Image-Based User Affinity Inference for Online Content Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online systems fail to infer user affinities for items or topics when content received from users or interacted with does not include descriptive information, leading to a degradation in user experience due to the inability to present relevant content.
Innovation Solution
An online system updates user profiles to include affinities based on images from content received or interacted with, using a machine-learning model trained on images and attributes to predict probabilities and select relevant content for presentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If online systems rely on descriptive information (tags, captions) to infer user affinities, then affinity inference accuracy is improved, but the system fails to identify affinities when such information is absent
Solution Approach 1:
The patent introduces image data as an intermediary element between users and the online system. Instead of relying solely on user-provided descriptive information (tags, captions), the system uses image processing techniques to automatically extract visual features and infer item identities. This intermediary approach allows the system to determine user affinities even when users do not provide explicit descriptions, thereby resolving the contradiction between accuracy and adaptability.
2Device complexity
If online systems use traditional methods based on user-provided information, then implementation simplicity is maintained, but user experience degrades when content lacks descriptive information
Solution Approach 1:
The patent replaces the manual, user-dependent information provision mechanism with an automated image-based recognition system. Instead of relying on users to mechanically add tags or captions, the system uses computer vision technology to automatically analyze images and extract item information. This substitution maintains implementation feasibility while significantly improving recommendation reliability, as the system can now function effectively regardless of user input quality.
3Measurement precision
If online systems wait for user interaction before inferring affinities, then data accuracy is improved, but response time and user engagement are delayed
Solution Approach 1:
The patent implements preliminary action by inferring user affinities proactively based on image content before explicit user interactions occur. When users upload or share images, the system immediately processes these images to identify items and infer affinities, rather than waiting for users to add tags, captions, or engage in additional interactions. This preliminary processing eliminates delays in presenting relevant content while maintaining accuracy through automated image analysis.
Data Source
AI summary
An online system receives a content item including an image from a content-providing user and/or receives an interaction with the content item from a viewing user. The online system accesses a machine-learning model that is trained based on a set of images of items associated with an entity and attributes of each image. The online system applies the model to predict a probability that the content item includes an image of an item associated with the entity based on attributes of the image included in the content item. Based on the predicted probability, the online system updates a profile of the user (i.e., the content-providing user and/or the viewing user) to include an affinity for the item. Upon determining an opportunity to present content to the user, the online system selects content for presentation to the user based on the profile and sends the content for presentation to the user.