University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program

title

Visualization and Interactive Data Analysis

speaker

Prof. Heer Jeffrey, PhD, Stanford University
Stanford, USA

date & place

Wednesday, 30.05.2012, 15:15 h
Room C 252

abstract

The increasing scale and accessibility of digital data provides an unprecedented resource for informing research, business and public policy. Yet acquiring and storing this data is, by itself, of little value. Turning data into knowledge is a fundamental challenge for both computer systems and user interface research: it requires integrating data management systems and analysis algorithms with human judgments of the meaning and significance of observed patterns. In this talk, I will discuss our research attempting to address this challenge through novel interactive systems for data visualization and manipulation.

First, visual representations are regularly used to aid perception of patterns, trends and outliers in data. To aid this process, we are investigating the design of declarative, domain-specific languages for custom visualization. Our resulting languages (Protovis and D3) simplify specification and enable performance optimization while preserving an expressive design space. These systems are now widely used throughout academia and industry.

Second, data analysts often expend an inordinate amount of effort manipulating data and assessing data quality issues. With our Wrangler system, users can construct data transformation scripts in a direct manipulation interface. Wrangler uses programming-by-demonstration methods to automatically suggest applicable transforms and preview their results. The end result is not simply transformed data, but a reusable program that can be run on other platforms (e.g., MapReduce) to process data at scale.

Collectively, these systems contribute new approaches for improving the efficiency and scale at which expert analysts work, while lowering the threshold for non-experts.