Profiling Graphs: Order from Chaos

Abstract

Graphs are being increasingly adopted as a flexible data model in scenarios (e.g., Google's Knowledge Graph, Facebook's Graph API, Wikidata, etc.) where multiple editors are involved in content creation, where the schema is ever changing, where data are incomplete, where the connectivity of resources plays a key rolescenarios where relational models traditionally struggle. But with this flexibility comes a conceptual cost: it can be difficult to summarise and understand, at a high level, the content that a given graph contains. Hence profiling graphs becomes of increasing importance to extract order, a posteriori, from the chaotic processes by which such graphs are generated. This talk will motivate the use of graphs as a data model, abstract recent trends in graph data management, and then turn to the issue of profiling and summarising graphs: what are the goals of such profiling, the principles by which graphs can be summarised, the main techniques by which this can/could be achieved The talk will emphasise the importance of profiling graphs while highlighting a variety of open research questions yet to be tackled.