Language Evolution and Computation Bibliography

We focus on the ``compositionality'' of semantics, a fundamental characteristic of human language, which is the ability to understand the meaning of a sentence as a combination of the meanings of words. We also pay much attention to the ``embodiment'' of a robot, which means that the robot should acquire semantics which matches its body, or sensory-motor system. The essential claim is that an embodied compositional semantic representation can be self-organized from generalized correspondences between sentences and behavioral patterns. This claim is examined and confirmed through simple experiments in which a robot generates corresponding behaviors from unlearned sentences by analogy with the correspondences between learned sentences and behaviors.

Lexicon learning systems need to be concerned with more than just producing symbol usage agreement between agents, which is easy to acquire through imitation. Lexicon learners should also explicitly attempt to increase the mutual information between their symbol usages (a measure of the usefulness of the symbols for transferring information between agents). This paper argues that, although many lexicon learning algorithms presented in the literature do attempt to create highly informative symbol usages implicitly, there are good reasons to make the mutual information of symbol usages an explicit goal of the lexicon learning system. Some first steps in this direction are provided in this paper. It presents lexicon learning experiments using both purely imitative and explicitly information maximizing algorithms. The results of these experiments are used to support the thesis of this paper, that lexicon learning algorithms should explicitly attempt to produce high mutual information symbol usages.