Advertisement

Home / Science-tech / Purrfact: Decoding an app that tries to translate cat-talk

Purrfact: Decoding an app that tries to translate cat-talk

US start-up Zoolingua is hoping to create an artificial-intelligence- powered dog translator that will analyse canine vocalisations and body language
NYTNS

Emily Anthes   |   Published 12.09.22, 03:53 AM

My cat is a bonafide chatterbox. Momo will meow when she is hungry and when she is full, when she wants to be picked up and when she wants to be put down, when I leave the room or when I enter it, or sometimes for what appears to be no real reason at all. 

But because she is a cat, she is also uncooperative. So the moment I downloaded MeowTalk Cat Translator, a mobile app that promised to convert Momo’s meows into plain English, she clammed right up. For two days I tried, and failed, to solicit a sound.

On Day 3, out of desperation, I decided to pick her up while she was wolfing down her dinner, an interruption guaranteed to elicit a howl of protest. Right on cue, Momo wailed. The app processed the sound, then rendered a translation: “I’m happy!”

I was dubious. But MeowTalk provided a more plausible translation about a week later, when I returned from a four-day trip. Upon seeing me, Momo meowed and then purred. “Nice to see you,” the app translated. Then: “Let me rest.”

MeowTalk is the product of a growing interest in enlisting additional intelligences — machine-learning algorithms — to decode animal communication. The idea is not as far-fetched as it may seem. For example, machine-learning systems, which are able to extract patterns from large data sets, can distinguish between the squeaks that rodents make when they are happy and those that they emit when they are in distress.

Applying the same advances to our creature companions has obvious appeal.

“We’re trying to understand what cats are saying and give them a voice,” said Javier Sanchez, a founder of MeowTalk. “We want to use this to help people build better and stronger relationships with their cats.”

To me, an animal lover in a three-species household — Momo the cranky cat begrudgingly shares space with Watson the overeager dog — the idea of a pet translation app was tantalising. But even MeowTalk’s creators acknowledge that there are still a few kinks to work out.

A meow contains multitudes. In the best of feline times — say, when a cat is being fed — meows tend to be short and high-pitched and have rising intonations, according to one recent study, which has not yet been published in a scientific journal. But in the worst of times (trapped in a cat carrier), cats generally make their distress known with long, low-pitched meows that have falling intonations.

“They tend to use different types of melody in their meows when they try to signal different things,” said Susanne Schötz, a phonetician at Lund University in Sweden who led the study as part of a research project called Meowsic.

And in a 2019 study, Stavros Ntalampiras, a computer scientist at the University of Milan, demonstrated that algorithms could automatically distinguish between the meows that cats made in three situations: when being brushed, when waiting for food or when left alone in a strange environment.

MeowTalk, whose founders enlisted Ntalampiras after the study appeared, expands on this research, using algorithms to identify cat vocalisations made in a variety of contexts.

The app detects and analyses cat utterances in real-time, assigning each one a broadly defined “intent”, such as happy, resting, hunting or “mating call”. It then displays a conversational, plain English “translation” of whatever intent it detects.

MeowTalk uses the sounds it collects to refine its algorithms and improve its performance, the founders said, and pet owners can provide in-the-moment feedback if the app gets it wrong.

In 2021, MeowTalk researchers reported that the software could distinguish among nine intents with 90 per cent accuracy overall.

I found that the app was, as advertised, especially good at detecting purring. But it’s much harder to determine what the calls in each category mean — if they carry a consistent meaning at all — without actually having a way of, you know, communicating with cats. (Cat-ch-22?)

Dogs could soon have their own day. Zoolingua, a start-up based in Arizona, US, is hoping to create an artificial-intelligence- powered dog translator that will analyse canine vocalisations and body language.

Dog owners have been overwhelmingly enthusiastic about the concept, said Con Slobodchikoff, founder and CEO of Zoolingua, who spent much of his academic career studying prairie dog communication. “Good communication between you and your dog means having a great relationship with your dog,” he said. “And a lot of people want a great relationship with their dog.” (However, he added, not everyone: “One small minority says, ‘I don’t think that I really want to know what my dog is trying to communicate to me because maybe my dog doesn’t like me.’”)

Still, even sophisticated algorithms may miss critical real-world context and cues, said Alexandra Horowitz, an expert on dog cognition at Barnard College. For instance, much of canine behaviour is driven by scent. “How is that going to be translated, when we don’t know the extent of it ourselves?” Horowitz said in an email.

The desire to understand what animals are “saying”, however, does not seem likely to abate. The world can be a lonely place, especially in the next few years. Finding new ways to connect with other creatures, other species, can be a much-needed balm.

NYTNS



Advertisement
Advertisement
Advertisement
 
 
 
Copyright © 2020 The Telegraph. All rights reserved.