The idea of high performance computing usually conjures notions of research such as rare isotopes, global warming, how a virus behaves at the molecular level, and other issues in the so-called hard sciences. Dr. Mark Kornbluh plans to use MSU’s High Performance Computing Center (HPCC) to study problems in the social sciences and humanities.
Dr. Kornbluh chairs the History Department and heads MATRIX: The Center for Humane Arts, Letters and Social Sciences Online. MATRIX hosts a wealth of content in the social sciences and humanities. Dr. Kornbluh envisions using the HPCC to study two of the content collections hosted by MATRIX: H-Net, a worldwide forum for scholarly discussion, and the Quilt Index, a database with images and descriptions of quilts.
H-Net, www.h-net.org, began in 1993 with four “Listserv” mailing lists – Internet discussion groups conducted via e-mail. Today it comprises over 160 public lists. Over the years over 1.2 million messages have been posted. Some 600 editors moderate the postings. A special focus of H-Net is book reviews; it is the largest publisher of on-line scholarly book reviews in the world. H-Net also publishes conference announcements, job postings, and multimedia materials.
Dr. Kornbluh seeks to use high performance computing to analyze the cumulative discussions on H-Net. For instance, how do discussion threads evolve over time, as events and new ideas enter the consciousness of scholars? For instance, when the movie Amistad was released, discussion of slavery and the slave trade spiked. “Simple keyword searches won’t work for the analysis we want to do; you’d get a result page with far too many hits,” Dr. Kornbluh says. “We need to do rich semantic data mining.”
The H-Net corpus has relatively little metadata, or data about the data. For example, each message has its date and time of posting, subject line, and author name. But subject lines are not always consistent; a posting responding to a previous message may have a different subject. Sophisticated clustering could tie those messages into a thread.
The National Historic Publication and Records Commission has awarded a grant to begin the research, and MATRIX has applied to the National Science Foundation for additional funding.
The Quilt Index, www.quiltindex.org, contains 15,000 images now and will grow to over a million images from 30 major collections over the next three years. Unlike the H-Net case, the Quilt Index houses a huge amount of metadata – detailed descriptions of each quilt. This provides rich fodder for analysis.
Dr. Kornbluh wants to explore whether cultural products, such as quilts can be analyzed systematically to explore key historical questions. For example, as African-Americans moved North and to Urban areas in the mid-twentieth century, did African-American quilt patterns and images change. Dr. Kornbluh wants to see if such analysis can provide a new window on key historical issues of segregation andcultural integration. “Similarly, we can ask what were the effects of 9/11 on quilt patterns in order to explore the broader impact of traumatic national events”
Dr. Rong Jin, Assistant Professor in Computer Science and Engineering, is developing software to use HPCC computers to perform the analysis. Dr. Kornbluh notes that one goal will be to compare the results of qualitative analysis by humans with computed results.
The Quilt Index research has been funded by the National Endowment for the Humanities and the Institute for Museum and Library Services in four separate grants totalling over $1 Million. Additional funding from the NSF is pending.
Dr. Kornbluh began using computers for social science research while a graduate student in History at The Johns Hopkins University. He used SPSS (Statistical Package for the Social Sciences) to study voting patterns, leading to a book, Why America Stopped Voting (New York university Press.) He joined the MSU faculty in 1994.
