#7: On Language Processing and Transfer Learning


Missed us?
The Natural Language Processing & Text Analytics Meetup is back. We are kicking it off in cooperation with the CorpusLab of the University of Zurich. Tanja Samardžić will introduce the Language and Space Lab and talk about the relationship of language and NLP, and Barbara Blank will introduce her research on transfer learning in NLP. Each presentation is 20−30 minutes followed by a short discussion time. Please RSVP on the meetup site.

19:00 Welcome
19:10 Tanja Samardžić (University of Zurich): Not so Natural Language Processing in a Lab
19:45 Barbara Plank (University of Groningen): Transfer Learning in Natural Language Processing - Where are we now?
20:15 Meet and greet

About Tanja:
Dr Tanja Samardžić is the head of the Text Group, Language and Space Lab, UZH, with a background in language theory and machine learning. Her research is about developing text processing methods and using them to test theoretical hypotheses on how language actually works. Her current projects include various activities on promoting and facilitating the use of computational approaches in the study of language.

Not so natural language processing in a lab

As odd as it sounds, the study of language is not part of NLP. NLP is even less part of the study of language. What does it take to fix this strange relationship, or rather lack of it? What does text tell us about language? These are overarching questions underlying projects and activities of my group. I will show in this talk how we approach these questions and how our findings can be used for practical applications.

About Barbara:
Barbara Plank is Assistant Professor (tenured) in Natural Language Processing at the University of Groningen. She received her PhD in 2011, and has previously held positions as assistant professor and postdoc at the University of Copenhagen and was postdoc at the University of Trento. Her main research includes learning under sample selection bias (domain adaptation, transfer learning), annotation bias and generally, semi-supervised, weakly-supervised and multi-task learning for cross-domain and cross-lingual NLP, applied to a range of NLP tasks covering tagging, parsing, relation extraction, opinion mining, and personality detection. She severs on the editorial board of the Computational Linguistics Journal, is area chair for NAACL 2018 and chair for the language and computation track at ESSLLI 2018.

Transfer Learning in Natural Language Processing - Where are we now?
Humans adapt their language use to fit an unbounded variety of contexts. In Natural Language Processing (NLP), however, the dominant paradigm is to treat language as uniform and static: models are trained from scarce and biased training data. The consequences are serious: these models work well only on similar text types, and suffer dramatically on texts from other types, authors with different demographic backgrounds and cannot transfer to other languages.
In this talk I will survey some current approaches to learning from diverse sources (domains, tasks and languages), including some of my work on learning from fortuitous data, cross-lingual learning, data selection and semi-supervised learning for transfer learning. I will outline current challenges and give some preliminary directions on where to go next for learning under limited (or absence) of annotated resources.

University of Zurich
Rämistrasse[masked] Zürich

CorpusLab UZH
Swiss Re