Abstract: This talk will cover a variety of setups and approaches to robustness in NLP. I will start by giving an overview (including limitations) of the task of lexical normalization: the conversion of social media data to its canonical form. Next, I will discuss the challenges in using multi-task learning to improve performance in low-resource setups. Finally, I will unveil remaining challenges in the first steps of the NLP pipeline: language identification and tokenization (i.e. word segmentation): are these tasks really solved?, and if no: what are still the open challenges that we should focus on in the future?
Location: Attend in person at room J411 or via Zoom, https://gu-se.zoom.us/j/66299274809?pwd=Yjc2ejc2VVhraXVJMmhWeWtOQ2NuUT09
Time: 13:15-15:00