Poultry Scientists are Using AI to Decode Chicken Speech

One of the first things you’re likely to learn when you raise a flock of chickens is that these birds are uniquely vocal creatures. Chickens are highly social animals that use a wide array of vocalizations to communicate with one another. From loud predator alarms to soft clucking sounds, researchers have cataloged at least 24 distinct chicken vocalizations, each of which has a distinct function and purpose.

Experienced poultry farmers tend to be pretty tuned-in to the language of their birds; by listening carefully to these vocalizations, farmers can often tell when chickens are stressed or sick. There’s a great deal of nuance in chicken speech, however, and as a result it can be difficult to reliably translate into terms humans can understand. Roosters, for example, react differently to predator threats depending on whether or not hens are nearby.

That’s why a group of engineers and poultry scientists from the University of Georgia and Georgia Institute of Technology have been working together to record chicken vocalizations and decode them using machine learning software. They might struggle with abstract reasoning, but artificially intelligent machines can reliably detect subtle differences in chicken vocalizations that human ears are unable to identify. The scientists hope that one day, their research will make it easier for farmers to respond to the needs of their birds.

“As a poultry farmer, you can always use more data to make better decisions,” said Wallace Berry, a poultry scientist at Auburn University in an interview with Scientific American. “This is a great way to continuously filter all the information available in a chicken house and learn as soon as possible that something is wrong.”

Modern poultry farms already utilize sensors to monitor the temperature, humidity and other environmental conditions in chicken houses. By adding audio sensors to their chicken houses, poultry farmers could be able to listen for signs of trouble as well.