PNNL

Abstract

This paper discusses our recent work across a number of disciplines, leading to a concept for a next generation analytical environment for scientific discovery within continuous, time-varying data-streams. First, we have created a stream-processing engine that processes multiple streams of interest. An analyst, via a client interface, reviews the data-stream format and specifies upstream filtering to define stream tokens of interest, leading to a highly specialized collection of time-variant material. We envision using this collection to drive an existing system that visualizes thematic variations over time across a corpus of information. This ‘ThemeRiver™’ helps analysts discern trends, relationships, anomalies, and structure in the data. Further, we make use of a number of technologies that allow us to investigate these elements in ambient environments that surround the user, placing them within their data. We discuss the HI-SPACE (Human Information Space) as a tool for bringing together the most desirable aspects of both physical and electronic information spaces to enhance the ability to interact with information, promote group dialog, and to facilitate group interaction with information to solve complex tasks. Here, we introduce a concept that combines these approaches to produce an advanced analytical environment for data stream analysis that provides a collaborative, ambient environment for scientific discovery in data-streams.