Content providers and analysts alike increasingly rely on combining multiple data sources to build comprehensive, up-to-date and properly interlinked information spaces. These organizations criticallydepend on technologies for integrating these sources and tracking their evolution. DIVINE aims to provide such technologies, with a lightweight seed ontology acting as the focal point for integrating new evidence derived from multiple, evolving data sources. As such, the project advances ontology evolution research characterized by single-source solutions, which exploit mostly textual and rather static data. DIVINE integrates structured, unstructured and social sources. A modu-
lar and scalable portfolio of evidence acquisition services crawls public Web documents, queries Linked Open Data repositories, aggregates resource annotations from Web 2.0 applications, and triggers validation processes for missing or conflicting evidence. Since evidence from third-party sources is inherently uncertain, source-specific transformation rules and impact factors assign a confidence value to each new fact. A spreading activation network utilizes the collected evidence in conjunction with the confidence values for extending the seed ontology.
DIVINE will monitor domain changes over time to derive knowledge evolution patterns. This domain-centric view makes DIVINE novel among existing change detection approaches, which tend to be domain-agnostic. Each ontology element is assigned a confidence matrix, which records the changes in confidence values over time. Data services and dynamic visualizations reveal rising, declining or cyclic patterns in the confidence matrices. Such patterns are important indicators - the rate of change or the date of a concept's first appearance, for example, shed light on the evolution of knowledge and on the underlying processes that drive this evolution.
Austrian Research Promotion Agency