A Universal Translator for Dog Language
Computers are being used to attempt to translate canine communication sounds.
Posted Dec 04, 2018
A colleague of mine recently came by to ask me a question. He had just returned from a short professional visit to Israel. According to him, the facility he visited had a security system which included guard dogs as well as human personnel; however, the dogs were connected to a computer which supposedly translated the sounds that the dogs made so that the human guards could determine the level of threat, and perhaps even the nature of the threat that they were observing. My colleague wanted to know whether our knowledge of canine language had advanced to the state that we now had an electronic translator which could interpret what our dogs were trying to say.
I told him that a number of years ago I became interested in the possibility of using computers to interpret canine communication. I suspected that the situation that he was referring to might have links to the research and applications that I had reviewed at that time, since the idea of electronic translation of dog language as part of a security system was first introduced by Eyal Zehavi, the founder of Bio-Sense Technologies in Israel.
Using dogs as guards is nothing new, since it is understood that dogs have better night vision than humans, plus significantly better senses of smell and hearing. However, when Zehavi was working on military bases, he found that that the human security personnel do not always understand what a particular dog's bark means. Guards often dismissed the dog’s barks as unimportant or insignificant—perhaps simply a response to a cat or another dog passing nearby, and because of those misinterpretations, break-ins still occurred.
Zehavi’s research lab used computers to analyze 350 different dog barks and his summary of their findings is that "Dogs have a specific bark when someone threatens their space. It doesn't matter what breed of dog they are, how big or small, or what sex, all that matters is that they bark in response to a threatening situation. An alarm bark is always the same."
On the basis of this search, Bio-Sense Technologies created the DBS, which stands for Dog Bio-Security System and can be integrated into existing security systems. The "language" input is from a sensor mounted on the dog’s collar that conveys the information to a receiver where a human guard may be monitoring a number of dogs. The data is filtered and passed through an analytical process to determine the state of the dog. This can include additional channels of information, such as the dog’s heart rate, in order to make interpretation clearer and avoid false alarms. The results are translated into three levels of alert, something like the national security system does using colors. Here we might have a “Code green” or “no alert” status, meaning that the watchdog is responding to routine events, a “Code orange” or “medium alert” meaning that there is suspicious activity in the vicinity, and a “Code red” or “high alert” meaning that an intrusion has occurred into the dog's territory.
Although Dog Bio-Security Systems can cost many thousands of dollars, it is typically still only one quarter the cost of video surveillance systems. At the time that I looked into this system, it had already been found to be sufficiently accurate and sensitive to do the kind of security work required. Information from the company and media reports showed that it was being used in Israel’s high-security Eshel Prison as well as some Israeli military bases, research laboratories, and water installations, as well as smaller installations for farms, ranches, and garages, and apparently it had also been installed in some of the Jewish settlements on the West Bank.
When I learned about this use of computers to interpret canine communication, I spoke to a colleague of mine in the linguistics department at my university. During the course of our conversation, he mentioned to me that a former student of his was working on an electronic translation system that was much more grand and elaborate. It was supposed to ultimately be the equivalent of the “universal translators” depicted in many science fiction shows like Star Trek—namely, a computer-based device to translate communications from any language, even if you could not recognize which language the message was in. The theory behind the research is that in all human language systems, the sounds that indicate emotions have certain similarities, and the sounds associated with certain people, events and things have notable commonalities. For example, the word for “mother” in many languages has a prominent “m” sound, as in the French Mere; Italian Madre; Serbian Majka; Dutch Moeder; Estonian Ema; Russian Mat; Greek Mana; Hindi Maji; Hawaiian Makuahine; Urdu Ammee; and Swahili Mzazi.
My colleague then mused: “Since all animals use much the same variations in sound quality in their communication, I wonder what would happen if a sample of dog sounds was put into the translator?”
I got in touch with the researcher who was working in a government research facility in Bethesda, Maryland. The idea of trying to translate the language of dogs intrigued him. He added some cautionary notes, however.
“First, you must understand that the translator is a work in progress and is still far from perfect,” he explained. “It does not interpret single words, per se, but rather tries to put them in the context of the word sounds before and after them. It then comes up with a set of candidate words, and selects the one that it feels is most likely. This means that you often get ungrammatical sentences, but at the stage that we are now at, you can at least pick out the general meaning of what the person is saying. Because it requires a context, I need an utterance that goes on for at least 30 seconds to work on.”
Thirty seconds is a long time for a continuous bit of dog language, however, there was one dog that I had lived with for a short time that “talked” all the time. She was a Cavalier King Charles Spaniel named Amy, who became mine when she was well past seven years of age. She was to be a gift for my parents who wanted an adult dog that already had some basic training. Amy stayed with me for nearly a month while I arranged for her transportation from Vancouver to Philadelphia. She was the ultimate dog orator, and would walk around the house giving canine monologues—combinations of barks, whines, little yodeling sounds, low-key growls, and so forth. Just for fun, I had taped one of her “speeches” to play over the phone for my parents with a comedic explanation that this was Amy’s list of what she wanted them to do to prepare for her arrival.
I dug up the tape, which for some reason I had kept, and sent it off to be translated. A week or so later I got an email, which said, in part, “When I put it through the translator at first it didn’t make any sense to me. However, when I looked at it a while it became clear that it wasn’t just straight prose, but was really poetry. So I took the liberty to parse it out the way I think it goes.”
This is Amy’s poem:
Happy closets speak falling biscuits to lazy cats.
Shadow puppies dance sunshine music.
Warm dreaming goes loudly through days.
I’m not convinced that there is anything valid in this “translation”; however, a free-form poem about biscuits, cats, puppies, and daytime naps could well be what an old dog might sing about as she wandered happily and “loudly through days.”
Copyright SC Psychological Enterprises Ltd. May not be reprinted or reposted without permission.