On June 9, 2006, PED Seminar Series Presents
The Roles of Segmentation Ability in Language Evolution
by Kazutoshi SasaharaWe focus on segmentation ability as a prerequisite of language, studying the part it plays in language evolution. Language is mediated by distinct sounds and has a characteristic of 'duality'. For such a structure, it is necessary to have segmentation ability, which is an ability to discern discrete units in continuous sound sequences.
We model the evolution of discourse where agents utter strings by turns. Consider N agents in society. Each agent is modeled by a recurrent network (RNN) that studies the transition rate of sound elements in the sound strings it hears. The utterances of the agents consist of the outputs of their RNN and they are translated into letters (here, A, B, ...., Z), considered sound elements. When one agent utters a sequence of sounds, the other agents hears it one-by-one and predicts the next sound element in the utterance. After that, the agent's RNN is trained with supervised learning in such a way that it can better predict the transition of sound elements. Then, the agents take turns uttering and hearing. The procedures are repeated within a certain number of discourses. We demonstrate how common shared words (i.e. frequently used sound patterns) emerge and how the distribution of sound elements changes from a random initial state as common words increase.
Our results show that if agents have simple statistical ability, the frequently appearing patterns in sound strings may become established as words through the interaction of the agents and that emerging words may affect succeeding discourses in evolution.
upcoming PED seminars