ITU bachelor's system to identify informative Corona tweets in top rank
Competing with academics and IT professionals from all over the world, Anders Giovanni Møller shared first place in developing a system to identify Corona-related tweets.
The amount of tweets on Corona is incredible. While some tweets are untrue or misleading, others can be useful, for example in terms of helping us identify new local or regional transmission chains. In order to identify the useful tweets, however, we must be able to sort out the misleading tweets, and by applying efficient Natural Language Processing (NLP) methods we can do so. Therefore, the task at the international workshop Noisy User-generated Text (W-NUT) was a competition where the 55 participants from all over the world competed in developing the best NLP system to sort the wheat from the chaff.
Up against acknowledged researchers
Unlike many of his competitors – students and employees from universities and companies from all over the world – Anders Giovanni Møller had just received his bachelor's degree in Data Science before the workshop started, as part of Denmark’s first cohort of graduates at ITU. He showed a particular talent for using machine learning and NLP. With guidance from two of his teachers, Associate Professor Barbara Plank and Postdoc Rob van der Goot, the team managed to get ranked on a shared first place for the system developed by Anders.
Anders Giovanni Møller's two teachers are impressed by his approach to the competition:
- Anders did an amazing job on this project. He didn't only manage to answer an interesting research question, he also developed a state-of-the-art system for a timely problem. This competition drew strong teams from universities and companies from all over the world, including acknowledged researchers. It is a great accomplishment that one of our students reached a first place, says Rob van der Goot.
- Anders used his spare time during the COVID-19 summer to use NLP to help fight the pandemic. It is fantastic to see your BSc graduate get so motivated, work on an important problem, and doing so well at the international level, says Barbara Plank.
A taste for NLP
At the time of writing, Anders Giovanni Møller works as a Teaching Assistant at ITU and for the company Flowplan (where he also works with machine learning) while he waits for his candidate courses to start. He wishes to continue to focus on Natural Language Processing in his studies and future work and therefore he's been co-authoring a research paper on how to identify informative user-generated content based on his experiences in the competition.
- It's been great fun to work on these tasks. Previously, I tended to focus on photo and video but now it's text, which can be applied in so many different ways. There is a big potential in the machine learning technics that we used in our project, he says and adds that he would like to investigate how Natural Language Processing would work in a Danish setting: - Looking at Danish, the potential is huge. Most of the research in Natural Language Processing is focused on English. That's what everyone understands, right. But it looks like Danish hold a wide range of opportunities too and in the future, I would love to explore that.
Jari Kickbusch, phone 7218 5304, email firstname.lastname@example.org