Anyone who convinced with a pet, probably would like to find out what is hidden behind the tatters, mugolii and megolii of their four -legged friend. For a simultaneous translation it is still early – even if you are working on – but there is a new model of artificial intelligence that you get closer. In fact, it promises to decode the verses of a wide range of animals and to understand if they express positive or negative emotions.
The details of the discovery were recently described in a study published on Scientific Reports From Stavros Ntalatti, researcher of the University of Milan. And they represent the last frontier in a field, that of the decoding of the languages and verses of the animals, which in recent times is experiencing an incredible acceleration thanks to the arrival of artificial intelligence and machine learning.
Projects such as Ceti (Cetacean Translation Initiative), an international initiative based in New York, are actively at work to translate the social meanings that are hidden in the intricate songs of the whales, in the whistles of the dolphins in human terms (it is the goal of the Google Dolphingemma model), and in the verses and in the behavioral repertoires of many other social animals, such as elephants, dogs, chimpanzees and others species of large monkeys.
The model developed by Ntalattiras does not aim to so much: it is designed to demonstrate that there are clues contained in the tone, frequency and other acoustic characteristics of animal verses, which can help recognize their emotional content in different animal species (seven have been studied in the research). The verses emitted in correspondence with negative emotions, for example, have shown to tend more often towards medium or low frequencies, while those related to positive emotions would be distributed more uniformly in the acoustic spectrum.
The algorithm has also shown the presence of species-specific differences in studied animals: for pigs, it is the high frequencies to prove particularly useful to understand their moods, while in sheep and horses it is the average frequencies to transport more information.
The result of the work of Ntalattiras is therefore not an app for the translation of animal verses, but an experimental model that has shown the presence of salient characteristics in animal vocalizations that can represent a clue of their mood. And it’s not a little. In fact, it could open the doors to the development of programs with which to monitor the state of health and the level of stress of breeding animals, those kept in the zoo, and also to monitor the behavior of wild specimens more effectively.