Biology and Compositionality: Empirical Considerations for Emergent-Communication Protocols

Abstract

Many researchers in language origins and emergent communication take compositionality as their primary target for explaining how simple communication systems can become more like natural language. I suggest that, if machine learning research in emergent communication is to take cues from biological systems and language origins, then compositionality is not the correct target. * http://travislacroix.github.io 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada.