A Neo4j knowledge graph with world's journal and patent data on 5 billion nodes and 50 bio edges is built for local and global traversals on tasks such as author and organization disambiguation, executed via a java-embedded neo4j-enterprise server, built with maven, and deployed as a docker build.
Janez Ales, PhD
Mathematician, Graph Algorithms Research Scientist, Knowledge Architecture&Innovation, BASF
Author's love for all graphs goes back to the late 80's when he implemented his first graph back-end for a successful heuristic for Hamilton cycles on cubic graphs. All three of his degrees encompass algorithmic graph theory and complexity. He currently works at BASF on a world’s journal and patent graph of 5 billion nodes and 50 billion edges, with emphasis on local and global traversals, for example entity disambiguation. Development pipeline includes python pre-processing with a reverse lookup for 5 billion nodes, a 7Tb neostore, and a java server code with an embedded neo4j-enterprise, for an efficient interplay of dense and sparse traversals, enabling real time local traversals and global off-line analysis of a graph on 55 billion entities. His past work includes linear and combinatorial optimization, ML, and AI projects.