DeepSqueak neural network will learn the language of rats

The DeepSqueak algorithm developed by American scientists will be used to analyze ultrasonic signals made by rats. This will bring us closer to a better understanding of their communication patterns.\n

Science has traditionally treated rats as little more than good experimental subjects. This was mostly due to the striking similarities between rodents and humans across a broad range of genetic traits. But rats are also extremely intelligent animals that live in their own societies, which means that their behavior and communication tools are inherently interesting.

Researchers from the University of Washington reminded us of this fact, developing a neural network called DeepSqueak to study ultrasonic signals made by rats. These sounds are unintelligible to the human ear, but machines are equipped to analyze and sort audio recordings. This is a fairly long process: sorting and analysis take 20 times the length of the recording itself.

The neural network can identify certain regularities and patterns that could easily be a sign of a primitive syntaxis. This new development might