Hanna Y.’s research and web app that combine her two passions—computer science and music—have earned her a travel grant to present at the International Society for Music Information Retrieval conference in Suzhou, China.

 

Hanna, a senior at Spence, said she was incredibly surprised and excited to learn in August that she had won the travel grant to go to China from the Women in Music Information Retrieval, which is a society that seeks to increase opportunities for women in the field.

 

“I submitted [my paper] because my mentor showed me this opportunity, and I thought I might as well give it a shot. But I wasn’t expecting anything at all,” Hanna said, noting that the conference typically accepts contributions from researchers with advanced graduate degrees.

 

As a research assistant at the NYU Steinhardt Music and Audio Research Laboratory for the past three years, Hanna has worked closely with her mentor, Rachel Bittner, a Ph.D. candidate. Last fall, Hanna started working on her own research project on machine learning and training a computer to recognize individual musical instruments in an audio file. Similar to the app Shazam, which listens to a portion of a song and tells the user the name of the song and the musician, Hanna’s creation listens to a song and can help distinguish the sound of a violin from a piano, for example. Her model can recognize 18 instruments, including the banjo, distorted electric guitar and synthesizer.

 

There had been previous work on this kind of tool, but Hanna noted that they were not open-source files so it was hard for researchers to use it. Now, someone who needs a perfected instrument classifier can find Hanna’s web app on Github and download it to tackle tasks such as automatically organizing a large collection of recordings.

 

“I basically dropped myself into this area,” she said. “I had no idea how to build a web app or anything like that, so it was a process of self-teaching and getting help from my mentor. But the end result was that I got a web app that works to identify musical instruments and a better understanding of how machine learning works in general.”

 

Hanna said she appreciated having the independence to research and figure out problems on her own; she said one of her biggest takeaways was how to teach herself to learn. She also gained experience writing a technical research paper (which was co-authored with her mentor). Hanna explained that research papers are different from lab reports in that they are not so much about the steps involved as the big picture of the project. She also learned more about what a conference expects from a paper, including how her research builds on previous research and improves it.

 

“This knowledge I’ve gained about coding and computer science is going to be invaluable for whatever field I go into, whether it be finance or physics or anything,” she said. “So I’m super excited to have this experience in a research lab early so I can use it in the future.”

 

Hanna will be in China for three days in October for the conference and will present her paper and app in the Late-Breaking Demo session, which is a part of the conference where new ideas and research are shared with the music technology community. Researchers from around the globe and individuals from companies such as Spotify and Pandora attend to learn what’s new in the world of music information retrieval.

 

Before this experience, Hanna said she wasn’t sure how or if she could combine her interests in music and computer science. She said she saw almost no connection between the two fields.

 

“They seemed like opposite fields because music is all about emotion and the arts and trying to convey what you can’t with words, whereas computer science is a lot about trying to put into numbers and code what the human brain thinks,” she said. “But after doing all this research, I found that music—because it’s so nuanced in its interpretations—is the perfect field to be approached by computer science.”