MACS - Multi-Annotator Captioned Soundscapes

This is a dataset containing <strong>audio captions </strong>and corresponding<strong> audio tags</strong> for a number of 3930 audio files of the TAU Urban Acoustic Scenes 2019 development dataset (airport, public square, and park). The files were annotated using a web-based tool. Each file is annotated by multiple annotators that provided tags and a one-sentence description of the audio content. The data also includes<strong> annotator competence </strong>estimated using MACE (Multi-Annotator Competence Estimation). The annotation procedure, processing and analysis of the data are presented in the following papers: Irene Martin-Morato, Annamaria Mesaros. <em>What is the ground truth? Reliability of multi-annotator data for audio tagging, </em>29th European Signal Processing Conference, EUSIPCO 2021 Irene Martin-Morato, Annamaria Mesaros. <em>Diversity and bias in audio captioning datasets, </em>submitted to DCASE 2021 Workshop (to be updated with arxiv link) Data is provided as two files: <strong>MACS.yaml</strong> - containing the complete annotations in the following format: - filename: file1.wav <br> annotations:<br> - annotator_id: ann_1<br> sentence: caption text<br> tags:<br> - tag1<br> - tag2 <br> - annotator_id: ann_2<br> sentence: caption text<br> tags:<br> - tag1 <strong>MACS_competence.csv</strong> - containing the estimated annotator competence; for each annotator_id in the yaml file, competence is a number between 0 (considered as annotating at random) and 1 id [tab] competence The audio files can be downloaded from https://zenodo.org/record/2589280 and are covered by their own license.本数据集包含TAU城市声学场景2019开发数据集（TAU Urban Acoustic Scenes 2019 Development Dataset）中3930个音频文件的音频字幕（audio captions）与对应音频标签（audio tags），涵盖机场、公共广场与公园三类场景。所有音频文件均通过基于网页的标注工具完成标注，每个文件由多名标注者进行标注，标注内容包含音频标签与一段描述音频内容的单句说明。数据集还包含使用MACE（Multi-Annotator Competence Estimation，多标注者能力评估）方法估算得到的标注者能力（annotator competence）数据。 本数据集的标注流程、数据处理与分析方法详见以下两篇论文： 1. Irene Martin-Morato、Annamaria Mesaros. 《何为真实标签？音频标注多标注者数据的可靠性》，发表于第29届欧洲信号处理会议（EUSIPCO 2021） 2. Irene Martin-Morato、Annamaria Mesaros. 《音频字幕数据集的多样性与偏差》，已提交至DCASE 2021研讨会（后续将补充arXiv链接） 数据集以两个文件形式提供： - MACS.yaml：包含完整标注信息，格式如下： - 文件名：file1.wav 标注项： - 标注者ID：ann_1 句子：字幕文本 标签： - tag1 - tag2 - 标注者ID：ann_2 句子：字幕文本 标签： - tag1 - MACS_competence.csv：包含估算得到的标注者能力数据；对于YAML文件中的每个标注者ID，其能力值为介于0（随机标注）与1之间的数值，格式为制表符分隔的标注者ID与能力值。 音频文件可从https://zenodo.org/record/2589280下载，其使用需遵循对应授权协议。