Web navigation is a curious universe. First of all we talk about the web, but there are no spiders. (There are virtual viruses, though.) Then we mention ‘navigation’, but there are no sailing boats. Yet we all navigate across the web’s virtual oceans in a sophisticated labyrinth of entangled entrances and exits. This is where information travels at the speed of…. your Internet connection.
It is pretty nerve-racking when your favorite video clip does not load in a reasonable time frame on Spotify, isn’t it? Or when there is a random ad popping up in the middle of the indie-rock hit that you are passionately singing along with. Or even when YouTube suggests you hear some less than thrilling tracks, which happen to start right after your favorite band and kill your good mood. At this stage you ask yourself: “How does this music platform recommend music to a random listener?” Sometimes the suggestions are aggressive, but they’re not too bad. But other times they are just a mood wrecker. How is this possible? Fate? Not really! Data analysts are behind the scenes here.
In music platforms, we all experience loads of publicity and tracks suggestions. In my PhD, I’m working on elaborating algorithms that will suggest songs that only reﬂect your music listening habits -- in other words, those that reﬂect your musical tastes and the type of music you seem to appreciate. If you are an eclectic listener, I have a wider margin of suggestions to play with. If you listen only to classical music, that narrows down my proposals. Some other studies take into account information such as online purchases, social media likes (Facebook likes, for example) and even personal information as the user’s location or details on his/her age or genre. Such approaches may ameliorate the aggressiveness of the propositions -- that is, the mathematical models that the play books are based on -- but these researchers may find themselves invading the delicate world of online privacy. In my case, my ultimate goal is to improve user satisfaction by making adapted music recommendations and respecting your privacy as much as possible.
What I really enjoy in my thesis is the fact that I work directly with fun applied mathematics. I have to understand the users, analyze their musical behavior and find optimal solutions to help them have a great time. In order to achieve this goal, I work with the human/machine interface, which accounts for human decision processes and psychology. Humans are very complex, and it is exciting to be able to figure out their needs -- at least in terms of information!
I am Amaury L’Huillier, a PhD student working at the LORIA laboratory of the University of Lorraine. My project is funded by the university’s research department and the Lorraine Region along with Grand- Nancy.