On the Effectiveness of Speech Self-supervised Learning for Music
원문 초록
Self-supervised learning (SSL) has shown promising results in various speech and natural language processing applications. However, its efficacy in music information retrieval (MIR) still remains largely unexplored. While previous SSL models pre-trained on music recordings may have been mostly closed-sourced, recent speech models such as wav2vec2.0 have shown promise in music modelling. Nevertheless, research exploring the effectiveness of applying speech SSL models to music recordings has been limited. We explore the music adaption of SSL with two distinctive speech-related models, data2vec1.0 and Hubert, and refer to them as music2vec and musicHuBERT, respectively. We train $12$ SSL models with 95M parameters under various pre-training configurations and systematically evaluate the MIR task performances with 13 different MIR tasks. Our findings suggest that training with music data can generally improve performance on MIR tasks, even when models are trained using paradigms designed for speech. However, we identify the limitations of such existing speech-oriented designs, especially in modelling polyphonic information. Based on the experimental results, empirical suggestions are also given for designing future musical SSL strategies and paradigms.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Zero-note samba: Self-supervised beat tracking
Deeper Convolutional Neural Networks and Broad Augmentation Policies Improve Performance in Musical Key Estimation
JointRank: Rank Large Set with Single Pass
A Hierarchical Deep Learning Approach for Minority Instrument Detection
From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agents
Universal Music Representations? Evaluating Foundation Models on World Music Corpora
토론
이 논문과 연결된 포럼 글이 아직 없습니다.