CONSTRAINT BASED COMPOSITIONAL SEMANTICS

摘要

This paper presents a computational system that handles the grounding,the formation, the interpretation and the conceptualisation of rich, compo-sitional meaning for use in grounded, multi-agent simulations of the emer-gence and evolution of artificial languages. Compositional meaning is de-constructed in terms of semantic building blocks which bundle a semanticfunction together with the relevant grounding and learning methods. Theseblocks are computationally modelled as procedural constraints, while thecompositional meaning is declaratively represented as constraint programs.The flexibility of the data flow in such programs is utilized to adaptively dealwith interpretation and learning. The conceptualisation is performed by asub-system that composes suitable constraint programs. The various meth-ods used for managing the combinatorial explosion are discussed.