Graph Neural Network Subgraph Personalization for Enterprise Event Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems for employee engagement in a digital working environment are inefficient in terms of resource utilization and fail to provide personalized, explainable recommendations that reflect current employee interests and needs.
Innovation Solution
A recommendation system utilizing a graph neural network (GNN) and a large language model (LLM) to generate user-specific recommendations for events within an enterprise, by processing user and event data to update sub-GNNs and provide personalized recommendations with explanations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional analytics-based tools are used for tracking employee engagement, then data can be collected from surveys and feedback channels, but the analytics remain static and retrospective, unable to reflect current engagement status or predict future trends
Solution Approach 1:
The system performs preliminary actions by continuously monitoring employee communications and behaviors in real-time, maintaining an updated profile of employee interests and engagement status. This allows the system to predict future engagement trends and proactively recommend events before engagement issues arise, rather than waiting for retrospective survey data.
Solution Approach 2:
The system implements continuous feedback loops by analyzing employee responses to recommended events and updating the employee engagement model in real-time. This feedback mechanism allows the system to adapt to changing employee interests and maintain accurate engagement measurements, resolving the contradiction between measurement precision and timeliness.
2Adaptability or versatility
If comprehensive employee data is processed to generate personalized recommendations, then recommendation accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The system segments the large-scale employee data processing into modular components: event embedding generation, employee profile maintenance, compatibility scoring, and recommendation generation. Each component processes only the necessary data subset, allowing personalized recommendations to be generated efficiently without processing the entire employee database for each recommendation query.
Solution Approach 2:
The system applies local quality by maintaining detailed, personalized employee engagement profiles that capture individual interests, preferences, and communication patterns. Rather than applying uniform processing to all employees, the system tailors the analysis depth and data processing scope to each employee's specific profile, optimizing resource usage while maintaining high personalization quality.
3Loss of information
If real-time analysis of employee communications is performed to identify relevant events, then recommendation relevance improves, but system complexity and processing requirements increase
Solution Approach 1:
The system introduces event embedding vectors as an intermediary representation that captures the essential characteristics of events in a standardized format. These embeddings serve as a bridge between diverse event data sources and employee profiles, enabling real-time relevance assessment without requiring complex analysis of raw communication data for each recommendation query.
Solution Approach 2:
The system transforms unstructured employee communication data into structured parameters such as interest categories, engagement levels, and communication patterns. By changing the data parameters from raw text to structured features, the system enables efficient real-time analysis and event matching while reducing the complexity of processing unstructured communication data.
Data Source
AI summary
Methods, systems, and computer-readable storage media for receiving a communication, aggregating user data and event data, the user data representative of a user that as an addressee of the communication, the event data representative of an event described in the communication, determining, from a GNN, a sub-GNN that is specific to the user, the GNN including a data structure that represents users of an enterprise and relationships between users, the sub-GNN representing a portion of the GNN, providing an updated sub-GNN based on the event data, generating a recommendation regarding the event using the updated sub-GNN, the recommendation being specific to the user, providing reason text from a LLM responsive to a prompt, and transmitting a notification to the user, the notification including the recommendation and the reason text.