Researchers at Eotvos Lorand University in Hungary are using computers to help decode and translate different dog barks.
The study began with recording the barks from fourteen different Hungarian sheepdogs. Barking was recorded in six different situations:
- When a stranger approached the house when the owner was away and the dog was home alone.
- When a trainer encouraged the dog to bite on protective gear and bark aggressively (think of police dog training).
- When an owner was preparing to take the dog for a walk.
- When an owner was holding a ball in front of the dog.
- When an owner was playing tug-of-war or a similar tug game with the dog.
- When the owner tied the dog to a tree in a park and walked out of sight.
Scientists recorded more than six thousand barks and used artificially intelligent software to process the results.
The computer program could correctly identify the type of barking forty-three percent of the time — that’s slightly more than the forty percent humans scored in the study. However, humans and computers excelled in different areas. Humans were better at identifying barks for play and barks for loneliness than the computer. The program could single out barks for walk and ball more accurately than humans.
Score one more for the computer program — it could identify which dog made which bark more than half the time. Most humans in the study couldn’t do that well with the fourteen dogs of the same breed.
These results suggest that computers can help people comprehend animal communication — especially because the computer program picked up on things that the humans didn’t and vice versa. Researchers on the study are hopeful that this could work with vocal signals from any animal, not just dogs.
Future studies are planned using different breeds.