Datasets:
metadata
license: mit
task_categories:
- audio-classification
language:
- ja
- en
tags:
- music
pretty_name: NeuroPiano
size_categories:
- 1K<n<10K
Dataset Card for NeuroPiano-data
This dataset contains 2255 entries of audio-question-answer pairs that specializes in music education. Questions ranges from cleaness of attack to hands balancing, each one come with verbal response as well as a rating within 6. There are 104 unique student recordings of piano excercises and part of them are rated by multiple teachers.
Dataset Details
Dataset Description
- Curated by: Hayato Nishioka, Vincent Cheung, Huan Zhang, Shinichi Furuya
- Funded by [optional]: Japan Science and Technology Agency CREST grant [number JPMJCR20D4]
- Shared by [optional]: Huan Zhang
- Language(s) : Japanese, English
- License: MIT License
Dataset Sources [optional]
- Repository: [Coming Soon]
- Paper: [Coming Soon]
Uses
This is a great resource in analyzing performance and piano education.
Dataset Structure
Each entry contains:
- audio_path:
- array: loaded audio in 48000
- piece: name of the excercise. pdf score of the excercises is attached in repo.
- subject: the id of rater
- question: The question in Japanese
- answer: The answer in Japanese
- q_eng: The question in English
- a_eng: The answer in English
- score: numerical score that rates the question, in a scale of 6.
Dataset Creation
Who are the source data producers?
The pianists and raters in this dataset are student and teachers from the NeuroPiano Academy (https://www.neuropiano.org/).
Personal and Sensitive Information
There should no be any sensitive or personal information, as the identities are all anonimized in the dataset.
Citation
BibTeX:
@inproceedings{Zhang2024HowDataset,
author = {Zhang, Huan and Cheung, Vincent and Nishioka, Hayato and Dixon, Simon and Furuya, Shinichi},
booktitle = {International Society for Music Information Retrieval (ISMIR) Late Breaking Demo (LBD)},
title = {How does the teacher rate? Observations from the NeuroPiano dataset},
year = {2024}
}