Novas abordagens computacionais em linguística histórica

Novas abordagens computacionais em linguística histórica

Presentation in Portuguese used in the first day of the Workshop "New Computational Approaches in Historical Linguistics" as part of the ABRALIN50 conference in Maceió (Brazil).

The basic techniques of historical linguistics, the “comparative method”, have not changed significantly in the last centuries: intensive comparison of linguistic material, identification of regular correspondences, and reconstruction of the history of languages and their families. Among its limits are an essentially manual execution (even when aided by computers), limited collaborative networks (especially in terms of data sharing), and the fact that some crucial tasks (such as the identification of cognates) are sometimes based on non-formalized knowledge. Computational proposals to overcome these limits have been advanced for more than 50 years, with limited acceptance by the community (see the justified reception of the M. Swadesh glottochronology and the proposals on Nostratic); however, during the last decade we have witnessed a widespread and progressive acceptance of a new category of methods, influenced by Bayesian statistics and practices from biology, in response to challenges imposed by certain language families (such as Sino-Tibetan or sign languages) and by the demands of transdisciplinarity and open access. Starting from a proposal to complement the traditional approach, in which results must always be interpretable and whose purpose is to assist and not replace researchers, this workshop will begin by reviewing the comparative method and exposing the principles of the new approaches, illustrating their limits and presenting the criticisms already advanced. Participants will learn to perform phonetic alignments, identify patterns of sound correspondence, detect cognates, constructo phylogenies, and investigate total and partial colexifications. Emphasis will be placed on the appropriate preparation of linguistic databases for collaboration and processing. Although it is not necessary for the participants to reproduce the demonstrations, all computer interaction can be reproduced individually in an open web interface ( The workshop is offered with the aim of disseminating the new methods, arousing a new interest in historical. This workshop wil be taught in Portuguese.


Tiago Tresoldi

May 02, 2019