Box Graph: how we built a spontaneous social network

Box has developed an innovative framework called Box Skills that integrates machine learning with content stored in Box to derive value from exponentially growing digital content, exemplified by features like video intelligence for transcription and facial recognition. To extend this capability, Box has introduced Box Graph, a machine learning model designed to understand and map relationships between content and users within an organization, forming a dynamic, real-time social network that reflects user interactions and collaborations. Box Graph computes collaboration scores in real time, scales to manage billions of interactions, and uses a horizontally scalable, versioned graph system supported by the Score Keeper Service (SKS) and Message Queue Service (MQS) for efficient data processing and recommendations. The system employs collaborative filtering, leveraging mathematical techniques like LogSumExp and LogDiffExp, to update scores incrementally and prioritize relevant content and collaborators. By utilizing backtesting and A/B testing, Box fine-tunes algorithmic parameters to tailor its recommendations for specific enterprise needs, facilitating efficient, real-time collaboration insights.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.