Information Technology Reference
In-Depth Information
Direct search does not always work. For ex-
ample, the song
Smells Like Teen Spirit
by Nirvana
appeared first on the 1991 album
Nevermind
, then
on numerous compilations. If MusicStory looks
up
Smells Like Teen Spirit
lyrics from Nirvana's
self-titled 2002 compilation, our lyrics database
will return no results. Similarly, Tori Amos's cover
of this song may not also return any results. When
a search failure is encountered, the agent performs
a roll back strategy, dropping the album from the
query. If no results are returned again, the artist
is dropped from the search key and only the song
is used. This ensures the most exact lyric match
can be found. If the song is not in the database,
we prompt for lyrics. As search extraction im-
proves, it is possible to “screen scrape” the lyrics
from a general-purpose search using engines like
Yahoo! Search. A similar rollback strategy would
be equally applicable in this situation.
finding pace
When building a Network Arts installation, one
must remember the audience. It is important the
audience be engaged and connected with the
installation and its performance. Keeping this
connection will allow the piece to create more
emotional and captivating moments with the
viewer, following Kandinsky's model of expres-
sionism (Kandinsky, 1994). Foremost, the agent
must determine the pacing of the installation
and hence performance. There are two pacing
metrics, the tempo of the source (input) media
and the desired tempo of the overall (output)
performance.
The tempo of a slow ballad does not match
that of a live speech or a fast hip-hop song. Mu-
sicStory bases its rate for presenting images on
the pace of the media. To accommodate several
media sources, we created a model of presentation
for the agent. The model's presentation pace is
set to complement the pace of the source media.
As a result, an effective flow state for the overall
installation is achieved. To keep the flow state
engaging, thresholds are set to keep the images
from changing too quickly or too slowly, which
prevents the audience from being overwhelmed
or becoming bored.
A simple implementation of Mihaly Csikszent-
mihalyi's flow model (Csikszentmihalyi, 1990)
suffices for our purposes. While, Csikszentmih-
alyi describes human activities as a compromise
between two components, challenges and skills,
we focus on the flow channel itself and the neigh-
boring anxiety and boredom outside the channel.
MusicStory needed a descriptor of the pace of the
song to implement our modified Csikszentmihalyi
pacing model. In a previous Network Arts instal-
lation, which used broadcast media, Shamma et
al. (2004) estimated pace by using the rate of
the closed captioning feed. Given the input rate
and knowledge of how fast information can be
displayed (output rate), Csikszentmihalyi's flow
finding Images
MusicStory searches for images from a user's
social network space. While many users keep
photos on their personal computers, often they
are unorganized, uncommented, and untagged.
We use Flickr to find images within the personal
network of an individual. First, the lyrics are
stripped of common stop words (common words,
“if”, “he”, “the”, and so forth, in the English and
Spanish languages). It then searches for each
salient term in the lyrics, using Flickr's public
API. MusicStory looks for the photos which match
each term by searching tags and comments. If the
personalized (user-specific) search returns no im-
ages, MusicStory then begins searching through
the user's Flickr contacts, first searching contacts
marked Family then contacts marked Friends
and finally, upon failure, a general Flickr search
is performed. Like the lyric search, MusicStory
uses this rollback strategy to ensure an image is
found for each candidate term, while providing
the most personalized relevant image.
Search WWH ::
Custom Search