by audioset

audioset /ontology

The Audio Set Ontology aims to provide a comprehensive set of categories to describe sound events.

441 Stars 106 Forks Last release: over 2 years ago (v1) 5 Commits 1 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:

Audio Set Ontology

Dan Ellis [email protected]

v1 2017-03-06

The file ontology.json contains the current definition of the AudioSet ontology, a hierarchical set of audio event classes.

The json file describes a list of sound entity objects. Each object contains the fields:

  • id
    : The machine identifier for this class, a short sequence like
    . Where possible, these are based on Knowledge Graph IDs, which were originally used in Freebase.
  • name
    : The Display Name that refers to the class. Should be unique and unambiguous. Several classes have one or more synonyms included in their display names, for example "Male speech, man speaking".
  • description
    : A description of the class in a few lines.
  • citation_uri
    : A pointer to any text used as the basis for the description. Frequently points to a Wikipedia page.
  • positive_examples
    : A list of compact URLs to segments within YouTube files that provide confirmed positive examples of this class. Each entry is of the form
    , meaning the example is the 10 seconds starting 30 seconds in to the the YouTube video whose ID is
  • child_ids
    : A list of the
    fields for any classes that children of this class in the class hierarchy.
  • restrictions
    : A list that can include the following values:
    • abstract
      for a class that is principally a container within the hierarchy, but will not have any explicit examples for itself. "Human voice" is an abstract class. Abstract classes will always have children.
    • blacklist
      for classes that have been excluded from rating for the time being. These are classes that we found were too difficult for raters to mark reliably, or for which we had too much trouble finding candidates, or which we decided to drop from labeling for some other reason.

The ontology is made available by Google Inc. under a Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.