🇳🇴 Tromsø Uncovered: Abdera – Machine learning to tackle online trolling

By Arctic Business Journal July 10, 2023
641

This is the third article in a new series by Pro Tromsø and the Tromsø Chamber of Commerce, highlighting northern Norway’s unique innovative capabilities through a number of interviews with cutting-edge start-ups and the people behind them, profiles of leading companies, and insight into the entrepreneurial work being undertaken in the capital of the Arctic. 

The third article profiles Abdera’s efforts to combat online hate speech and trolling using algorithms and models to solve the problem in a simple and cost-effective way. This article features Abdera’s four founders, Johannes H. Nilsen, Erland Gromstad, Jonas N. Myhre and Markus Tiller, all of whom have backgrounds from Tromsø-based UiT – The Arctic University of Norway.


The Abdera team (from left): Johannes H. Nilsen, Erland Grimstad, Markus Tiller and Jonas N. Myhre. Photo: Lars Åke Andersen

Social media and the internet have become omnipresent in recent years. In line with the development, the spread of hate speech and offensive comments has increased at an alarming rate. Abdera AS wants to contribute to doing something about that.

Traditional methods of moderating comments sections are very labour- intensive and not always effective enough, says the CEO and Co-founder of Abdera, Johannes H. Nilsen.

“We are working to create algorithms and models that can solve the problem in real-time, and in a simple and cost-effective way,” he says.

Abdera’s four founders, who alongside Nilsen are Erland Gromstad, Jonas N. Myhre and Markus Tiller, all have backgrounds from UiT – The Arctic University of Norway. Abdera was founded in May 2022 and, at the time of going to press, the first version of the software is in test production at two local newspapers – Nordlys and iTromsø – and the local football club, Tromsø IL.

“The starting point for Abdera and our technology was a research project where the objective was to recognise logical flaws in debates. Since then, we have taken the technology further,” says Nilsen, who has a background in economics.

But what exactly is the software looking for?

Legal definition

“We base our model on the legal definition of hate speech,” explains Nilsen.

“But the model can of course be adapted to any area of application and trained to follow the rules that the specific publication wishes to enforce,” he explains.

Like all other machine learning systems, Abdera’s model must be trained on data sets from comments sections and research. The aim is to train the model to recognise different language and special orthography that often occurs in comments sections.

“For instance, many people use orthography that reflects their own dialect, or simply have poor spelling. The model will capture that,” says Nilsen.

When the model detects unwanted comments, the comment may be flagged for further manual follow-up or deleted automatically. The machine learning model continues to learn, even after it has been adopted.

“That’s one of the advantages of machine learning. The model gets better and better as it’s used because it learns along the way,” says Nilsen. “So it may be natural to use the software as a support tool to begin with and then switch to more and more automation.”

“Of course, that’s entirely up to the newspaper or publication that uses the tool.”

Important for democracy

It may seem like only online newspapers and magazines have use for such a tool, but Nilsen believes that is not the case.

“Hate speech and offensive comments also appear on the pages of banks, online stores and indeed everywhere where people can leave comments. So the potential market for our products is enormous,” he says.

“It doesn’t take much to translate the tool into other languages, so our ambition is to go international as soon as possible.”

”The model gets better and better as it’s used.”

Abdera’s vision is to create a system that safeguards freedom of expression in a positive way and contributes to supporting a democratic public debate, says Nilsen.

“The minority destroy what could be a very good democratic tool. We wish to contribute to putting an end to that.”

Local and national support

Abdera has received support from Norinnova in Tromsø through the Startup IX programme for entrepreneurs as well as a grant from Innovation Norway. Abdera was also one of 10 businesses selected to take part in Norinnova’s Arctic Ignite, which is a programme to promote exciting start-ups with financial means and market exposure. Abdera was one of three awarded a jury prize. The business is now based in Norinnova’s incubator.

“The support has made it possible for us to get the company started. It’s very important and positive for us to have this environment around us when it comes to both starting capital and competence,” emphasises Nilsen.

“Several start-ups are gathered here. Some are in an early phase like us, while others have come further. There is always a lot to learn, and here the threshold for asking is low. It’s a very inspiring and positive environment, he says.

Now Abdera and Nilsen are looking for capital.

“We want to grow. That requires capital and we are working continuously to get that in place, he concludes.


This article was originally produced as sponsored content by Pro Tromsø in their appendix from the Norwegian Business Journal, produced in cooperation with the Tromsø Chamber of Commerce.

This article was originally published in Norwegian and has been translated by Pro Tromsø.

To join the Arctic Business Journal network as a sponsored content partner, contact us at [email protected].