ViewDesigner: A Tool for Designing Viewscopes in Discovery System HypothesisCreator

Knowledge discovery from complete genome data demands more discovery-oriented computational methods than homology search or keyword search. Currently, over 20 complete genomes have been determined, and 82 prokaryotic and 24 eukaryotic genome sequencing projects are in progress [3]. However, software tools to strongly support the discovery process of genomic researches have not established yet. HypothesisCreator is a multi-strategy discovery-oriented knowledge discovery system, which allows users (i.e., domain experts) to design views on data [1, 2]. A view on data means a particular way to interpret the data. In the scientific literature, devising new view on data is a key to discovery. To deal with views in a systematic way, the concept of ”viewscope” is formulated in [1, 2]. Informally, a viewscope V is defined as a pair of a polynomial-time algorithm ψ of interpreting data and a set P of parameters of the algorithms. We call ψ and P the interpreter and the viewpoint set of V , respectively. A view on data is a viewscope V = (ψ,P ) with a fixed viewpoint, i.e., |P | = 1. One of the features of HypothesisCreator is the automatic generation of viewscopes in the process of search for good views along with hypothesis generation. In the current version of the system, clustering programs and a decision tree generator are available as hypothesis generators. Another is to realize human intervention in the discovery process by designing viewscopes by experts. ViewDesigner is a component of HypothesisCreator, which manages viewscopes and provides an interface between the system and users to design and operate viewscopes. In this abstract, we introduce a new version of the component, which is designed to provide environment for users to design viewscopes intuitively.