In this paper, we develop an algorithm to mine message boards, e.g., Yahoo! Clubs. Each message is evaluated by three indices:
(i) the degree of inclusion of given information, (ii) the degree of inclusion of new information, and (iii)the degree of effect for the successive messages. We show these indices are useful for characterizing a message. We also show a prototype system of summarization applied to a message board, as an example of application of the indices.