For our experiments we used a chat corpus collected by
Lin and described in [3]. It comprises chat posts from public
online chat rooms that were based on participant age
and not specific topic. Accordingly, the expectation was
that there would tend to be a wider variety of topics discussed
within these session without being bound to a certain
domain as would be the case in topic-based chat rooms
or IRC channels (e.g., channels about politics or computer
programming).