SUGAR

The Sugar Dataset is a corpus of spoken commands, consisting of 2293 audio files produced by 36 different speakers and annotated by two linguistic experts for the action expressed. Actions are annotated as a finite set of predicates accepting an open set of parameters. For example, the action of putting may refer to a pot being placed on the fire and expressed with the predicate “put(pot, fire)”. The training set consists of 1721 utterances (and therefore 1721 audio files) produced by 36 different speakers annotated by two linguistic experts. The test set consists of about 572 audio files containing uttered commands without annotations. The dataset was created and used in the context of the SUGAR task (https://sites.google.com/view/sugar-evalita/), organised as part of the EVALITA 2018 evaluation campaign (http://www.evalita.it/2018). The task challenged participants to train a voice-controlled robotic agent to act as a cooking assistant.