In Phase I of DINLAS (Distributed Knowledge Graph Service), Language Computer Corporation will conduct detailed experiments using both currently available open source distributed processing and graph frameworks and new novel methodologies for the fusion of knowledge graphs to determine what is in the state of possible in regards to very large knowledge graphs. As a part of this study we will define a common service interface and query framework , which will facilitate testing of the scalability, reliability, consistency, performance characteristics, and the bandwidth and storage requirements of major open source graph storage engines and memory-mapped implementations. This will result in being able to select the best storage engines for different operational environments. Additionally, we will test the feasibility of a distributed message queue-based architecture for population of the knowledge graph, which will ensure high availability and fault tolerance, and facilitate processing of any data modality. In order to provide a deduplicated and merged knowledge graph, we will build on our scalable state-of-the-art cross-document coreference system by introducing a novel mechanism which significantly reduces the overall bandwidth requirements. Finally, we propose to test a novel one-way hashing procedure that can optimize trade-offs between querying power and potential information leakage.Knowledge Graph,Data Fusion,One-Way Graph Analytics,Cross-Document Coreference,micro-services,Cloud-Based Architecture,Real-Time Data Transport Layer,Distributed Computing
