The research, published Thursday in the journal, Scientific Reports, explains how they did it. First, the team spent 75 days recording two groups of 11 bats held in separate cages. The team then went through the video footage to suss out which individuals were squeaking at each other, what they were squeaking about -- food, sleep, perch or sex (or lack thereof) -- and the ultimate outcome of the argument. Finally, they trained the machine learning algorithm with 15,000 calls from seven adult females using those variables.
In the end, the algorithm managed to correctly identify the bat making the call (compared to the video footage) 71 percent of the time, the subject of that argument 61 percent of the time and the eventual outcome 41 percent of the time.
"What we find is there are certain pitch differences that characterise the different categories - but it is not as if you can say mating [calls] are high vocalisations and eating are low," Yossi Yovel, co-author of the study, told The Guardian. "We have shown that a big bulk of bat vocalisations that previously were thought to all mean the same thing, something like 'get out of here!' actually contain a lot of information."
The team hopes to further decipher bat language based on the squeak patterns and inflections. Eventually this research could reveal insights into not just bat behavior but also how human language itself evolved.