Can we teach AI to feel music? The challenge of building reliable datasets for music emotion recognition

Speaker: Patrick Donnelly, Associate Professor, Computer Science

Music has a remarkable ability to evoke emotions, from joy and excitement to sadness and nostalgia. But can artificial intelligence learn to recognize these emotions, and how do we know whether it gets them right? Music Emotion Recognition (MER) uses machine learning to identify the emotional qualities of music, with applications in music recommendation, personalized playlists, and music therapy. Yet progress in this field faces a fundamental challenge: the scarcity of large-scale, high-quality datasets for training and evaluating AI models. Because emotions are inherently subjective, assigning reliable emotional labels to music is challenging and existing datasets may not accurately capture the diversity of listeners’ emotional experiences. In this talk, we explore novel techniques for creating large-scale music emotion datasets and discuss how these approaches can help advance AI’s ability to recognize emotion in music.

Oregon State University-Cascades | Website | View More Events

OSU-Cascades Event Services