AI-driven advertising is often presented as one of the main successes of deep learning, after heavy investments in machine learning algorithms that discover and exploit patterns in consumer behaviour. In a new project financed the Marcus and Marianne Wallenberg Foundation, researchers from Umeå University, Malmö University, Stockholm School of Economics, and the University of Gothenburg collaborate to understand the implications for citizens and society linked to different types of targeting methods used in online advertising.
On Monday the 29th of September, Hannah Devinney presented the EQUITBL project at the AI for Good Breakthrough Days main stage event. The project is one of three winners in the Breakthrough Days’ Gender Equity Track.
This interdisciplinary project explores ways of combining qualitative and quantitative methods in order to explore and understand how bias and stereotypes manifest themselves in large text collections, such as those commonly used to train machine learning models in language technology. We also develop tools for mitigating the detrimental effects bias, stereotyping, and underrepresentation can have when the ML models are integrated into AI systems used for decision making.
The project members are:
Hannah Devinney, Computing Science, Centre for Gender Studies, and LPCN, Umeå University
Henrik Björklund, Computing Science and LPCN, Umeå University
Jenny Björklund, Centre for Gender Research, Uppsala University
Suna Bensch gave a presentation about social robots at Kafe Station in Umeå. Vetenskapsluncher is a series of popular science presentations organized by Umeå University (https://www.umu.se/nyheter/europa-i-fokus-pa-varens-livesanda-vetenskapsluncher-_7483184/). In her presentation she focused on language and dialogue processing techniques and difficulties for social robots.
Professor Thomas Hellström’s research deals with teaching robots to act and to communicate. One of his projects deals with developing methods for intention recognition from verbal utterances. Here’s a video about Thomas Hellström and his research.
On Wednesday, LPCN’s Anna Jonsson visited a group of young and fantastic women all enrolled in Teknikprogrammet at Dragonskolan. The mission was to recruit as many (future) PhD students as possible. The first step was to tell them what a PhD student is and does; the second step was to get them interested in choosing a computer science programme at the University after finishing high school. Sadly, Anna immediately scared them off by introducing formal grammars (as usual). At least the attendants got a great deal of fika for the trouble from the awsome woman (also named Anna) who organises these tech-for-girls meetings. Better luck with the recruitment next time!
LPCN member Suna Bensch blogged two weeks on the Umeå University website about Artificial Intelligence (AI). Her first blog post was about two almost opposing methodological approaches in AI, embedded into childhood memories about the autonomous car with language capabilities KITT.
Two junior members of LPCN, PhD students Adam and Anna, travelled to Charlottetown on Prince Edward Island in Canada in the middle of the summer for a vacationlike CIAA 2018. Adam made sure that they got there on time by running as a champion through Heathrow after a delayed flight. When safely at the conference, Anna presented A Comparison of Two N-Best Extraction Methods for Weighted Tree Automata (Johanna Björklund, Frank Drewes and Anna Jonsson). Moreover, both students showed great eating skills throughout the stay. Below follows a bunch of photos that hopefully prove to the concerned supervisors that their students actually went despite the heat wave.
On the 8th of June, Sara Stymne, from the Department of Linguistics and Philology at Uppsala University presented her work on learning dependency parsers from heterogenous treebanks in Umeå.
Three LPCN members gave much appreciated presentations of their work on the 9th International Workshop Weighted Automata: Theory and Applications (WATA 2018) which was held in Leipzig (Germany), May 22–26:
Johanna Björklund presented work in progress on The N-Best Problem for Weighted Hypergraphs over Complete Lattices (J. Björklund, F. Drewes, A. Maletti, A. Jonsson).
Frank Drewes gave a 2 x 90min tutorial on (Weighted) Regular DAG Languages – Properties and Algorithms.
Petter Ericson presented results on the Minimization and Characterization of Order-Preserving DAG Grammars (H. Björklund, J. Björklund, P. Ericsson)