On-Device Social Grouping via Local Multimodal Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional social grouping techniques often result in low accuracy due to a global view of social connections, overshadowing individual connections and raising privacy concerns by analyzing user data stored in external systems, which can lead to privacy violations and data breaches.
Innovation Solution
A system and method for user-centric social grouping based on spatial-temporal information and communication modalities, where attributes associated with interactions between a user and their contacts are determined and organized into groups using multimodal clustering, ensuring privacy and accuracy by processing interaction data on a user's device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional techniques analyze user data in external systems to identify social connections, then social grouping can be performed, but privacy risks increase due to data exposure and potential breaches
Solution Approach 1:
The patent extracts the social grouping functionality from external centralized systems and relocates it to the user's local device. By processing interaction data locally on the user's mobile device rather than uploading to external servers, the system eliminates data exposure risks while preserving the ability to perform social connection analysis.
Solution Approach 2:
The patent introduces an intermediary layer of local processing between data collection and analysis. Instead of direct analysis in external systems, the interaction data is first processed locally on the user's device through a mediator application that performs clustering algorithms, then only aggregated or anonymized results are shared externally if needed.
2Quantity of substance
If a global view of social connections is used for grouping, then comprehensive analysis is achieved, but grouping accuracy decreases due to overshadowing of individual connections
Solution Approach 1:
The patent segments the social connection analysis into hierarchical levels: individual connection analysis (pairwise interactions between user and each contact), small-group analysis (clusters of closely connected contacts), and overall network structure. This segmentation allows individual connections to be analyzed with high precision before being integrated into broader groupings, preventing them from being overshadowed by global patterns.
Solution Approach 2:
The patent applies different analysis qualities at different levels of the social network hierarchy. Individual connections receive detailed, high-resolution analysis using multiple interaction attributes (call frequency, text messaging patterns, location co-occurrence), while group-level analysis uses aggregated metrics. This local quality approach ensures that individual connection nuances are preserved in the final grouping results.
3Extent of automation
If interaction data is processed externally to determine social groups, then centralized control is achieved, but data security and privacy protection are compromised
Solution Approach 1:
The patent inverts the traditional centralized processing model by making the user's local device the primary processing unit. Instead of external systems collecting and analyzing all interaction data, the inversion places the analytical engine on the user's device, where only the user has access to the raw interaction data. This reverses the security risk model from external threats to local control.
Solution Approach 2:
The patent enables the user's device to perform social grouping analysis independently without requiring external processing infrastructure. The device itself gathers interaction data from its sensors and communication interfaces, processes this data through local clustering algorithms, and generates social groupings autonomously, making the system self-sufficient and eliminating the need for vulnerable centralized data repositories.
Data Source
AI summary
Methods, systems, and computer readable media for social grouping are provided to perform social grouping of a user's contacts based on the user's interactions with the contacts. A set of attributes associated with interactions between a user and a set of contacts may be determined by a first device. The set of attributes associated with the interactions may be related to the first device. The set of contacts may be organized into a set of groups based on the set of attributes.