El Grupo de Lingüística Formal del Instituto de Filología y Literaturas Hispánicas "Dr. Amado Alonso" invita a la charla que dará el Dr. Mattis List (Max Planck Institute for the Science of Human History) el día Lunes 24 de septiembre a las 18:00 con el título "Chances and Challenges of Computational Approaches in the Humanities (from the Perspective of a Historical Linguist)".
Resumen de la charla:
Chances and Challenges of Computational Approaches in the Humanities (from the Perspective of a Historical Linguist)
In times of steadily increasing amounts of digitally available resources in almost all branches of science, it is essential for all scholars to understand both the potential and the problems involved with computer-based or computer-assisted applications. The problems we face in this context are not only practice, concerning the power of available algorithms and methods to conduct tasks which have so far been exclusively carried out by humans, but also theoretical, concerning the limits of current computational frameworks in machine learning.
As a historical linguist who was trained together with biologist, I am in some sense a representative of a younger generation of scientists in our field, making active use of computational approaches in my daily research. Having originally been trained by philosophers and classical linguists, however, I have also learned to be careful with all new advances that the new methods promise. Furthermore, from my practical experience and my longstanding attempts to automatize the task of historical language comparison, I have also learned that the problems we are dealing with in historical linguistics (and probably also in many other branches of the humanities) are often far more complex than usually assumed by engineers and computational scientists.
In the talk, I will try to give an overview on the chances and challenges of computational approaches in the humanities. Based on my previous experience as a linguist trained both in the classical and the more recent computational approaches, I will give examples where computer can indeed provide actual help. On the other hand, however, I will also point to obvious misunderstandings involved in computational applications. These problems result in part from the black-box character of many machine learning approaches (Castelvecchi 2016), but in part they also result from the deeper misunderstanding of computer scientists regarding the nature of our data and our methodology in the humanities (see, e.g. Marcus 2017).
References
Castelvecchi, Davide. 2016. "Can We Open the Blackbox of AI." Nature 538: 20–23.
Marcus, Gary. 2017. "In Defense of Skepticism About Deep Learning." Medium Corporation. https://medium.com/@GaryMarcus/6e8bfd5ae0f1.