In this talk we will present a method for efficiently inferring evolutionary trees in Historical Linguistics which avoid the difficulties that have made this analysis intractable. We use primary data, and show that an appropriate optimization problem (based solidly upon traditional historical linguistic methodology) can be solved exactly for these data.
We have applied this method to the problem of inferring the evolutionary history for the Indo-European (IE) family of languages, and have made several surprising and strikingly strongly supported findings. We analyzed the IE data with particular interest in determining whether our new methodology could lay to rest the debate on two longstanding conjectures: the {\em Indo-Hittite hypothesis} and the {\em Italo-Celtic hypothesis.}
Our analysis indicates significant support for the Indo-Hittite hypothesis and preliminary (albeit weak) support for the Italo-Celtic hypothesis. It also proposes a reasonable explanation for the surprising dual allegiance of Germanic. Most importantly, it provides a firm methodology by which linguists can test the consistency of their judgements, thus enabling further research into problematic data.
This is joint work with linguists Donald Ringe and Ann Taylor of the University of Pennsylvania.
Note: If you would like to meet with the speaker, please send mail to Vineet Bafna at bafna@dimacs.rutgers.edu