Abstract
The anteater program is a method to acquire similar knowledge to experts, for example animal classification etc. This program acquires knowledge itself and can be also applied to other areas. And, this program does not have a built-in database. In this paper, an improved anteater program is proposed to narrow down main dishes potentially desired to be eaten as an association support system. The subjects firstly registered some dishes (main dishes) in the system as a process, which they had eaten before, and the impressions for each dish were recorded at the same time. Some questions were created to narrow down the dishes potentially desired to be eaten based on these impressions. As the subjects answered a series of questions, they could associate the dishes that they might potentially desire to eat, and this process continues until the name of the dish is finally concluded in the system. This system is a learning system. When the concluded dish name is not the dish the subject desires to eat, new questions are added for a binary tree so the subject can reach the name of the desired dish.
