HARVARD GAZETTE ARCHIVES

Research associate Malcolm Hyman (from left), Mark Schiefsky, assistant professor of the classics, and lecturer Elaheh Kheirandish discuss the Archimedes Project, which develops advanced computer technology to mine classic texts in a variety of languages. (Staff photo Kris Snibbe/Harvard News Office)

Today, a team of Harvard scholars in the Department of the Classics is leveraging 21st century technology to achieve wonders of its own. The aptly named Archimedes Project, now in its third year, aims to investigate the history of mechanics with new, penetrating tools that can, with a click or two of a mouse, peel back time and reveal heretofore hidden truths.

The goal of this academic research project, which is funded by the National Science Foundation and conducted in collaboration with the Max Planck Institute for the History of Science in Berlin, is to develop advanced computer technology that will help scholars mine myriad scientific texts in a variety of languages, but also to connect the dots between them. In this way, it can broaden the scope of scholarship, but also sharpen the scholar's ability to probe deeply into the past, thereby shedding new light on generally accepted assumptions.

"Now what we can do is take into account a vastly wider range of evidence, and that changes the picture quite considerably," explained Mark Schiefsky, an assistant professor of the classics at Harvard and the principal investigator of the Harvard team. "It changes our idea of what mechanics was like in antiquity, in such a way that it makes it seem both much more complicated and more modern than one might have thought."

In other words, Archimedes, Aristotle, and their ilk were far more advanced in their thinking than we generally give them credit for. They were truly ahead of their time. And Schiefsky and his team plan to prove it.

But for that you need sharp, smart tools, which are the chief responsibility of Malcolm D. Hyman, a research associate with the project. "What we want to do is to get scholars to be able to see the big picture," said Hyman. "This is a complex tradition. We have all these different languages, and we have lots of texts. By using this technology, what we hope to do is to allow people to work with a mass of material that they would otherwise not be able to work with."

In some ways, the technology driving the Archimedes Project resembles the sophisticated computer programming used more often in the hard sciences, where computers are designed to decode genes, or map complex chemical structures. But the Archimedes team intends to prove that such techniques can also bear fruit in the humanities and the softer sciences.

"It's a new direction for the study of the past and its future," said Elaheh Kheirandish, a Harvard lecturer and the third member of the Archimedes Project team. "It's very young, and there's no question that it's uncommon, but there are very interesting projects that use humanities and technology."

So far, the Archimedes Project technology can perform a variety of impressive feats, including automatic morphological analysis - which means that each word in a text, be it Greek, Arabic, or Latin, can be linked to its root form in the dictionary. This is a godsend for scholars who want to study primary sources in their original languages, but need a little extra help in mastering them.

A key software program for the project is called Arboreal. It facilitates linguistically complex searches, and also allows for the annotation and creation of data. In this way, you can scribble in the margins, as it were, of the ancient text inhabiting your computer screen. "What's novel about Arboreal is the integration of all these linguistic tools into one package," explained Hyman, who created it. "It's new in scale - there is no existing tool, actually, that will serve scholars working in the type of texts that we're working on. Arboreal is the first."

But it is nonetheless insufficient. Eager for more, the Archimedes team is in the process of acquiring yet another instrument for its forensic research, called latent semantic indexing. This technology uses statistics to characterize and identify texts by determining the structure and meaning of words, how frequently they are used, and in what context. In this way, for example, the identity of an anonymous author can be discovered, and translations can be distinguished from original content.

With such tools, the Archimedes Project members can start to ask, and answer, questions that have until now been too unwieldy or time-consuming to tackle.

In particular, the Archimedes team hopes to go back in history and re-create important ancient Greek texts, which were known to have existed but have since disappeared.

"For the first time I feel that there is the possibility of reconstructing lost fragments," said Kheirandish, speaking of missing documents by Archimedes, which are referred to in later texts on mechanics. "The technology gives you a range of vocabulary in a magnitude that is unimaginable. The fact that you have the linkage of all these roots, the morphology, the connections, allows you to reconstruct something almost on the spot."

In many cases, these lost Greek texts can only be reconstructed through careful analysis of a handful of critically important Arabic translations, which are the only copies of these works to have survived. During the early Middle Ages, while much of Europe stagnated, Arabic culture, with its headquarters in Baghdad, flourished. Starting in the ninth century A.D., caliphs and viziers funded an extensive and systematic translation movement, which transferred to their own time the ancient wisdom of the Greeks, thereby preserving many crucial texts, which would have otherwise perished.

"It's a process of transformation, where Greek originals are transformed into Arabic sources, which are then used for further development," Schiefsky said. "In those days, they really did appreciate what the Greeks had achieved and tried to bring it into their own language."

Certain Arabic translations - such as one by Qusta Ibn Luqa of a document called "The Mechanics" by Heron of Alexandria, a first century A.D. Greek scientist - are key texts for the Archimedes Project team as it sets out to reconstruct, to translate in reverse, missing Greek documents. In particular, Schiefsky and Kheirandish, who reads both Arabic and Persian, are working on a section of this text in which references to Archimedes' definition of the "center of gravity" appear.

"Archimedes apparently invented the concept of center of gravity. It persists throughout the entire history of mechanics, and one of the important things about this text is that it gives us insights into the introduction of this concept," explained Schiefsky. "So if one can reconstruct from this an actual argument of Archimedes that belongs to some lost work, that would be quite helpful."

Although Archimedes' notion of the center of gravity does not anticipate Newton's breakthrough 20 centuries later, it does reveal sophisticated abstract thinking on weights and balances. Archimedes realized that the entire weight of an object could be regarded as concentrated at a single point - its center of gravity. This concept, which is still accepted today, is essential in designing mechanical systems.

Such third century B.C. thinking was precocious, and according to Schiefsky, "extremely modern" on the part of Archimedes. It's Schiefsky's hope that the Archimedes Project will be able to not only reconstruct Archimedes' original argument on this concept, but also, in time, rescue from neglect dozens of Greek texts, which currently exist only as Arabic translations.

"This will take years of work from all three of us, and beyond, to actually get to where we want to be," explained Kheirandish. "But one must not shy away from these things. Three years back when we started, where we are today would have been unthinkable." Academic research of this nature is usually slow and painstaking, but the Archimedes team is betting that their technology has the potential to transform this process.

"We're taking off in leaps and bounds," said Schiefsky. "We're not trying to do everything by hand. This one little example on the center of gravity is only the first of the many results we could expect when we actually apply the full resources of modern technology to these questions - which has never been done. So what we can expect, I think, is really extremely impressive."