Real-time Trend and Sentiment Analysis and Tracking with Dynamic Knowledge Graph

This effort proposes to develop a real-time KG-based Trend and Sentiment Analysis (KG-TSA) system with dynamic Knowledge Graph. KG-TSA provides a solution for intelligence analysis where no obvious models or signatures are available to identify potential insider threats, especially when the risk indicators and warning signs are from low-signal-to-noise events and transactions. Specifically, three key components of the system, Dynamic Knowledge Graph for representation and inference, knowledge-aware deep learning for trend and sentiment analysis, and pattern mining for anomaly detection and latent-risk discovery, will be developed. With the help of the knowledge graph, it is expected that the final products could be incorporated and integrated multi-modal information and detect, track the entity’s behavior as well as indicate latent risks. The lambda framework is used in the implementation, which can efficiently integrate the streaming data and the historical data in a scalable way.