Full Citation
Title: Visualization Techniques for Rule-based Reasoning in Uncertain Knowledge Bases
Citation Type: Dissertation/Thesis
Publication Year: 2010
ISBN:
ISSN:
DOI:
NSFID:
PMCID:
PMID:
Abstract: In recent years, several projects have built large semantic knowledge bases, with the help of information extraction techniques. By applying these techniques to unstructured (or loosely structured) web sites like Wikipedia, the received knowledge bases may contain uncertainty or even inconsistency to some extend. To tackle the problem of this potential data uncertainty and inconsistency, the Max-Planck Institute Saarbrücken has developed URDF. URDF is an efficient reasoning framework for graph-based RDF knowledge bases. Thereby, URDF uses a SPARQL-like query model. Moreover, URDF augments first-order reasoning by a combination of soft and hard rules. In addition, URDF applies a novel approximation algorithm for a generalized version of the Weighted MAX-SAT problem to resolve inconsistencies between the underlying knowledge base and the inferencing rules at query time. The knowledge base currently used by URDF to answer user-given queries is YAGO. Thereby, URDF produces potentially complex lineage information during its reasoning process. These produced reasoning data, which are a valuable source of information for the user, pose some tough challenges for a suitable visualization. In this thesis, we present UViz (URDF Visualization), a complete visualization system, using URDF as reasoning backend. UViz is built in a client-server fashion. Thereby, UViFace (UViz Visualization InterFace), the visualization interface of UViz, uses Adobe Flex and Flash Player to run as a RIA (Rich Internet Application) inside a common web browser. Thereby, URDF runs on the server. Moreover, URDF applies the Flex-specific data service BlazeDS to guarantee fast data exchange between the visualization on the client and the URDF on the server. Finally, UViz integrates the Flare visualization toolkit to provide a dynamic and visually appealing graph visualization. This way, UViz is able to visualize the information produced by URDF in an intuitive and meaningful way. Thereby, UViFace supports three different operation modes to explore the visualization, examine the lineage information and compare query results with and without rule changes. Moreover, UViFace applies several state-of-the-art visualization techniques to support the user in working with the visualized data. Finally, UViFace provides a visualization that allows the user to accomplish URDF-related user interface tasks intuitively. This is demonstrated in this thesis
User Submitted?: No
Authors: Meiser, Timm
Institution: Saarland University
Department: Computer Science
Advisor:
Degree:
Publisher Location:
Pages:
Data Collections: IPUMS USA
Topics: Population Data Science
Countries: United States