| Zeit: | 28. Juni 2022, 18:00 – 19:00 Uhr |
|---|---|
| Veranstaltungsort: | Universität Stuttgart Abteilung SOWI VII Raum 2.076 Seidenstr. 36 70174 Stuttgart |
| Download als iCal: |
|
Ort: Seidenstraße 36, Raum 2.076
Im Rahmen des Kolloquiums der Abteilung SOWI VII freuen wir uns ganz herzlich Agnieska Falenska mit dem Vortrag "NLP Intelligent Systems -- Design and Methodological Perspectives" bei uns begrüßen zu dürfen.
Beschreibung:
Natural Language Processing (NLP) systems aim at analyzing, understanding, and generating human language. Possible applications of these tools are constantly growing and reaching domains beyond the NLP field, with such use cases as medical fact-checking, supporting political debates, or translating technical reports. Yet, standard NLP publications focus primarily on improving the accuracy of developed techniques, leaving out insight on how to apply them outside of the typical experimental setups.
In this talk, I will look into the design of NLP models from three methodological perspectives: run time, accuracy, and training resources. First, I will show how to incorporate these perspectives into guidelines for designing particular NLP architectures -- ensemble dependency parsers. Then, I will focus on the questions related to the training resources. Specifically, I will present my recent work on estimating gender-related biases in Wikipedia's content and their possible consequences for NLP models trained on this data source.
Alle Interessierten sind ebenso herzlich dazu eingeladen. Eine Anmeldung ist nicht nötig.
Zur Webseite von Agnieszka Falenska