. The filtering criteria are item name information and brand information. Based on real-time preference tags and core user preference tags, cloud video ads are retrieved from high-value cloud video ad sets using a deep learning model.

7. A cloud video advertising intelligent delivery management system based on information recognition, used to implement the construction method as described in any one of claims 1-6, characterized in that, include: The main control module is used to obtain core preference tags of a group based on user gaze focus data and behavior data; obtain a high-value cloud video ad set based on preference ad mining methods according to similar user groups; obtain the real-time gaze intersection of the user based on eye tracking based on user gaze focus data; obtain the real-time interaction intent vector based on behavior encoding network based on behavior data; obtain the real-time preference tag based on behavior semantic knowledge graph based on the real-time gaze intersection and real-time interaction intent vector; and obtain the cloud video ads to be delivered from the high-value cloud video ad set based on the real-time preference tag and the user's core preference tag and a deep learning model. The information processing module is used to obtain a cloud video advertisement image set based on image preprocessing according to cloud video advertisement information; obtain cloud video advertisement attribute information based on information recognition according to the cloud video advertisement image set; obtain user gaze focus data based on data preprocessing according to user feature information; obtain user preference index based on user gaze focus data; and obtain key element information of user preference based on eye tracking and target detection according to user preference index and cloud video advertisement attribute information. The cloud video ad pool module is used to obtain a comprehensive matching score for cloud video ads based on user gaze focus data and behavioral data. If the comprehensive matching score is greater than a preset threshold, the cloud video ad information is added to the cloud video ad pool. Based on user preference scores, the ad with the lowest user preference score in the cloud video ad pool is eliminated. Based on historical user gaze focus data, a first preference tag is obtained. Based on user preference scores and core user preference tags, a second preference tag is obtained. Based on big data analysis, keywords of current hot events are obtained as a third preference tag. Based on the second and third preference tags, the weight of the first preference tag is adjusted. Based on the first, second, and third preference tags, the cloud video ad pool is updated.

8. The cloud video advertising intelligent delivery management system based on information recognition according to claim 7, characterized in that, The main control module specifically includes: High-value advertising units are used to obtain core preference tags of a group based on user gaze focus data and behavioral data, and to obtain a set of high-value cloud video ads based on similar user groups and preference ad mining methods. The real-time preference unit is used to obtain the user's real-time gaze intersection based on eye tracking data, obtain the real-time interaction intent vector based on behavior data and behavior encoding network, and obtain the real-time preference label based on the user's real-time gaze intersection and real-time interaction intent vector and behavior semantic knowledge graph. An advertising delivery unit is used to obtain cloud video ads to be delivered from a high-value cloud video ad set based on a deep learning model, according to real-time preference tags and user core preference tags.

9. The cloud video advertising intelligent delivery management system based on information recognition according to claim 7, characterized in that, The information processing module specifically includes: The information acquisition unit is used to acquire a cloud video advertisement image set based on image preprocessing according to cloud video advertisement information, acquire cloud video advertisement attribute information based on information recognition according to the cloud video advertisement image set, and acquire user gaze focus data based on user feature information and data preprocessing. The information processing unit is used to obtain a user preference index based on user gaze focus data, and to obtain key element information of user preference based on eye tracking and target detection, using the user preference index and cloud video advertising attribute information.

10. A cloud video advertising intelligent delivery management system based on information recognition as described in claim 7 The system is characterized by, The cloud video advertising pool module specifically includes: The update unit is used to obtain the comprehensive matching score of the new cloud video advertisement based on the user's gaze focus data and behavior data. If the comprehensive matching score is greater than a preset threshold, the cloud video advertisement information is added to the cloud video advertisement pool. An elimination unit is used to eliminate the advertisement with the lowest user preference score in the cloud video advertisement pool based on the user preference score. The preference unit is used to obtain a first preference tag based on historical user gaze focus data, obtain a second preference tag based on user preference rating and user core preference tag, obtain keywords of current hot events as a third preference tag based on big data analysis, and adjust the weight of the first preference tag based on the second preference tag and the third preference tag.

Images