Random Walk 29: Cats and Crows with Sabrina Schalz


Have you heard about this? 

The cat’s meow: Ex-Amazon engineer launches cat speech translation app

Ok so first, every headline I found relating to this story made reference to the inventor as an “Ex-Amazon Engineer”… his name is Javier Sanchez. Wouldn’t it be great if scientists and researchers became household names like athletes? Places like France and Germany name streets after their scientists…

Just a second, let me hop off my soapbox here.

So the idea is that you have this app, and you let it listen to your cat’s meows, and it interprets the tone, pitch, etc, spitting out a human equivalent.

I think my favorite part of this particular article goes as such: 

Sanchez, a cat person, used his experience working on Amazon’s Alexa to develop MeowTalk as his pet project. He wanted to give cats more respect.

“You could meow at it all day. There’s no API for it,” he told GeekWire of Alexa. “They don’t care about cats.”

So, this made a decent amount of buzz near the end of 2020. About a week’s worth of buzz, I think mostly because it was 1) a slow news week in the world of science 2) a wholesome and cute story, and 3) about cats. But I was a little disappointed with the coverage about how it works. So here it is, right from the app description:

Each cat has their own unique vocabulary that they use to communicate with their owners consistently when in the same context. For example, a cat can have their own distinct meow for “food” or for “let me out.” This is not necessarily a language, as cats do not share the same meows to communicate the same things with each other, but we can use Machine Learning to interpret an individual cat’s meows and translate that into a human readable language. MeowTalk by Akvelon gives your cat a voice!

Using machine learning MeowTalk instantaneously translates your cat’s meows into one of nine general cat intents; these nine intents represent cat moods and states of mind and are as follows:

– Happy/Content

– Defense

– Angry

– Attack

– Mother Call

– Mating Call

– In Pain

– Resting

– Hunting

In addition to these nine general intents, each cat also has a unique vocabulary of meows. You can train the MeowTalk app to learn your cat’s unique vocabulary of meows by telling the app what each meow means when your cat makes it. When you give the app 5 to 10 examples of a specific meow for your cat (e.g. “food”, “let me out”) the app can start to recognize that meow when it hears it. The meow recognition is updated once a day so it may take up to 24 hours for the app to start recognizing the new word after you have provided the training information.

To train the app to learn a specific meow pick a context where you know with certainty what your cat is trying to say (e.g. at feeding time the cat is saying “food”, at the door the cat is saying “let me out”). Use the app to translate those meows. You will get one of the nine general intents as a translation, select “No, Change it” and then provide the correct translation. Repeat this 5 to 10 times then wait up to 24 hours for the app to update and learn your cat’s new word.

MeowTalk app on Google Play

So, this is a nice example of demystifying “machine learning”, since it basically requires you to be part of the machine that is learning. You need to feed the app plenty of audio from your own cat, and tell the app what the audio means. It has these presets, but as the app description explains, “there is no set language for cats” so you need to give the app plenty of reference material. AFTER that, it can start to recognize similar patterns in the audio you give it and match it back to the meanings you also gave it. 

It’s a fun concept, but not particularly groundbreaking. In fact, I am curious how much better the app can do compared to a human that really knows their cat. For example, could a human listen to the meows of a variety of cats and tell ANYTHING about what they are trying to express? How about the other way around? Can animals learn the subtle differences in human speech? Well, today’s guest studies exactly that. 

Sabrina Schalz is a PhD student in Urban Evolutionary Ecology at Middlesex University, focusing on the speech perception of wild Carrion Crows in London. However, in this interview, you will find that she’s broadened her scope to cats!




Thanks again, and I’ll see you next week!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s