Dolphin whistle data involves different parameters, such as maximum frequency, number of inflections, duration, and intensity. The analysis of such parameters is challenging, as they are used by the animals in the context of various behaviors. On the other hand, their study should increase the knowledge about dolphin communication, which is an essential means of these animals’ survival. This presentation displays an investigation on whistle data from bottlenose dolphins that live in Southern Brazil. Their community has a few hundred individuals, with two groups showing two distinct foraging behavior: some animals may forage in cooperation with fishermen, while others always forage independently from humans. Our data analysis involved an artificial intelligence tool (self-organizing maps, or SOM) that uses artificial neuron networks and unsupervised learning. In this investigation, we applied this tool searching for clusters that could show patterns within the data. Using an 8x8 net, we noted subtle variations among the whistles. Through an original approach, we pinpointed particular whistles that could be key to distinguish the two foraging behaviors.
Magalhaes et al. (2025) studied this question.