▽TOP▽活動実績▽2001年▽
(25) Evaluation of Spoken Dialogue Processing Method
Using Inductive Learning with Genetic Algorithm
Proceedings of the IASTED International Conference
ARTIFICIAL INTELLIGENCE AND SOFT COMPUTING,pp.231-236,2001-5
We propose the method that replies by response rules acquired automatically from spoken dialogue examples. In adding to dialogue examples, our method acquires response rules. Our method generates rapidly a suitable response using GA process(crossover, mutation, selection) and inductive learning. We call our method ELIZA-added GA-IL. We compared the ELIZA-added GA-IL system and the system without GA-IL response generation process. We confirmed that the total of a correct response rate and a semi correct response rate increased from 66.3% to 76.1% by using ELIZA-added GA-IL. Therefore, we confirmed that our method was an effective method for a chat.
|