Using human listeners to identify LENA-identified target child vocalizations that deserved that label.
One thing that is unique about this project is that it goes through all LENA-identified target child vocalizations (called "CHN" by LENA) within a day-long file, in order. This allows for taking some history and context into account while also being very quick, as no utterance boundaries need to be placed, only one judgment is made per sound, and no sounds other than probable target child vocalizations are included.
Currently, the instructions given to listeners for the relabelCHN task (which is the main task) are to indicate simply whether the sound contained any type of target child vocalization.
To facilitate sharing the data, we'll prioritize labeling the publicly shareable HomeBank recordings, and because this is related to our lab's interest in studying exploration/exploitation/foraging dynamics of infant vocal production, we will focus first on the Warlaumont Corpus data.
Author: Anne S. Warlaumont