New dataset Chordonomicon offers 666,000 song chord progressions

Researchers have introduced Chordonomicon, a new dataset comprising over 666,000 song-level symbolic chord progressions. This dataset is significantly larger than existing ones and includes annotations for structural parts, genre, and release dates. The accompanying research also proposes a benchmark suite for next chord prediction, evaluating recurrent neural network, gated recurrent unit, and long short-term memory architectures. Experiments indicate that structural annotations enhance prediction performance, and the dataset is available as an open benchmark for future music understanding models. AI

IMPACT Provides a large-scale dataset and benchmark for advancing AI models in music understanding and generation.

RANK_REASON The item is a research paper detailing a new dataset and benchmark for music information retrieval. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Chordonomicon
• DagsHub
• gated recurrent unit
• Gotit.pub
• Hugging Face
• long short-term memory
• recurrent neural network
• ScienceCast
• Spyridon Kantarelis

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →