TUTORIAL 1 INTRO: Uncovering the communication of sperm whales
Recent advances in the field of neural networks enable translation between languages even in cases where no dictionary is available, such as ancient languages. The backbone of this capability is access to vast amounts of training data. Given this breakthrough, there has been a great deal of interest in recent years in the application of unsupervised machine translation (UMT) methods to decipher the communication of potentially intelligent animals, with most efforts focused on sperm whales, which have a unique communication system. However, to cope with the amount of data required for UMT, an automated tool to detect and annotate these signals is a prerequisite. In this talk, we will present a detector-based annotator for sperm whale communication signals. We will describe the main ideas and the theoretical formulation. Participants will then apply the annotation software to a sample recording of a short conversation between two whales. Finally, participants will be divided into groups and given time to try to match the annotated signals to their corresponding speaker based on the features extracted by the software. Teams that successfully complete this task will be able to uncover the “language-like” structure of sperm whales.