Multi-Domain Geovisualization of News Stories

Previous

Students

Julie Dillemuth, Geography
Alex Villacorta, Statistics
Rama Hoetzlein, Media Arts & Tech
Carlos Castellanos, Media Arts & Tech
Shane Kendrick, Undergraduate Researcher

Faculty Advisors

Keith Clarke, Geography
S. Rao Jammalamadaka, Statistics

George Legrady, Media Arts & Tech
Lisa Parks, Film Studies

Next

Abstract

How does information, specifically news stories, move around the globe from where the news happens to where the public reads or hears about it? Does it spread like disease, a story ‘infecting’ the population as it is communicated through mass media or from one person to another, or can it be compared to point-source pollution, flowing outwards along particular vectors?

The goal of this project is to statistically analyze and visualize the movement of a news story over time as it is reported in U.S. newspapers. Traditionally, the idea of information flow pertains to the evolution of information from its inception to archival where its geographical path is of little significance. In this project we aim to consider both the temporal and geographical aspects of information flow.

Newspaper stories, unlike television, radio or Internet news sources, have been widely archived in digital databases (such as America’s Newspapers by Newsbank, Inc.) and are published and circulated in specific geographical locations By analyzing the statistical and geographical features of news stories from origin to archival and the dissemination of the papers to a readership it is hoped that a mathematical model may be created which helps to explain the spatial properties of communication channels in the US. Geovisualization theory and techniques provide the foundation for data exploration and communication through dynamic maps and graphics.

Stories are selected through a database query on a particular topic and parsed into a table with headline, newspaper, date, publication location, and word count. A geographic information system (GIS) translates the table into a series of maps at different time steps, indicating the publication of stories at a city level, and Flash software animates the map sequence to convey the temporal characteristics of the news stories. For the story content, latent semantic indexing (LSI) measures story similarity, identifying how relevant a story is to the main topic and tracking how a news topic evolves over time. Newspaper circulation and readership information provide statistics for modeling the spread of the news through the population.

While this research will start on a small scale of comparing the spread of individual news topics within specific regions, the ultimate goal is to be able to quickly analyze and visualize any dataset of news stories, from the scale of a city or urban area, to the nation as a whole, and potentially worldwide. In addition, there are several broader implications of this project. Contributing to a better understanding of the flow of information through a model will help to identify both strong and weak channels of information transfer. Such knowledge would be useful in preparing against attacks on communication systems within a region. Also, emergency response systems need to know which communication channels reach the most people in the shortest amount of time. Lastly, it is hoped that this project could lead to a determination of what effect, if any, the flow of information has on cross-cultural understanding and geopolitical (i.e. power) relationships.

a) March 21, 2005

b) March 22, 2005

The above images visualize the spread of news related to Terry Schiavo from March 21-March 22, 2005. Click an image for a full picture.