Social Network Label Propagation for Interest Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Social networks face challenges in accurately targeting online content, such as advertisements, to users with incomplete or unknown interests and disinterests, as existing methods rely heavily on user-provided information, which may be insufficient or absent, leading to ineffective content delivery.
Innovation Solution
A method that propagates labels representing user interests and disinterests through a social network graph, where labels are associated with user nodes and weights are determined based on similarity or dissimilarity, allowing for the inference of interests and disinterests from related users and groups, even if not explicitly stated, and normalizing these labels for targeted content delivery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user-provided information is used to determine user interests, then the system can target content to users with complete profiles, but users with incomplete or absent information cannot be effectively targeted
Solution Approach 1:
The patent uses social relationships as an intermediary to transfer interest information from users with complete profiles to users with incomplete profiles. By treating friends and connections as mediators, the system can infer interests for users who haven't provided sufficient information themselves, thus resolving the contradiction between measurement precision and adaptability
Solution Approach 2:
The system copies interest information from users with complete profiles and applies it to users with incomplete profiles through the social network graph. This copying mechanism allows the system to extend targeting capabilities to users who lack sufficient profile information, maintaining adaptability while preserving reasonable accuracy
2Adaptability or versatility
If labels are propagated through the entire social network graph, then more users can be characterized, but the influence of distant nodes may reduce accuracy
Solution Approach 1:
The patent applies local quality by making the propagation effect local rather than global. By introducing uncharacterized label nodes and using them to reduce the effect of faraway nodes, the system ensures that label propagation maintains high accuracy for local connections while still providing coverage for broader network participation
Solution Approach 2:
Uncharacterized label nodes serve as intermediaries that control and regulate the flow of label information through the network. These intermediary nodes prevent distant nodes from having excessive influence while still allowing information to propagate through the graph, thus maintaining both coverage and precision
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer-readable storage medium, including a method for propagating labels. The method comprises determining a set of labels to be associated with users of a social network, the labels including one or more designators for specifying areas of interest and areas of disinterest for a user. The method further comprises associating nodes in a graph representing the social network, where the users are represented by user nodes in the graph, and determining that a user is similar or dissimilar to another user in the social network. The method further comprises determining weights for the labels, each weight reflecting a magnitude of a contribution of an associated label to a characterization of the respective node, and propagating labels to other nodes that are related to the respective node by a relationship, including propagating labels in accordance with the determined similarity or dissimilarity.