 I am Umberto and I'm a PhD student at the Insight Center for Data Analytics in UCD, and I work on music recommendations. So why? Because ever since I was a kid I had this really strong connection with music. I build many memories and share amazing moments with the people I love surrounded by music. I should like remember the first time I got a Waltman. I was like four or five. My dad used to put new music into it every 10 or 15 days, only 10 songs of course. Then I got an iPod, a place where I could have all my illegally downloaded music library in a single place. More than that, the iPod was like an open window to my soul. Just by looking at what playlists I had you could understand if I was going through a happy face, an angry teenage face, or even if I was in love. We don't have a personal music collection anymore. We listen to music through things like Spotify, where we have access to virtually every single song ever written. And that's great, but then the question is what do we listen now? Well, before I would go around to my friends and ask them What's the new cool band in town? What should I watch do like kind of fit in, right? Well right now I do kind of the same thing because I still want to fit in and be cool, but I do it with my phone. So my phone knows what I'm listening to and it tells me there's this new band They are playing in Dublin next week. Go check them out. That's great. How does it work? Well, basically it analyzes millions and millions of users and their behavior and some music meta data, magic algorithms working there. And it tells you because five million people listen to this song of Lady Gaga that you're listening right now. Your next song will be Wrecking Ball by Miley Cyrus. That's great. I love Lady Gaga, but I hate Miley Cyrus. Computers are not able to really understand that because they rely on this type of very simple information like they are both pop or they are both listened by people who like pop music. So when I embarked in this journey of research, I really wanted to make music personal again and somehow in the way, I kind of understood that music was about sentiment, it was about feelings and more importantly music is about people, right? So we developed this very simple, elegant and yet efficient model to classify music into moods or sentiments. We analyzed the lyrics of the songs, such as the ones you see here, and we are able to understand which words are related to particular sentiments. So we know if a song is hot or if a song is happy. With the previous approach that everyone uses right now, Spotify uses, you are not able to distinguish between two indie and acoustic writers. They might seem really similar, but actually they are not. The top one is very sad and melancholic, whether the one at the bottom is actually really happy, catchy and lovely. And the only way we can find this type of relations and make personalized music recommendations to you, which are cool and interesting, is doing this type of approach. Thank you very much.