Self-Organized Linguistic Systems: From traditional AI to bottom-up generative processes

Abstract

This work seeks to explore the potential of bottom-up generative processes in the context of conlang production, aiming to describe the basis of a new field of research: Self-Organized Linguistic Systems or SOLS, specified under the perspective of both self-organized systems and constructed languages. SOLS approach provides a framework for the creation of self-generated artificial languages and may serve as a starting point for the development of context-dependent or domain-specific languages. It acknowledges that the development of conlangs can happen in artificial societies of simple agents, as the output of social interactions in computational simulations under the agent-based modelling paradigm.In the proposed initial SOLS model, automatic generation of lexicon takes place in the context of a digital environment with objects, actions and agents with embodied cognition through peer-to-peer interactions. Specifically, this paper exposes how SOLS can be developed with bi-dimensional games and simulations. An initial work has been done with the xmunch-atomspace and the SciArt simulator, which constitute the first implementations of both our knowledge representation toolbox and our bi-dimensional simulator of P2P Social Dynamics. Non-interactive agent-based SOLS can allow artificial agents to independently evolve emergent languages as part of their self-organizing or adaptation processes.