To provide a conversation analysis method for specifying a person who will be the next speaker with high probabilities.

The conversation analysis method for analyzing voice data of each speaker in conversations acquires sentence data by using a conversation analyzer 1 having an input part 100, a buffer 105, and an analysis part 110 has steps of; inputting voice data of conversations among three or more speakers from the outside by the input part 100; storing voice data of conversations in the buffer 105; and segmenting voice data of speakers per prescribed unit from voice data of conversations stored in the buffer and analyzing a flow of conversations to specify a person who will be the next speaker with high probability, by the analysis part 110.