Stories and social networks - Narrative Intelligence - page 310

of messages, but no one responds to those messages, then that someone will
not show up in the social network.
2. Themes : The upper middle panel is a menu of discussion themes. Themes
listed at the top of the menu are those themes that are most commonly used in
the conversation. The list of discussion themes is extracted from the archives by
examining the words and synonyms of words in quotations and replies to pre-
vious messages. In linguistics, this analysis is properly described as an analysis
of lexical cohesion between messages (see Halliday & Hasan 1976). The links be-
tween participants in the social network are labeled with the discussion themes
from the menu of themes.
3. Semantic Network : The upper right-hand panel displays a semantic network.
If two terms in the semantic network are linked together, then those two terms
have been found to be synonyms - or terms that may have similar meanings -
in the conversation. The semantic network is produced through the application
of corpus-based linguistics techniques referred to in the literature as techniques
of “semantic extraction” and “automatic thesaurus construction” (cf., Hindle
1990; Hearst 1992).
4. Message Threads : The panel that occupies the lower half of the window is
a graphical representation of all the messages that have been exchanged in the
newsgroup conversation over a given period of time. The messages are orga-
nized into ”threads,” i.e., groups of messages that are responses, responses to re-
sponses, etc. of some given initial message. The threads are organized chrono-
logically, from upper-left to lower-right. The oldest messages can be found in
the upper left-hand corner.
For a newsgroup devoted to a television program, the computed themes
and terms in the semantic network often include names of characters and
episodes from the television show. Thus, these are the pieces of the televi-
sion story that one can empirically observe as being appropriated into and
employed by the audience's discussions of the story. Obviously, with a more
sophisticated set of computational linguistic analysis tools one might observe
larger portions of the narrative structure being woven into the audience's dis-
cussion (e.g., like the sorts of appropriations observed by Jenkins (1992)).
However, the set of computational linguistic procedures we employ and have
developed expressly for our system are more sophisticated than any others
compared to contemporary, computational work on the social and linguistic