3

We propose a novel methodology - Causal Disentanglement with Exteme Values (CDEV) - to identify representations learned by GANs. When trained on raw whale communication, it finds - for the first time - specific acoustic attributes that might serve as carriers of meaning.

Modeling the translation and review processes as zero-inflated fat-tailed distributions, we show how to extract useful information on translators’ reliability even with just one review per translation.