摘要 |
This apparatus calculates, from the spoken dialogue data between a customer and an operator, the speech length of each speech. Then, it estimates the structure of the spoken dialogue data based on the speech length and specifies a speech-length-characteristic change time point (time) when the speaker who makes longer speeches changes. Then it extracts an information search time point (time) when the operator performs information search from operation log information about operator's operations performed during this conversation and determined both the ending time point of a part-to-elicit from the information search time point and the starting time point of the part-to-elicit from the speech length-characteristic change time point.
|