Personalized Content Recommendation System and Method for Illustration and Character Design Platform
Patent Information
- Authority / Receiving Office
- KR · KR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-02-05
- Publication Date
- 2026-08-12
Smart Images
Figure PAT00001_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to the field of providing user-customized content by utilizing artificial intelligence (AI) and recommendation system technologies. More specifically, it relates to a system and method for recommending personalized content by analyzing user preferences and behavioral patterns in a digital illustration and character design platform.
[0002] With the recent development of online platforms, a massive amount of digital content is being produced, and users desire to effectively find and consume content that suits their tastes. In particular, in the fields of illustration and character design, there is a rising demand for customized content that aligns with users' individuality and preferences.
[0003] Accordingly, there is a growing need for a system that recommends personalized content by analyzing user data based on artificial intelligence technology. In response to this technical requirement, the present invention aims to enhance the user experience and improve the competitiveness of the platform by providing an intelligent recommendation system that matches user profiles with a content database.
[0004] Therefore, the technical field to which the present invention belongs can be described as a convergence technology area encompassing artificial intelligence, recommendation systems, user-customized content provision, digital illustration, and character design platforms. Background Technology
[0005] With the advancement of digital content platforms, the demand for personalized content recommendation technology is increasing. Particularly in the fields of illustration and character design, the provision of customized content that reflects users' individuality and preferences is emerging as a key factor determining the competitiveness of platforms.
[0006] Existing content recommendation systems primarily utilized collaborative filtering techniques to analyze user similarity and provided recommendations based on the content consumption patterns of users with similar preferences. However, this approach had limitations, such as difficulty in precisely reflecting an individual's unique characteristics and preferences, and a decline in recommendation accuracy for new users or content with low popularity.
[0007] Recently, driven by advancements in artificial intelligence technology, sophisticated recommendation systems utilizing deep learning and natural language processing techniques are being proposed. These systems can provide more precise personalized recommendations by comprehensively analyzing various data, such as user profile information, content consumption history, and real-time feedback.
[0008] However, intelligent recommendation systems specialized in the fields of illustration and character design have not yet been sufficiently developed, and there is a growing demand for technology that can effectively analyze and reflect users' aesthetic preferences and design tastes.
[0009] Furthermore, since existing recommendation systems primarily relied on standardized metadata to identify content characteristics, they were difficult to apply to content where visual and subjective characteristics are important, such as illustrations or character designs.
[0010] Accordingly, the present invention aims to provide an advanced customized recommendation system by utilizing artificial intelligence technology to deeply analyze user preferences and the characteristics of illustration and character design content. The technology underlying this invention encompasses user profiling, image and text analysis, and AI-based recommendation algorithms, and it is expected that this will further enhance the user experience on illustration and character design platforms. Prior art literature
[0011] Patent Document 1: KR Patent No. 10-2021-0052276 Patent Document 2: US Patent No. 10,102,559 Patent Document 3: US Patent No. 9,852,444 Patent Document 4: US Patent No. 8,560,545
[0012] Non-patent Document 1: Ricci, Francesco, Lior Rokach, and Bracha Shapira. “Recommender systems: introduction and challenges.” Recommender systems handbook. Springer, Boston, MA, 2015. 1-34. Non-patent Document 2: He, Xiangnan, et al. “Neural collaborative filtering.” Proceedings of the 26th international conference on world wide web. 2017. The problem to be solved
[0013] The present invention aims to solve the following technical problems.
[0014] First, the present invention aims to provide a customized recommendation system capable of effectively reflecting the characteristics of illustration and character design content. Existing recommendation systems rely primarily on text-based metadata or user rating information to identify content characteristics, which has limitations in their application to the fields of illustration and character design, where visual and subjective characteristics are critical. Accordingly, the present invention utilizes artificial intelligence technology to deeply analyze the visual and semantic characteristics of illustration and character design content and matches them with user profiles to provide advanced customized recommendations.
[0015] Second, the present invention aims to implement a recommendation system capable of accurately identifying and reflecting users' personalized preferences and needs. Existing recommendation systems primarily provide recommendations based on similarities among user groups, making it difficult to finely reflect an individual's unique tastes and needs. The present invention seeks to improve user satisfaction and engagement by providing personalized recommendations through the comprehensive analysis of various data, such as user profile information, content consumption history, and real-time feedback.
[0016] Third, the present invention aims to secure the comprehensiveness and diversity of recommendation systems by solving the cold start and long-tail problems. Existing recommendation systems had limitations in that the accuracy of recommendations was low for new users or content with low popularity. By utilizing artificial intelligence technology to identify the inherent characteristics of content and predict users' potential preferences, the present invention can provide effective recommendations even in situations where sufficient data has not been accumulated.
[0017] Fourth, the present invention aims to enhance user trust by improving the explainability of recommendation results. Since existing recommendation systems derive recommendation results using complex algorithms, it has often been difficult for users to understand the basis for the recommendations. The present invention aims to increase the transparency of recommendations and enhance user trust by presenting the reasons for the recommendations to the user along with the results. Through this, it is expected that the usability of the recommendation system and user satisfaction can be improved.
[0018] By solving the above technical problems, the present invention is expected to innovate the user experience in illustration and character design platforms and ultimately contribute to the development of the relevant industry. means of solving the problem
[0019] The present invention provides the following configuration to solve the above technical problem.
[0020] A user-customized content recommendation system in an illustration and character design platform according to one embodiment of the present invention comprises: a data collection unit that collects user activity data within the platform and stores it in a database; a preference analysis unit that analyzes user preferences based on the collected activity data and generates a preference profile; a content feature extraction unit that extracts and classifies visual features of content within the platform through artificial intelligence-based image analysis; a recommendation engine unit that generates a personalized recommended content list by combining user preference profiles and content feature information; and a content provision unit that provides the generated recommended content through a user-customized interface.
[0021] The data collection unit collects users' content viewing history and dwell time, content saving, download, and purchase history, history of likes, comments, and sharing activities, search keywords and browsing patterns, content creation and upload history, and information on interactions with other users. Through this, it secures foundational data that enables a comprehensive analysis of users' activities within the platform.
[0022] The preference analysis unit identifies user groups with similar preferences using a deep learning-based collaborative filtering algorithm, analyzes the visual and semantic characteristics of content preferred by users through content-based filtering, and tracks patterns of change in user preferences through time-series analysis. Through this multifaceted analysis, users' tastes and needs can be accurately identified.
[0023] The content feature extraction unit classifies the art styles of illustrations and characters, analyzes major color palettes and color harmony, measures composition, layout, and visual balance, identifies applied techniques such as line art, coloring, and texture, analyzes character facial expressions, poses, and emotional expressions, classifies the genre and theme of the work, and analyzes the artist's unique stylistic characteristics to store them as metadata. This enables the systematic classification and management of the visual characteristics of the content.
[0024] The recommendation engine utilizes a deep neural network-based machine learning model to learn changes in users' short- and long-term preferences, recommends content tailored to the user's current situation and purpose through context-aware algorithms, and continuously improves recommendation accuracy through reinforcement learning. Through this intelligent recommendation system, content optimized for the user can be provided.
[0025] The present invention also includes a feedback processing unit that collects explicit and implicit feedback from users and reflects it in a recommendation system in real time. The feedback processing unit collects and analyzes a 5-point scale satisfaction evaluation of recommended content, the display of content of no interest and selection of reasons, responses agreeing or disagreeing with the reasons for recommendation, an evaluation of preferences by feature of recommended content, and user suggestions for improving the recommendation system.
[0026] The content delivery unit converts and delivers content in a form optimized for the user's device characteristics and network environment, dynamically loads content by tracking the user's scroll position and areas of interest in real time, and presents the reason for the recommendation and related content for each recommended item. Through this, it is possible to provide the user with an optimized content consumption experience.
[0027] The present invention additionally includes a creator support unit that analyzes activity data of content creators within a platform to identify trends and recommends directions for new content production based on these trends. The creator support unit provides functions for analyzing popular styles, themes, and techniques; forecasting user demand and identifying niche markets; suggesting differentiation points from similar works; and identifying potential target user groups.
[0028] In addition, it includes a timing optimization unit that analyzes users' content consumption patterns to determine the optimal recommendation timing, which provides functions such as analyzing users' platform access times, identifying preferred consumption times by content type, optimizing the exposure frequency of recommended content, and adjusting the timing of push notification transmission.
[0029] In addition, it includes a Community Management Department that forms communities based on common interests among multiple users and recommends group-customized content tailored to the characteristics of each community. The Community Management Department provides features such as interest-based user grouping, analysis of popular content by community, filtering of recommended content based on group activity, and adjustment of recommendation weights based on community participation.
[0030] Finally, including a performance evaluation unit that monitors and evaluates the performance of the recommendation system, it provides functions for measuring recommendation accuracy and variety, tracking user satisfaction indicators, comparing algorithms through A / B testing, monitoring system response time and resource usage, and identifying and applying improvement points.
[0031] Through this configuration, the present invention can effectively recommend user-customized content on an illustration and character design platform, and can significantly improve user satisfaction and the usability of the platform. Effects of the invention
[0032] The user-customized content recommendation system in the illustration and character design platform of the present invention provides the following effects.
[0033] First, the present invention can automatically extract and classify the visual characteristics of illustration and character design content through AI-based image analysis technology. By systematically analyzing various visual elements such as art style, color palette, composition, and emotional expression, it overcomes the limitations of relying on existing text-based metadata and enables a more accurate identification of content characteristics. Through this, customized content that matches the user's visual tastes and preferences can be effectively recommended.
[0034] Second, the present invention constructs a personalized preference profile by collecting and analyzing various user activity data. By comprehensively analyzing not only explicit activities such as viewing, saving, and sharing content, but also implicit behavioral data such as dwell time and scrolling patterns, the user's tastes and needs can be identified more accurately. This significantly contributes to increasing the accuracy and relevance of recommendations.
[0035] Third, the deep neural network-based recommendation engine of the present invention continuously learns and reflects changes in the user's short-term and long-term preferences. Through context-aware algorithms and reinforcement learning, it can provide recommendations optimized for the user's current situation and purpose, and the system's performance improves over time.
[0036] Fourth, the present invention enhances the transparency of the system and user trust by providing explainability for recommendation results. By presenting the reason for each recommended content and showing related content together, user understanding and acceptance can be improved.
[0037] Fifth, the creator support function of the present invention provides useful insights to content creators. By suggesting effective content production directions through trend analysis and demand forecasting, it can contribute to improving the quality and diversity of content within the platform.
[0038] Sixth, the community management function of the present invention promotes interaction among users with similar interests and supports the formation of active communities within the platform through group-customized content recommendations. This leads to the activation of the platform and improved user loyalty.
[0039] Seventh, the timing optimization function of the present invention can enhance the effectiveness of recommendations and user acceptance by analyzing the user's content consumption patterns and providing recommended content at the optimal time.
[0040] Eighth, the present invention can continuously monitor and evaluate the performance of the system to identify and apply areas for improvement. The quality of the system can be continuously improved through algorithm comparison via A / B testing and tracking of performance indicators.
[0041] Consequently, the present invention can significantly improve the user experience in illustration and character design platforms and provide valuable services to both content creators and consumers. This is expected to contribute positively to strengthening the competitiveness of the platform and promoting industrial development. Brief explanation of the drawing
[0042] Fig. 1: Overall configuration diagram of the user-customized content recommendation system for an illustration and character design platform Fig. 2: Activity of the data collection unit, data collection and processing system configuration diagram Fig. 3: Structure diagram of the deep learning-based collaborative filtering and content-based filtering system of the preference analysis unit Fig. 4: Configuration diagram of the image analysis and metadata generation system of the content feature extraction unit Fig. 5: Architecture of a deep neural network-based personalized recommendation system for the recommendation engine Fig. 6: Structure diagram of the real-time user feedback collection and analysis system of the feedback processing unit Fig. 7: Configuration diagram of the Creator Support Department's trend analysis and content creation recommendation system Fig. 8: Structure diagram of the user pattern analysis and recommended timing determination system of the timing optimization unit Fig. 9: Architecture of the Community Management Department's group-customized content recommendation system Fig. 10: Configuration diagram of the system monitoring and performance improvement process of the Performance Evaluation Department Specific details for implementing the invention
[0043] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.
[0044] Referring to FIG. 1, a user-customized content recommendation system (100) in an illustration and character design platform according to an embodiment of the present invention includes a data collection unit (110), a preference analysis unit (120), a content feature extraction unit (130), a recommendation engine unit (140), a content provision unit (150), a feedback processing unit (160), a creator support unit (170), a timing optimization unit (180), a community management unit (190), and a performance evaluation unit (200).
[0045] As illustrated in FIG. 2, the data collection unit (110) performs the role of collecting and storing all activity data within the user's platform. Specifically, it includes a content viewing history collection module (111), a save and purchase history collection module (112), an interaction history collection module (113), a search pattern analysis module (114), and a dwell time measurement module (115). Data collected from each module is stored in an activity data database (116) and processed into an analyzable form through a data preprocessing module (117).
[0046] The preference analysis unit (120) is configured as shown in FIG. 3 and includes a deep learning-based collaborative filtering engine (121) and a content-based filtering engine (122). The collaborative filtering engine (121) generates a user-content interaction matrix and identifies similar user groups by extracting potential factors based on it. The content-based filtering engine (122) generates a preference profile by analyzing the characteristics of the content preferred by the user. The time series analysis module (123) tracks and predicts changes in the user's preference.
[0047] As illustrated in FIG. 4, the content feature extraction unit (130) automatically extracts and analyzes visual features of illustrations and character designs using computer vision technology. It includes a style analysis module (131), a color analysis module (132), a composition analysis module (133), a technique identification module (134), and an emotion expression analysis module (135), and the analysis results of each module are converted into integrated feature vectors and stored in a metadata database (136).
[0048] The recommendation engine unit (140) is configured as shown in FIG. 5 and operates based on a deep neural network-based personalized recommendation model (141). This model receives a user preference profile and a content feature vector as input and calculates a recommendation score. A context recognition module (142) adjusts recommendations by considering the user's current situation, and a reinforcement learning module (143) continuously optimizes the model based on user feedback.
[0049] The feedback processing unit (160) is a system that collects and processes explicit / implicit feedback from users, as illustrated in FIG. 6. It includes a satisfaction evaluation collection module (161), a non-preferred content analysis module (162), and a recommendation reason evaluation module (163), and immediately reflects the collected data into the recommendation system through a real-time feedback processing engine (164).
[0050] The creator support unit (170) is configured as shown in FIG. 7, identifies content trends within the platform through a trend analysis engine (171), and predicts future popular content through a demand forecasting module (172). The differentiation strategy proposal module (173) suggests the optimal content production direction to the creator based on the analysis results.
[0051] As illustrated in FIG. 8, the timing optimization unit (180) includes a time-based activity analysis module (181) that analyzes the user's platform usage pattern, a content consumption pattern analysis module (182), and an exposure frequency optimization module (183). Through this, recommended content can be provided to each user at an optimal time.
[0052] The community management unit (190) is configured as shown in FIG. 9, connects users with common interests through a user grouping engine (191), and identifies the characteristics of the community through a group activity analysis module (192). A group-specific recommendation module (193) selects and provides content suitable for the characteristics of each community.
[0053] The performance evaluation unit (200) measures and analyzes various performance indicators of the recommendation system as illustrated in FIG. 10. It includes an accuracy measurement module (201), a diversity evaluation module (202), an A / B test management module (203), and a system monitoring module (204), and supports continuous improvement of the system based on the analysis results.
[0054] The above components are organically connected and operate together, and the processing results of each component are stored in a central database to contribute to the improvement of the overall system performance. The system of the present invention is designed with scalability in mind and has a modular structure that facilitates the addition of new functions or algorithms. Explanation of the symbols
[0055] 100: Personalized Content Recommendation System 110: Data Collection Unit 120: Preference Analysis Department 130: Content Feature Extraction Unit 140: Recommended Engine Section 150: Content Provider 160: Feedback processing unit 170: Creator Support Department 180: Timing Optimization Section 190: Community Management Department 200: Performance Evaluation Department 110: Data Collection Unit 111: Content View History Collection Module 112: Save and Purchase History Collection Module 113: Interaction History Collection Module 114: Search Pattern Analysis Module 115: Residence time measurement module 116: Activity Data Database 117: Data Preprocessing Module 120: Preference Analysis Department 121: Collaborative Filtering Engine 121a: User-Content Matrix Generator 121b: Latent Factor Extractor 122: Content-based filtering engine 122a: Content Feature Analyzer 122b: Preference Profile Generator 123: Time Series Analysis Module 123a: Pattern detector 123b: Change Prediction System 130: Content Feature Extraction Unit 131: Style Analysis Module (Art Style Analysis of Illustrations and Characters) 132: Color Analysis Module (Major Color Palette and Color Harmony Analysis) 133: Composition Analysis Module (Layout and Visual Balance Measurement) 134: Technique Identification Module (Identifies applied techniques such as line art, coloring, texture, etc.) 135: Emotion Expression Analysis Module (Character Facial Expression, Pose, and Emotion Expression Analysis) 136: Metadata Database 140: Recommended Engine Section 141: Personalized Recommendation Model 141a: Input Handler (Preprocessing of User Preferences and Content Features Data) 141b: Deep Neural Networks (Core Recommendation Algorithm Processing) 141c: Output Generator (Generate Recommended Results) 142: Context Awareness Module (Analysis of user's current situation and purpose. Reflected in recommendations) 142a: Situation Analyzer 142b: Adaptive filter 143: Reinforcement Learning Module (Continuously optimizing model performance based on user feedback) 143a: Compensation Calculator 143b: Policy Optimizer 160: Feedback processing unit 161: Satisfaction Evaluation Collection Module 162: Non-preferred Content Analysis Module 163: Recommendation Reason Evaluation Module 164: Real-time feedback processing engine 170: Creator Support Department 171: Trend Analysis Engine 172: Demand Forecasting Module 173: Differentiation Strategy Proposal Module 180: Timing Optimization Section 181: Activity Analysis Module by Time of Day 182: Content Consumption Pattern Analysis Module 183: Exposure Frequency Optimization Module 190: Community Management Department 191: User Grouping Engine 192: Group Activity Analysis Module 193: Group-based recommendation module 200: Performance Evaluation Department 201: Accuracy Measurement Module 202: Diversity Assessment Module 203: A / B Test Management Module 204: System Monitoring Module
Claims
Claim 1 A user-customized content recommendation system in an illustration and character design platform comprises: a data collection unit that collects user activity data within the platform and stores said activity data in a database; a preference analysis unit that analyzes user preferences based on said collected activity data and generates a preference profile; a content feature extraction unit that extracts and classifies visual features of content within the platform through AI-based image analysis; a recommendation engine unit that generates a personalized recommended content list by combining said user preference profile and content feature information; and a content provision unit that provides said generated recommended content through a user-customized interface; wherein the data collection unit is characterized by collecting user content viewing history and dwell time, content saving, download and purchase history, like, comment and sharing activity history, search keywords and browsing patterns, content creation and upload history, and interaction information with other users. Claim 2 A user-customized content recommendation system according to claim 1, wherein the preference analysis unit identifies user groups with similar preferences using a deep learning-based collaborative filtering algorithm, analyzes the visual and semantic characteristics of content preferred by the user through content-based filtering, and tracks the user's preference change patterns through time-series analysis. Claim 3 A user-customized content recommendation system according to claim 1, wherein the content feature extraction unit classifies the art style of illustrations and characters, analyzes the main color palette and color harmony, measures composition, layout and visual balance, identifies applied techniques such as line art, coloring, and texture, analyzes the facial expressions, poses, and emotional expressions of characters, classifies the genre and theme of the work, and analyzes the unique style characteristics of the artist and stores them as metadata. Claim 4 A user-customized content recommendation system according to claim 1, wherein the recommendation engine unit learns changes in the user's short-term and long-term preferences using a deep neural network-based machine learning model, recommends content suitable for the user's current situation and purpose through a context recognition algorithm, and continuously improves recommendation accuracy through reinforcement learning. Claim 5 A user-customized content recommendation system according to claim 1, further comprising a feedback processing unit that collects explicit and implicit feedback from a user and reflects it in a recommendation system in real time, wherein the feedback includes a 5-point scale satisfaction evaluation of recommended content, display of content of no interest and selection of reason, a response of agreement / disagreement regarding the reason for recommendation, an evaluation of preference by characteristic of recommended content, and user suggestions for improving the recommendation system. Claim 6 A user-customized content recommendation system according to claim 1, wherein the content providing unit converts and provides content in a form optimized for the user's device characteristics and network environment, dynamically loads content by tracking the user's scroll position and area of interest in real time, and presents the reason for recommendation and related content together for each recommended content. Claim 7 A user-customized content recommendation system according to claim 1, further comprising a creator support unit that analyzes activity data of content creators within the platform to identify trends and recommends directions for new content production based thereon, wherein the creator support unit provides functions for analyzing popular styles, themes, and techniques, predicting user demand and discovering niche markets, suggesting differentiation points from similar works, and identifying potential target user groups. Claim 8 A user-customized content recommendation system according to claim 1, further comprising a timing optimization unit that determines the optimal recommendation timing by analyzing the user's content consumption pattern, wherein the timing optimization unit provides functions such as analyzing the user's platform access time, identifying preferred consumption times by content type, optimizing the exposure frequency of recommended content, and adjusting the timing of push notification transmission. Claim 9 A user-customized content recommendation system according to claim 1, further comprising a community management unit that forms a community based on common interests among multiple users and recommends group-customized content suitable for the characteristics of the community, wherein the community management unit provides interest-based user grouping, analysis of popular content by community, filtering of recommended content based on group activity, and adjustment of recommendation weights according to community participation. Claim 10 A user-customized content recommendation system according to claim 1, further comprising a performance evaluation unit for monitoring and evaluating the performance of a recommendation system, wherein the performance evaluation unit provides functions for measuring recommendation accuracy and diversity, tracking user satisfaction indicators, comparing algorithms through A / B testing, monitoring system response time and resource usage, and deriving and applying improvement points. Claim 11 A method for recommending personalized content to a user using a system according to any one of claims 1 to 10, comprising: (a) collecting user activity data within a platform and storing it in a database; (b) analyzing user preferences based on the collected activity data and generating a preference profile; (c) extracting and classifying visual features of content within a platform through artificial intelligence-based image analysis; (d) generating a personalized list of recommended content by combining the user preference profile and content feature information; (e) providing the generated recommended content through a user-customized interface; (f) collecting and analyzing user feedback on the recommended content; and (g) continuously improving a recommendation algorithm based on the collected feedback. Claim 12 A user-customized content recommendation method according to claim 11, wherein step (b) comprises: (b-1) a step of identifying a group of users with similar preferences through collaborative filtering; (b-2) a step of analyzing the characteristics of preferred content through content-based filtering; and (b-3) a step of tracking preference change patterns through time series analysis. Claim 13 A user-customized content recommendation method according to claim 11, wherein the above step (c) comprises: (c-1) a step of analyzing visual elements of illustrations and characters; (c-2) a step of classifying the style and technique of the work; (c-3) a step of analyzing the emotional expression of the characters; and (c-4) a step of storing the analyzed features as metadata. Claim 14 A user-customized content recommendation method according to claim 11, wherein step (d) comprises: (d-1) a step of learning changes in user preferences through a machine learning model; (d-2) a step of selecting content suitable for the current situation through a context recognition algorithm; and (d-3) a step of improving recommendation accuracy through reinforcement learning. Claim 15 A user-customized content recommendation method according to claim 11, wherein step (e) comprises: (e-1) a step of converting content into a form optimized for the user device; (e-2) a step of improving the user experience through dynamic content loading; and (e-3) a step of presenting the reason for recommendation and related content together.