Experiments with Chat Rooms and Bulletin Boards
In this example, we present an auditory display for three chat
rooms, each concerned with a different topic. For the sonification,
each different room is assigned a different pitch group,
reinforcing subject changes when they occur.
For each room, we build three streams of information:
- Topics: A continuously updated list of up to ten "topics" (the
most frequently appearing words and/or phrases mined from the chat stream);
- Content samples: A selection of sample sentences, identified
as typical or representative of the chat at that point in time;
- Content entropy: A vector that represents the changing level of
entropy in the source data.
The topics are spoken by a TTS system at regular intervals in a
pitched monotone, and are panned alternately hard left and hard right
in the stereo field, creating a sort of rhythmic "call and response."
The sample sentences are panned center, and rendered with limited
inflection (as opposed to the pitched monotone of the topics). The
tonal, rhythmic and spatial qualities of the topics contrasts
sufficiently with the sample sentences to create two distinctly
comprehensible streams of verbal information.
The entropy vector controls an algorithmic piano score. When entropy
is minimal and the discussion in the chat room is
very focused on one subject, chords are played rhythmically in time
with the rhythmic recitation of the topics. As entropy increases and
the conversation diverges, a Gaussian distribution is used to expand
the number, range and dynamics of notes that fall between the chords.
With this audio component, one can easily differentiate a
well-moderated content source from a more free-form, public chat
without distracting from the TTS annotations. The piano score also
serves a secondary function as an accompaniment to the vocal
foreground, enhancing the compositional balance and overall musicality
of the sound design.
A more formal description of the mechanism used to sonify these data is given
in our
ICAD 2001 paper.
|
Samples [Created 8/00]
|
Chat Sample
|
Three separate topics are contained in one file.
|
Timestamp:
00:00.00-00:01.47
|
The first portion of this sample concerns the recent recall of
Bridgestone tires and was based on a
www.cnn.com chat
room. This discussion was heavily moderated and hence the backing
piano score frequently reduces to a simple rhythm.
|
Timestamp:
00:01.47-00:03.21
|
For our second topic we recorded chat exchanges on
www.financialchat.com
one morning when Yahoo's stock opened low.
In this example, we hear day traders frantically exchanging
predictions about when Yahoo's stock will ``bounce.''
|
Timestamp:
00:03.21-end
|
The final topic
in this sample is again from
www.cnn.com
and treats a recent strike by the Screen Actor's Guild and the
American Federation of Television and Radio Artists. This chat room
was much less moderated than the previous CNN chat, and the backing
piano score reflects that.
|