The results of Experiment 1 show that a meaning-based method dividing each microtask as short as possible is optimal for task division, the required minimum numbers of groups are 5 for Signed Japanese and Pidgin Signed Japanese
or 6 for Japanese Sign Language, and the necessary numbers of interpreters in a group are 3 for Signed Japanese and Pidgin Signed Japanese or 4 for Japanese Sign Language. The results of Experiment 2 show that different counters can
have the agreement on delimiters under the policy for task division that is determined in Experiment 1. The results of
Experiment 3 show our system can achieve a practical level of captioning for a continuous speech given by sign language
based on the setting and conditions determined in Experiment 1 and Experiment 2.