Professor Profile: Barbara Plank wants to broaden the field of NLP to deal with variability in language use

Barbara Plank has been appointed professor at the Computer Science department at ITU. Her ambition is to broaden the scale of Natural Language Processing and make it more inclusive. She will be giving her inaugural lecture entitled “NLP for everyone” in auditorium 3 at ITU on September 2.

Barbara PlankComputer Science Departmentartificial intelligence

Written August 24, 2021 6:36 AM by Theis Duelund Jensen

Barbara Plank, who is a recipient of the prestigious Sapere Aude research leader grant from the Independent Research Fund Denmark, wants to explore her passion for technology and language to broaden up Natural Language Processing. As an undergraduate in Computer Science, she took a course on computational linguistics which turned everything around: “I realized how complex—and fascinating—it is to model language. Ever since I wanted to work on this challenging problem and build algorithms to understand language,” says Barbara Plank, who most recently with her NLP lab won an outstanding paper award at EACL 2021’s demo track for a project on multitask learning in Natural Language Processing (NLP).

In recent years, NLP has become an enormously popular field of research in part due to the many types of application for the technology which is advancing at a rapid pace. “Today, NLP plays a large part in society. It is increasingly adopted by industry, academia and government, for example”. According to Barbara Plank the advent of deep learning – applying neural networks to AI, so systems can learn on their own – has revolutionised the field and opened it up to the world in a new way:

“Before the 1990’s researchers worked with a rule-based approach to NLP,” says Barbara Plank. “Since then, the statistical revolution took place, which ultimately culminated in the deep learning tsunami”. Today, most people use NLP in some way – whether because of digital assistants or translation services, or even the apps categorizing or spell checking their emails.

Language lessons for machines
At ITU, where Barbara Plank started as Associate Professor in 2018, the researcher has worked extensively on building deep learning systems for machine learning and other types of artificial intelligence. One of the things that attracted her to the field was the wide scope of possibilities it offers:

“Natural Language Processing can be approached from many different directions – from Computer to Cognitive Science – but what has always fascinated me is the linguistics perspective; the fact that we can study language from a computational perspective and develop technologies that benefits everyone – regardless of how widely spoken your native language is.”

Still, there is a tangible difference in the quality of applications for smaller languages compared to bigger languages like for instance English – and when it comes to regional dialects, a technological breakthrough has yet to come. There is a great need for better models for smaller language areas, something Barbara Plank is keenly aware of coming from a small minority group in Northern Italy with its own dialect.

In 2019, she received a Sapere Aude research leader grant from the Independent Research Fund Denmark for a project on broadening the application of NLP technology. “Take customer service conversation systems, for instance,” says Barbara Plank. “They work relatively well for English, German, and Chinese – for bigger languages – but languages like Danish, Swedish, or Icelandic have very little support yet. It is important to develop technologies so more people can benefit from NLP.”

Technology should always be human centred
Ultimately, her goal is to use deep learning technology to enable machines to learn languages faster and more accurately. This entails, teaching AI how to adapt to what is known as data shifts: underlying shifts due to variability in language use (shifting from one language to another or from one language style to another, like from formal to informal language).

“I think, in the long run, we need to focus more on making NLP technology human centred. It is ultimately an enabling technology. We must ensure that everyone benefits from it.”

On September 2, Professor Barbara Plank will present her inaugural lecture in auditorium 3 at 3:15 pm followed by a reception at 4:15 pm.

Further information