Abstract—In evolutionary linguistics, the Iterated Learning Model (ILM) is often used for simulating a first language acquisition. Whereas an infant agent acquires a grammar through communication with his/her parent in ILM, the length of syntax rules tends to increase rapidly over generations due to the addition of symbols of meaningless terminal symbols. In the case learning agents potentially have more than one teacher agent, this problem causes an unnatural learning, which results in a combinatorial explosion. In this paper, we propose a learning method in ILM to solve the problem by string clipping. Our experimental result showed that the length of utterances decreases without potential influence in intergenerational language propagation.

Index Terms—Language acquisition, simulation, iterated learning model, clipping.

Ryuichi Matoba, Hiroki Sudo, Shingo Hagiwara are with the National Institute of Technology, Toyama College, 1-2 Neriya, Ebie, Imizu-City, Toyama, 933-0293, Japan.
Makoto Nakamura is with the Japan Legal Information Institute, Graduate School of Law, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, 464-8601, Japan.
Satoshi Tojo is with the School of Information Science, JAIST, 1-1 Asahidai, Nomi, Ishikawa, 923-1292, Japan.

Cite:Ryuichi Matoba, Hiroki Sudo, Makoto Nakamura, Shingo Hagiwara, Satoshi Tojo, "Process Acceleration in the Iterated Learning Model with String Clipping," International Journal of Computer and Communication Engineering vol. 4, no. 2, pp. 100-106, 2015.