Skip to content
printicon
Main menu hidden.
Staff photo Kalle Prorok

Kalle Prorok

I do research in AI, especially self-learning systems where the computer learns from data or experienced experiences - it is called machine learning.

Works at

Affiliation
Location
X, Håken Gullessons väg 20, Teknikhuset, TB413 Umeå universitet, 901 87 Umeå

Right now, 2020-2021. I am researching how to generate texts, such as answers to questions on large documents, such as Environmental Impact Assessments on 200 A3 pages, make summaries or to facilitate applications. The studies take place in collaboration with the Swedish Transport Administration / Norrbottniabanan and Sweco. We use a method called Natural Language Processing, NLP, which is a very hot topic where Facebook, Google, Microsoft and others are major players and we use their free software. We have concentrated on the special case Swedish with the help of the Royal Library.

Deep Learning is a revolutionary way of handling images, sounds and more and there I have created a popular course and also teach about autonomous vehicles.

With graduate students, we have, among other things, won the European Championship in robot football in Paris and then competed in Seattle. The robot dogs (Sony Aibo) were completely on their own and not remotely controlled.

Previous research

In a research group, we have studied how forest machines can be made self-propelled. In my dissertation I studied how computers can learn to cooperate and then for simplicity in a game (the card game Bridge) with the technology Multi-agent Reinforcement Learning (RL).

IEEE International Conference on Robots and Agents, IEEE conference proceedings 2009 : 447-452
Rönnbäck, Sven; Westerberg, Simon; Prorok, Kalle

Research projects

1 March 2021 until 30 November 2021
1 January 2002 until 31 December 2011