JerusML Meetup #36: Artificial Melodies

JerusML continues its strike of face to face events, this time at the ancient city of David. We will have a very entertaining meetup that will be about the use of Machine learning tools in the music domain.

** All lectures will be held in Hebrew.

Agenda:
18:00 - 18:20: Gathering, networking, food and drinks

18:20 - 18:30: Opening remarks, Shimon Yitshak (JerusML)

18:30 - 19:10: “ Do Androids Listen to Electro Funk” Oded Zewi, DSP Lead @ JoyTunes

19:00 - 19:40: 'Approaches in song similarity given multi-layered data: lyrics, composition, production and sound' by Amir Graitzer Research Lead and Sivan Kollnescher VP R&D @ MyPart Inc

19:40 - 20:00: More Drinks & networking

More Information:

“Do Androids Listen to Electro Funk” - Oded Zewi, DSP Lead @ JoyTunes

Abstract:
At JoyTunes we inspire millions to realize their musical dreams. Our apps Simply Piano, Simply Guitar, and Simply Sing make music learning accessible, simple, and fun.
At the heart of our technology is the ability to provide learners with accurate, real time feedback on their music. This is done by sophisticated signal processing and machine learning algorithms, making estimations, predictions, and decisions within a super small delay.
In this lecture we will dive into the fast developing field of Music Information Retrieval (MIR), learn the basics of music processing, and discuss the benefits of combining data and domain knowledge for music ML.

Lecturers Bio:
Oded Zewi, DSP Lead at JoyTunes:
Oded leads the digital signal processing guild at JoyTunes, which is responsible for research and development of algorithms for music education. In his role, he builds systems for music recognition in acoustic signals, researches machine learning generation of musical content, and plays the trumpet. Oded is a graduate of Talpiot program, with a Bsc in physics and mathematics from the Hebrew University. In parallel to service at the Ministry of Defense, completed an Msc in physics at the Technion where he researched the use of compressed sensing to improve optical microscopic diagnostics.

Lecture by Amir Graitzer, Research Lead and Sivan Kollnescher VP R&D @ MyPart :

Nowadays, music search and analysis for various industry needs rely mostly on three aspects:
1. Users music consumption behavior
2. Metadata (e.g - who the artist is, in which decade the piece was released etc)
3. Sonic analysis - the piece’s genre, bpm, mood etc

While the above might be enough for some market needs, critical aspects of what makes up a musical piece are widely ignored. Such aspects are for example:
Composition - compositional aspects like harmony, melody, the song structure and more
Lyrics - what is the song about, and what lyrical aesthetics means it uses to portray this

In this talk we will share with you how referring to these aspects benefits with high-profile industry requirements, as well as some of the data and unique features we extract.
We will continue to discuss how this higher-resolution music understanding capabilities translates to deeper music similarity analysis, that utilizes purely mathematical methods, unsupervised & supervised techniques

Sivan Kollnescher: Over ten years of experience in artificial intelligence, cyber security, and web development. Having graduated from TAU with a BSc in Computer Science concentrating in Artificial Intelligence, Natural Language Processing, and Neuroscience. A music producer and guitarist with a music education from Rimon & Muzik schools and a background in Intelligence Corps.

Amir Graitzer: TAU Computer Scientist, ML and Statistical Analysis expert
Full Stack Developer
10+ years experience as music theory & guitar tutor