CS Colloquium - Davide Mottin:Are smart devices really smart? An algorithmic road towards smartness

OK Google, I would like to visit a city like Aarhus". A seemingly simple question like this finds only disappointing answers from modern smart devices.

2019.03.01 | Dorthe Haagen Nielsen

Date Fri 15 Mar
Time 15:15 16:00
Location Building 5335, room 016 (Peter Bøgh Auditorium)

The real problem is that the devices cannot interpret what similar means in this context, and therefore resorts to matching word-by-word the question posed by us.

Interestingly, into the heart of these devices lay what is called a knowledge graph, a network of entities, such as cities, people, objects connected through labeled relationships. The question's answer is obtained by physically matching words in the questions with entities and traversing such graph with graph algorithms. Clearly, the original question cannot be obtained by simple matching, but rather treating "Aarhus" as an example of other cities to find. 

Such an example-based view displays a significant shift in the ways current algorithms retrieve answers from the graph, challenging our ability to predict the user intention. Treating users as first-class-citizens and redesigning algorithms to be interactive, explainable, efficient, and adaptive has driven most of my research starting from my Ph.D. 

This talk is my personal exploratory journey in the landscape of algorithms that support principled and easier user intentions to obtain better answers.  

As we are surrounded by smarter devices, wouldn't be worth investing also time into smarter algorithms? 

Inaugural lecture

On October 1, 2018, Davide Mottin started as an Assistant Professor in the Data Intensive Systems group. Davide's research focuses on graph exploration, which lays on the broad areas of database, data mining, and machine learning. In particular, Davide enjoy working and understanding graphs with or without labels interactively, understanding phenomena and users in a principled way. Before coming to Aarhus, Davide was a Research Associate at the Center for Geoscience (GFZ) in Potsdam and a Postdoctoral Researcher at the Hasso Plattner Institute, in the group of Knowledge Discovery and Data Mining working with Emmanuel Müller. He got his PhD in Computer Science at the dbTrento group in the University of Trento under the supervision of Themis Palpanas and Yannis Velegrakis.

The lecture will be followed by a reception.


CS frontpage, Site Specific, Reception, Debate, Future students, Private and public companies, Public/media, Staff, Students