Computer Science Department News

There will be a Ph.D. Candidacy on May 4, 2011 at 8:30 AM by Hui Shi

Title: Adaptive Reasoning or ScienceWeb
Location: E&CS Building Conference Room RM 3316
Student: Hui Shi
Advisor: Dr. Kurt Maly

Abstract:
ScienceWeb is a system that provides answers to qualitative and quantitative queries of a large evolving knowledge base covering science research. The system will support the community joining together, sharing the insights of its members, to evolve: (a) the type of data to be gathered and how to organize them, (b) the methods for collecting the data, their sources, the process of collecting them, and validating them, and (c) the meaning of qualitative descriptors and queries most needed and how they can be computationally realized.

ScienceWeb will need to scale to accommodate the substantial corpus of information about researchers, their projects and their publications. It will need to accommodate the inherent heterogeneity of both its information sources and of its user community. Finally, it must allow the semantic (qualitative) descriptors to evolve with time as the knowledge of the community grows and the problems the community researches change. As the size of knowledge base increases, scalability becomes a challenge for the reasoning system.

In this thesis we will research the issues involved in developing a reasoning architecture and designing an adaptive reasoning system whose scalability and efficiency are able to meet the requirements of query and answering in ScienceWeb. For evaluation purposes, we shall develop a base query and rule set as well as instance data from a variety of sources.

My system is adaptive under changing situations to ensure efficient reasoning. First, we will combine both Query-invoked Inference and Materialization Inference with incremental inference for more efficient query and answering. Second, we will introduce adaptive mechanisms that switch between Query-invoked Inference with Materialization Inference after changes have been made. Third, we will introduce new ways of storing, grouping and indexing objects in the knowledge base for faster searching and reasoning as the size of the triple set scales to the millions and the complexity of the Abox increases. Since reasoning is necessary to many semantic applications, my contributions to the field of efficient reasoning should enable one to extend these methods to other than the ScienceWeb applications.