Interdisziplinäres Zentrum für Kognitive Sprachforschung

Links und Funktionen



Abstract De Beule

Fluid Construction Grammar

The past few years have seen significant progress in the understanding of the dynamics and mechanisms by which autonomous agents can self-organize a communication system through iterated peer-to-peer interactions. Most of the experiments carried out so far were centered on purely lexical languages. In order to carry this research even further we need a formal theory of language and language processing that is powerful enough to deal with the complexities found in natural languages in a computational manner. For this reason we have been developing Fluid Construction Grammar (FCG.).

Although developed from the viewpoint of Artificial Intelligence and with the main goal of building working systems in the form of autonomously learning and communicating robots, FCG shares many things with other flavors of construction grammar and other unification-based feature-structure grammars. But it also exhibits some more or less unique characteristics like bidirectional rule application (for parsing and production), openended semantic and syntactic categories, fluidity in rule application, mechanisms for learning and creating new rules of language and, recently, a solid connection to a grounded procedural semantics called IRL (Incremental Representation Language.) Although many issues still need to be addressed, FCG has become a fully operational grammar formalism and implementation for representing, learning and applying lexical and grammatical inventories. We therefore believe that FCG opens many new research directions for linguists, especially those interested in cognitive and computational linguistics. In particular, at the moment we are performing experiments in which robotic agents need to communicate the location of objects in the world. They are equipped with FCG on top of a procedural semantics and low level sensor-motor skills. It is shown that in order to maximize their communicative success and to reduce ambiguity, the agents evolve a recursive grammar that allows perspective-marking.



Steels L., De Beule J. and Neubauer N. (2005). Linking in Fluid Construction Grammar. Proceedings of BNAIC-05, royal Flemish Academy for Science and Art.

De Beule J, Steels L. (2005). Hierarchy in Fluid Construction Grammar. Proceedings of the 28th German Conference on Artificial Intelligence. Berlin: Springer.

Steels, L. and Bleys, J. (2005). Planning what to say: Second order semantics for Fluid Construction Grammars. Proceedings of CAEPIA ’05. Lecture Notes in AI. Berlin: Springer.

Steels L., De Beule J. (2006). Unify and merge in Fluid Construction Grammar. Proceedings of the Third International Symposium on the Emergence and Evolution of Linguistic Communication. Berlin: Springer.

De Beule J. and Bergen, B.K. (2006). On the emergence of compositionality. Proceedings of the 6th Evolution of Language Conference, Rome, 12-15 April 2006.

Steels, L. (2006). How to do experiments in artificial language evolution and why. In Cangelosi, A., Smith A. and Smith K. (eds). Proceedings of the 6th International Conference on. London: World Scientific Publishing.

Steels, L. (2006). The recruitment theory of language origins. Proceedings of the 6th evolution of language conference, Rome, 12-15 April 2006.