Lora Aroyo

Lora Aroyo (born in Bulgaria) is a Dutch computer scientist and professor at The VU University of Amsterdam, Netherlands, best known for her work in user modeling, digital humanities, and for the Crowd Truth crowdsourcing method.^[1] She is the head of the user-centric data science (UCDS) research group in the VU Computer Science department, president of the User Modeling Society,^[2] a former vice-president of Semantic Technology Institute,^[3] Chief Scientist at Tagasauris, and a member of over 100 scientific programme committees and editorial boards. She is one of the few female full professors of Computer Science in the Netherlands, and one one of only two women (out of over 30 full professors) at the VU.

After leaving Bulgaria during the aftermath of the fall of communism, Aroyo obtained a Phd in Computer Science from Eindhoven University. She worked with Paul de Bra in the area of intelligent tutoring systems, focusing primarily on understanding and modeling the different needs of users to make the experience more productive.^[4] Her work in user modeling grew in reputation and she chaired the primary conference in the area in 2010 and became president of the User Modeling Society in 2016.

Aroyo moved to the VU University in 2008 and began her seminal work in cultural heritage, later called digital humanities.^[5] With colleagues Guus Schreiber and others, she is credited with pioneering niche-sourcing,^[6] a method to motivate small communities of crowd workers who are passionate about specific topics. Often these communities are hobbyists, amateur scientists, etc., who enjoy sharing their knowledge and expertise, especially with public institutions such as museums. Aroyo led several successful campaigns with the Rijksmuseum using niche crowds to generate large amounts of metadata about the Reichsmuseum collections on line.

In 2013, Aroyo spent her sabbatical working with the IBM Watson team, just after the famous Jeopardy! match, where she developed the Crowd Truth methodology with Chris Welty.^[7] Artificial intelligence systems are trained and evaluated using human-labeled data, Aroyo and Welty observed that when the people labelling the data disagree on what the right answer is, this disagreement may signal some other problem, such as ambiguity or vagueness. This is considered important, because most human-labelled data ignores the difference between high and low agreement examples.

In 2018, she was listed among the top women semantic web researchers without a wikipedia page, and was nominated to have this page authored during the Ada Lovelace women in computing hackathon,^[8] as a way to combat Gender bias on Wikipedia.

References