Implementation of Biases Observed in Child Development into Concept Learning Agent

R. Taguchi, M. Kimura, S. Shinohara, K. Katsurada, and T. Nitta (Japan)

Keywords

Knowledge Acquisition, Intelligent Agents, Symbol Grounding, Learning Bias, and OnlineEM algorithm.

Abstract

This paper describes efficient concept acquisition for an infant agent (IA) based on learning biases that are observed in child language development. An IA acquires concepts through learning relations between visual features of objects and acoustic features of human speech. In this task, the IA has to find out which visual features are indicated by a speech. Previous concept acquisition systems find out them by using probabilistic methods, however, such approaches need much samples to achieve high accuracy. In this paper, firstly, we propose basic concept acquisition system using Online-EM algorithm without the biases. And then, we implement two types of learning biases to accelerate a learning process into our system. The experimental results show that the proposed method can achieve efficient learning.

Important Links:



Go Back