Summarizing Text Conversations - Methods for Mining and Summarizing Text Conversations - page 80

￿ From Joannie: Erica, Due to the fact that Jeff is unable to attend on July 19, I believe it would be
better to reschedule the call for sometime next quarter.
￿ From Erica: Joannie: As I mentioned, Jerry is in transit in Europe at the moment, and it will be
extremely difficult for me to get hold of him again today to reschedule. Given the time difference
involved I will not be able to contact him before the appointed time tomorrow, and therefore I'd very
much appreciate if we could go ahead with the call as planned. At least that way Mr. Skilling can
decide which of the dinners he would prefer to attend, I assume that the more notice you have the
better.
￿ From Sherri: Erica, thanks for the note. Joannie has left the office for the day, but will return
tomorrow morning. In the meantime, I'll run this by Jeff to see if he thinks it makes sense. Do you
have any dates for future dinners in mind? The vast majority of Jeff 's time is committed through
February 2002, so knowing what the dates/timeframes are would be most helpful in the event we
need to try to free up some time. Thank you.
￿ From Erica: Sheri: Thanks for the response. Our next dinner is scheduled for September 25, at
the moment I'm not sure who the other invitees would be. Jerry would probably have a better idea
as he puts together the guest list for each event and may already have something in mind. That's
actually another reason why I think we should leave the call tomorrow on the calendar. Depending
on Mr. Skillings availability/interest Jerry might want to reconfigure his invitee list for the most
appropriate mix of people.
Because we have chosen a relatively brief email for purposes of explication, summarization
might not strictly be necessary here. However, other email threads are much longer and feature more
complicated thread structures. Before we even consider summarization and the question of salience
or importance, there is a variety of dimensions on which we can characterize this conversation. We
can enumerate the actual participants in the discussion, in this case Erica, Joannie and Sherri. We
can identify a number of topics such as phone calls , travel and dinners . We can extract dates ( July 19 ,
September 25 ) and named entities ( Europe , Aspen , Insight Capital , etc.). We can observe that there is
an initial email that contains a great deal of information, and several shorter follow-up emails. We
can detect sentiment and opinions, such as Joannie and Erica disagreeing about whether the phone
call should go ahead. Essentially, we can enlist all of the mining techniques described in Chapter 3
in order to derive some structure from this conversation. As previously stated, many of those mining
techniques themselves can be considered a type of focused summarization, where particular types of
information are being extracted, and a large part of a general summarization system may consist of
combining those pieces of information.
Given the email and the derived information, how might one summarize this conversation
for a third party? The answer partly depends on the audience. The two people who are the focus of
the conversation, Mr. Murdock and Mr. Skilling, did not take part in the discussion, and a summary
generated for the benefit of Murdock may differ from the summary provided to Mr. Skilling. The