Linguistics paper co-authored by Warnow settled key questions, honored for its influence


David Mercer, Computer Science

When Tandy Warnow and historical linguist Donald Ringe met in 1994 as faculty members at the University of Pennsylvania, the evolution of Indo-European languages was not well understood.[figure="" class="align-right" width="300"]

CSL Professor Tandy Warnow says recognition for the 2005 paper has led she and Professor Donald Ringe to extend their collaboration.

Warnow and Ringe began a partnership that resulted in 10 papers, one of which, written with Luay Nakhleh, was published in 2005 in the journal Language and has answered many key questions about that evolution. (Nakhleh was then a PhD student and is now the chair of the Department of Computer Science at Rice University).

The paper, “Perfect Phylogenetic Networks: A New Methodology for Reconstructing the Evolutionary History of Natural Languages,” has been selected as part of Volume 3 of The Best of Language, a compilation of the best articles published by the journal between 1986 and 2016.

“I was tremendously pleased.  It is a great honor to be recognized for contributing to the scholarship in linguistics,” said Warnow, now a Founder Professor of Engineering with Illinois Computer Science.

“I'm simply happy that it will give our line of collaborative work exposure to a broader audience,” added Ringe, who remains a professor of Linguistics at Penn.

The selection of the paper also provided a bonus for Warnow: inspiration to return to linguistics work, she said.

“Don and I are continuing to work together, and I have a new (PhD) student, Marc Canby, starting in the Fall who will work with us on extending the mathematical modelling in language evolution,” she said.

When Warnow and Ringe began working together, the evolution of Indo-European languages -- English, Sanskrit, Russian, French, Albanian, Celtic, and many others – was poorly understood. Many questions such as precisely how some languages were related to others and which language developed first from the root were controversial, she said.[figure="" class="align-right" width="300"]

Professor Donald Ringe says linguists were more comfortable with the models created by he and Warnow because they were based in linguistic research and methodology.

“Don and I hoped that by combining our expertise -- his in historical linguistics and mine in the mathematical aspects of phylogenetic tree estimation -- we'd make some progress on resolving Indo-European's history,” Warnow said.

The two developed a series of mathematical models, she said, that assumed language evolution followed a treelike pattern.

As they worked together over the following years, they improved their models to account for borrowing between languages, words that have shared meaning, and other factors. They found strong consistency in the models they developed, but not absolute consistency. And they realized that there was probably borrowing between languages that could not be detected.

The two researchers decided they needed a model that went beyond a tree and instead was based on a phylogenetic network, Warnow said.

“By extending our evolutionary models from trees to networks, we were able to produce a phylogenetic analysis of Indo-European that made sense,” she said, adding that it settled a number of controversies – that the now-extinct Anatolian family of languages was the first to split off from the root; that another extinct language, Tocharian, was the second language to split off; that the Italic and Celtic language groups are very closely related, and more.

The researchers' results also found acceptance in the linguistic community, Ring said.

“By and large, historical linguists are more comfortable with our phylogenetic trees because our approach is firmly based in linguistic research and methodology, unlike some other research," Ringe said.