Who among us has not wished that our dogs could talk to us, clearly expressing how they feel and what they want? We all want to understand our dogs better and it’s exhilarating to imagine that we could translate what dogs say, either vocally or with body language. Combining behavioral research with artificial intelligence just may be the path forward towards this goal.
Biologist Con Slobodchikoff, author of the book Chasing Doctor Doolittle: Learning the Language of Animals, is an expert in animal communication. He has spent decades studying prairie dogs, and has shown that their system of communication is so sophisticated that it deserves to be considered a language. For example, their alarm calls can convey information about the size and type of predator. There is evidence that these calls are even nuanced enough to describe the color of clothing worn by a person.
In a collaboration with a computer scientist, Slobodchikoff developed an algorithm to express those vocalizations in English. The goal of his new company, Zoolingua, is to develop similar algorithms that will “translate” animal sounds, movements of the body and facial expressions a variety of species.
One of the first projects of the company was to develop an algorithm that would determine if a sheep was in pain based on the positions of the mouth, eyes, and ears. Research about sheep facial expressions guided the process. (A related system can be used to evaluate whether a sheep is happy or sad.)
GET THE BARK IN YOUR INBOX!
Sign up for our newsletter and stay in the know.
Similarly, in order to understand dogs better, Slobodchikoff is collecting thousands of videos of dogs in which they vocalize or communicate in other ways. He uses these videos to develop a translation algorithm based on these signals. People’s interpretations of the meanings of these signals are part of the process of developing a translation. That means that the output (the translation) is only as good as the input (what the humans think the signals mean). As research on the meaning of animal signals advances, so will the accuracy and precision of the algorithms.
Interpretations of meaning when changing languages are always problematic, even for humans—as anyone who has ever used Google Translate knows. Something really does get lost in the translation. When languages are not just within a species but between different species, the potential for inaccuracies and mistakes increases.
A truly thought-provoking discussion with our dogs may be forever out of reach. However, greater knowledge about how dogs communicate will allow progress towards a deeper understanding of what they are saying. That’s a prospect that sounds great in any language.