Maternal and paternal instinct is a wonderful thing and has more than sufficed to get homo sapiens through parenthood for a couple of hundred thousand years now. However, that doesn’t mean that new mothers and fathers always feel like they know what they are doing. Frustration can particularly set in when a new addition to the family is relentlessly crying for no obvious reason. Luckily, two hundred thousand years later, we now have an app to tell us why.
Originally developed by researchers at the University of California LA (UCLA), the original conception behind the ChatterBaby app was an aid to deaf parents, alerting them to when their baby is crying. Babies have different ways of crying depending upon what they are communicating, be it hunger, that they are tired, hot, in pain or simply want attention or the comfort of a hug. Often circumstances are such that it is obvious what a baby wants or is complaining of. Other times, especially for less experienced first-time parents, but certainly not exclusively, the cause of distress or need is less clear. For deaf parents, or a carer still not very familiar with an infant, an additional layer of complexity is added when it comes to translating a crying baby.
Ariana Anderson, the statistician, assistant professor of psychiatry and mother to three behind the ChatterBaby app came to the conclusion that the similarities she empirically spotted between the way her own three babies cried according to circumstances might well represent a wider pattern. If she was right and the cries of all babies demonstrate common traits depending upon why they are crying, she thought it was likely a baby-cry translation algorithm could be created.
Creating the database of cries involved recording crying babies under circumstances where the reason why they were crying was clear, such as cries of pain immediately following a vaccination. Other categories of cry the app translates are fussy, hungry, separation anxiety, colic, or scared. ‘Veteran mothers’ were enlisted to label cries and an audio sample only added to the database when they were in complete agreement. Cries of pain, for example, says Anderson “typically have louder, longer bursts and there’s very little silence between sounds.”
Extensive testing of the algorithm’s database-informed learning has demonstrated that it is able to decipher and translate any given baby’s cry with 90% accuracy.