Our keynote speaker will be Michael J. Kurtz of the Smithsonian Astrophysical Observatory. Dr. Kurtz will be speaking on The Astronomical Information Network.

Astronomical Objects (stars, galaxies, ...) are bound together by a complex and often surprising assortment of shared ond/or similar interactions and histories. As a purely observational science it is the task of astronomy to disentangle the vast network of objects with shared or similar properties and discover the underlying causal relationships which govern our universe.

As an example some galaxies are blue, and exibit spectral features typical of a hot, ionized gas; other galaxies are red, and show none of these gaseous features. These red galaxies tend, very strongly, to cluster together in space, while the blue galaxies do not. Edwin Hubble first noticed this 75 years ago. What causes this effect, were the galaxies formed this way, or did they become this way over the history of the universe?

Astronomy research exists now within a large and growing man-made network of tightly interconnected data sources and services. Based on a culture of freely shared information and shared goals astronomers are building a Virtual Observatory, using the internet to bring the totality of astronomical information to anyone, anywhere in the world.

Astronomy research also exists within more abstract networks of thought and behavior. The structure of astronomy research can be seen in the co-citation network of astronomy research articles, but it can also be seen in the co-reader network of those articles, and the co-keyword network of those same articles again. Are these structures the same?

Despite the wide and bewildering array of data mining, machine learning, visualization, and classification methods available to researchers, one finds a core set of concepts and issues common to most of them. Methods like similarity computation, and concerns over variable selection, replication, interpretation and validation are important in most studies, no matter which algorithms are chosen.

This short course provides an introduction to these issues and techniques, as well as hands-on instruction. Our vehicle will be an introduction to cluster analysis, with simple methods for authorship attribution as the running example. But our discussion and list of resources will extend into a number of other domains and techniques.

The course will be offered on the afternoon of Thursday, June 7, following the last regular session of the Digital Humanities meeting. David Dubin will be the instructor, and DH/CSNA attendees can register for the course on the website for either meeting. A laptop computer will be helpful, but not essential for participating in the course. Participants can save some time by installing the R statistical analysis software, available free of charge from the R-Project website.

Techniques like cluster analysis, supervised classification, and many other mapping, ordination, scaling, and visualization methods have been applied in humanities scholarship. These applications often produce results that are provocative and interesting, but it's not always clear precisely what inferences the analytic methods do (or are intended to) support. Exploratory application of these techniques to new domains may indeed be fruitful. But if humanistic scholarship is the starting point, how does one recognize that a particular method may suit a scholar's needs? How do understandings of evidence in humanistic disciplines compare to those in the social and natural sciences, and how should this inform decisions about information processing and analysis?

We will explore these questions at a joint DH/CSNA workshop, held on the morning of Friday, June 8. Beginning with a discussion of three case studies, we will work with the aim of expressing guidelines and compiling a set of useful resources. All Digital Humanities and CSNA attendees are invited to participate.