LingLunch : Timothée Bernard

Timothée Bernard (LLF)
Studying the syntax and semantics of emergent languages

In signaling games, two or more agents (human or otherwise) have to cooperate through information exchange in order to retrieve some target. In particular, we start with an experimental setup in which one agent has to describe an image using arbitrary symbols so that another agent can recognise it among a set of distractors. The agents are implemented as neural networks and their training is based on reinforcement. Our long term goal is to understand under which conditions the language developed by the two agents, similarly to natural language, exhibits a syntax and a compositional semantics. As first steps, we introduce an array of automatic metrics designed to uncover syntactic and semantic aspects of the messages exchanged by successful pairs of agents, and we study experimentally how various factors impact the communication of high-level concepts (vs image-specific feature).