Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Knowledge representation for decision-centered visualization

: Kohlhammer, J.
: Encarnacao, J.L.

Herdecke: GCA-Verlag, 2005, 250 pp.
Zugl.: Darmstadt, TU, Diss., 2005
Forschung und Wissen - Informatik
ISBN: 3-89863-197-4
Fraunhofer IGD ()
knowledge representation; decision support; information visualization; human centered computing; time critical visualization

Users of information systems in timecritical domains are under pressure to digest and process information that is vital for their task. Too often analysts receive the important information buried within a collection of insignificant data and information. Sorting through such a cluttered display and finding the critical information requires a high level of attention and becomes even more challenging with the increasing amounts of information available. Especially in timecritical domains, the problem of information assessment is replaced by the problem of avoiding the timeintensive efforts to review all the available information.
The focus of this dissertation was the combination of knowledge representation, information visualization, and human-centered computing to reduce the probability of such information overload, defining the field of decision-centered visualization (DCV). The DCV approach is unique in how it ties together information visualization and knowledge representation in time-critical and information-intensive environments to support the user's situation awareness. The knowledge representation for DCV as it is defined in this thesis corresponds to the conceptual and architectural requirements of DCV to enable decision- centered applications.
The goal of this work was a knowledge representation approach that is user-centered, allows scalable information handling, supports knowledge- based information visualization, flexibly supports various kinds of display and presentation systems, and can be realized in a dynamic network environment - all aspects of information systems typically encountered in time-critical domains like air traffic control.
Part of the solution introduced in this work was the scalable combination of domain ontologies and domain databases for human-centered and knowledge-based information visualization. The domain ontology as the knowledge structuring element of the representation is integrated with the scalability advantages of databases, together representing the visualization-relevant aspects of a domain.
This thesis extends the current information visualization theory by socalled presentation requirements (PRs). PRs allow the abstract analytical description of information, particularly for the support of situation awareness. By describing entity types, their attributes, and relationships between entity types, knowledge about the information in the domain can be combined with knowledge about the necessary information visualization to users under time pressure. This is enabled by using presentation types and data types for the description of information visualization environments, leaning on and extending the current information visualization models with aspects for time-critical and information-intensive domains.
With the help of so-called DCV transformations, this work enables the flexible guidance of various display systems by a DCV system tuning the visualization to the current situation, to the task that the user is working working on, and to the role of the user in the current situation. Depending on the needs of the user and the used display system, filtered and prioritized information can be flexibly mapped to visual structures and presented in an effective and efficient way.
The results of this work are most applicable to the area of emergency management and are also applied to the field of visual analytics, an integrated field between data mining and information visualization in the area of very information-intensive domains. Here, the presentation requirements enable the reduction of the complexity of the information on an abstract level based on the provided semantics, before any data instances have to be considered.