Conversation map

From Driscollwiki

Jump to: navigation, search


Sack, W. (2000) "Conversation map: A content-based usenet newsgroup browse". Proceedings from IUI, New Orleans, pp. 233-240.

"How can all of the messages in an archive be graphically displayed and organized according to content of the messages and the social structure representative of the participants' interactions?" (239)

Contents

Introduction

Content-based browser, software that leverages computational linguistics and quantitative sociology to construct new interface to electronic communication

  • Especially large-scale conversations
  • Possibly a "better" USENET browser, esp. for sociologists, anthropologists

Conversation Map

Tech

  • Text analysis in Perl
  • GUI in Java

Graphical interface

  • Can be run on any newsgroup archive

Social networks

  • Nodes represent people
  • Links represent responses and/or quotations
    • Strength of connection rendered via inverse proximity (strong tie, short link)
  • Nodes are dragable
  • Selecting two nodes highlights the common threads between two users

Discussion themes

  • Lexical cohesion analysis
    • Not strictly speaking a "discussion theme" analysis, according to linguistics
  • Clicking on a theme highlights links among nodes in social network where two users have posted on this topic
    • Analysis of social cohesion, combo of lexical cohesion and social network

The messages

  • Messages are threaded and conceptually follow a tree structure
    • Original post is a root
    • Replies are branches
    • Replies to replies are sub-branches (236)
  • Tree structures may also be represented as "spider webs"
    • Original post at origin, reponses in concentric circles outward
    • Double-click opens a zoomed-in network diagram of the spider web

Semantic network

  • If 2 terms are linked, they have been "talked about" in similar ways (237)
    • Used with same verbs
    • Appear together with the same nouns
    • Share a large number of adjectives with which they are both modified (238)
  • Clicking on terms also highlights themes
  • Double-clicking on terms opens up "associations" from the archive
    • e.g. "Actions done by ___", "actions done to ____", etc.
  • Possible also to compare two terms

Text analysis (238-9)

  • Messages are threaded
  • Quotations are identified, sources are found
  • Signatures of posters are identified and distinguished from message content
  • Index of posters:messages is built
  • For each poster, list of all posters that replied is recorded
    • IDing reciprocal relationships
  • Messages broken down linguistically
  • Lexical and social relationships processed together

Related work

Jump offs

Personal tools