Dobbelt professortiltrædelsesforelæsning: Anders Søgaard og Robert Glück
Den 4. maj 2018 fejrer vi DIKUs to nye professorer, Anders Søgaard, professor i Machine Learning og NLP, Natural Languages Processing, og Robert Glück, professor i Programming Models and Techniques, med en dobbelt tiltrædelsesforelæsning.
Anders Søgaard: More nightmares at test time
I give a crash-course on why NLP is hard. One reason, a result of sparsity and bias, is feature swamping, i.e., under-training of features that covary with more frequent and predictive ones. Such features may be false friends if not present at test time. I generalize the notion of false friends to unrealistic experimental conditions and give examples of how such conditions have led to inflated results in the NLP literature. Finally, I give examples of some of my work on obtaining more robust NLP models by learning from more diverse signals (multi-source cross-lingual transfer, multi-task learning, etc.).
14.50-15.00 Kort pause
Robert Glück: Metacomputation - extending the horizons of computing
Despite tremendous technological progress over several decades, the development of software still relies to a greater extent on non-automated processes. While our dependency on computer systems has grown almost as rapidly as their complexity, developing the programs requires enormous amounts of human eff ort and high levels of investment. An important research goal of computer science is, therefore, to deal with the analysis, transformation and generation of programs by programs. More specifically, we aim to advance three fundamental meta-operations: the composition, inversion, and specialization of programs. A long-term endeavor well illustrated with the development of the 'Copenhagen interpretation' of reversible computing at DIKU.
16.00 ff. Reception
Anders Søgaard is professor in Natural Language Processing (NLP) and Machine Learning, Dept. of Computer Science, University of Copenhagen. He holds a BA in Linguistics from University of Copenhagen (2003) and an MA in Computational Linguistics from CBS (2004).
He received his PhD in Language Technology at University of Copenhagen in 2007 and his Dr.Phil (habilitation) also in Language Technology (University of Copenhagen) in 2014. In 2015 he relocated his NLP research group from the Centre of Language Technlogy (CST) to the Department of Computer Science (DIKU).
Previous positions: Senior researcher (University of Potsdam), Postdoctoral researcher (CBS).
Robert Glück is Professor of Computer Science specialising in Programming Models and Techniques at the University of Copenhagen. His main research interests are programming languages and software systems. His current research focus is on reversible computing, program generation, and metaprogramming.
He has chaired and edited the proceedings of several conferences and workshops published by Springer and by the ACM. According to Google Scholar, Glück's publications had an h-index of 34.
Glück has been the principle investigator of a number of research awards and projects. He received several fellowships from the Austrian Science Fund (FWF) and the Japan Society for the Promotion of Science (JSPS), and a PRESTO Basic Research grant from the Japan Science and Technology Agency (JST). He led projects funded by the Danish Councils for Strategic Research (DSF) and for Independent Research (FNU).
He currently serves as the Head of Studies of the Master in Computer Science and on the Study Board of Mathematics and Computer Science at the Faculty of Science.