The SCHAEFFER dataset (Spectro-morphogical Corpus of Human-annotated Audio with Electroacoustic Features For Experimental Research), is a compilation of 1000 raw audio files accompanied by human annotations and morphological acoustic features. The audio files adhere to the concept of Sound Objects introduced by Pierre Scaheffer, a framework for the analysis and creation of sound that focuses on its typological and morphological characteristics. Inside the dataset, the annotation are provided in the form of free text, while the labels are pre-chosen from a list of classes, making the sound description fit into a suitable framework for digital analysis. All the sounds within the dataset are under a "CC-By-4.0-attribution" license.
The dataset can be found at the following links - Huggingface Kaggle