https://www.math.aau.at/talks/115/pdf

Two decades into the 21st century, it is fair to say that the internationalisation of higher education and, concurrently, the use of English to achieve such internationalisation is an indisputable reality. However, under the seemingly homogenising label of ‘English-medium education’ (EME) we find a wide range of complex and highly situated phenomena that come in different shapes and forms. To examine these multi-faceted realities in a comprehensive and holistic way, this session will argue for the use of a conceptual framework known under the acronym of ROAD-MAPPING, which is anchored conceptually in sociolinguistic and ecolinguistic approaches as well as language policy research. After introducing the ROAD-MAPPING framework, I will illustrate its wide usage by elaborating, firstly, how it can be applied to describe EME in different countries, secondly, how it can be used to design a research project and, thirdly, how it allows for re-interpreting existing EME research.

Vasilij Starostin stellt in seinem Vortrag ein Projekt der Menschenrechtsorganisation Memorial vor. Das Ziel dieses Projektes ist die Sichtbarmachung der Tatorte des „Großen Terrors“ 1937/38 in Moskau.Die 1989 gegründete Organisation Memorial dokumentiert politische Unterdrückung und Massenverbrechen in der Sowjetunion und im postsowjetischen Russland. Außerdem beherbergte Memorial eines der größten unabhängigen Archive in Russland und sowie ein Museum und eine Bibliothek. Die Menschenrechtorganisation wurde bereits seit langem von russischen Behörden drangsaliert, unmittelbar nach dem Angriff Russlands auf die Ukraine endgültig aufgelöst. Die Räumlichkeiten von Memorial wurden am 4. März 2022 durchsucht und verwüstet, die Mitarbeiter und Mitarbeiterinnen massiv bedroht.

When talking about new developments in Machine Learning, we typically think about new algorithms, better optimization techniques, or optimized hyperparameters. However, one important aspect is often neglected: the quality and the structure of training data: measurement noise, label noise, and correct but ambiguous labels. In this talk, we address the latter problem, trying to deal with high intra-class and small inter-class variability in the data, following two different strategies. First, we consider the problem of metric learning, showing that by selecting/learning a better metric for a specific problem, better results can be obtained: using the same learning method and the same data. Second, focusing on neural networks, we analyze the influence of specific hyperparameters, namely the activation functions. For both directions, we show that the quality of the finally learned model is highly dependent on the data. To illustrate these aspects, we will further discuss a visualization technique, namely information planes, providing better insights into the current state of the learning system.