You probably believe you know what your cat is telling you, but now you can use machine learning to confirm it.
Javier Sanchez, group technical program manager at the Seattle-area company Akvelon, leads a team that’s created MeowTalk, an app that translates feline vocalizations into human speech.
The app is premised on the notion that cats typically have nine different “intents” that they express through meowing.
“These nine intents represent cat moods and states of mind,” the app’s description says. “But each cat also has its own unique vocalization and vocabulary of meows that goes beyond these nine general intents.”
These messages are pretty basic, such as “Feed me,” “I’m in love!” and “I’m in pain.” Akvelon’s goal is to employ machine learning so the app can identify the specific meows of an app user’s cat. The cat’s people can help along the machine’s study habits by inputting information about their pet — that is, recording their cat and telling the app what they believe the sounds mean.
He hopes to get the technology to the point that it will be useful even without an owner’s input, so that, as the app states, “if you don’t have your own cat, or you don’t have time to create its profile, or maybe you just met a cat on a walk or at neighbors, you can still translate any cat’s meows with the modified recognition algorithm.”
Sanchez hopes eventually to create a collar that will turn a cat’s meows into audible human words in real time. (He should start compiling a database of expletives now.)
IPhone and Android owners can download and use MeowTalk, but it remains in development. And it appears the app’s developers do indeed still have work to do. The app’s user rating so far on Google Play and the Apple Store is 2.8 (out of 5).