抄録
A grammar of natural language is a result of statistical analyses, and is different from that of formal language. Thus, the grammar should be represented in a flexible, changeable formalism. In this paper, we propose a model of common language acquisition in multi-agents. Agents are divided into two groups, each of which has its own language. They try to communicate each other, especially with those which are in the different language group. Becuase the grammars of the languages are represented in the mutable form, the agents change their own grammars and temporarily build up a common grammar, that is of an artificial ‘pidgin,’ that enables the communication. We applied Genetic Algorithm to the statistical mutation of grammar, and utilize Lexicalized Tree Adjoining Grammar that tolerates the change of word-order. In this formalism, we could observe the phenomenon that the majority group assimilates the minority group, preserving grammatical features of each language.
