Sie sind hier: Startseite Kolloquium Let’s get more data! Or learn smarter? Or fairer?

Let’s get more data! Or learn smarter? Or fairer?

Dr. Lucie Flek, Depart. of Mathematics and Informatics, Philipps University of Marburg

Art des Termins
    Wann 21.06.2022
    von 15:00 bis 16:00
    Wo b-it, room 2.113
    Termin übernehmen vCal



    Research Talk: Let’s get more data! Or learn smarter? Or fairer?

    While the online world tends to give us an illusion of abundant big data to analyze, there is a constant lack of accessible labeled resources in many sensitive domains or underrepresented languages. In this talk, I will discuss our journey from data augmentation strategies, across the improvements in representation learning methods and model architectures, towards fairer and more efficient learning with a lower carbon footprint. I will demonstrate our approaches on applications such as mental health analysis from text, and toxicity detection and reduction, inter alia.



    Lehrprobe:  Recurrent neural networks

    Understanding RNNs is an important theoretical step for contextualizing the research development from traditional sequential modeling approaches to contemporary attention models. In this short sample I aim at explaining the learning mechanism of RNNs and its pros and cons that led to further recent developments.


    Location (for both talks): b-it , Friedrich-Hirzebruch-Allee 6, 53115 Bonn, Room no. 2.113