Email Conversation Maps
In the past much more effort has been spent on visualising email conversation than visualising blog diffusion. Though it’s a different domain blog analysis can learn a lot from work previous work done in the field of email archives.
In both domains there are huge archives that reflect our interests and conversations with others. Default functionality like ordering archives by time or author and search functionalities are proper means for finding information, but there’s almost no functionality for browsing archives in order to explore personal developments, the cyclical up and down of interests or the path of conversation between two persons.
In the following I refer to a paper by Judith Donath, Fernanda B. Viégas, and Scott Golder “Visualizing Email Content: Portraying Relationships from Conversational Histories” in which they present a prototype called “Themail”. It is based on Salton’s TFIDF algorithm which compares the frequency of a certain word in a limited time span (e.g. a month) to its frequency in the whole archive. If there’s a relative high frequency it is displayed in a larger font in order to highlight its importance. This technique was used to analyse the email communication between the owner of the mail archive and a specific conversation partner over a series of months (see Figure 1). User testing, done by the authors, demonstrated that this form of analysis shows good results in characterising developments in the personal relationship and helps in assigning hot topics to specific months. E.g. in email conversations before a wedding, words like “invitation”, “tables”; “drinks” and guests names were used more frequent. Or a travel to Asia resulted in words like “Bangkok,” “thai” and “kuala”.
Figure 2 shows that after the return the conversation turned back to programming and other usual themes of conversation. Users like to explore their archive together with friends and conversation partners, because it tells a lot about the evolution of a relationship (like from classmates to lovers)
Similar analysis could be done for Weblogs; either for characterising the archive as a whole or single months or categories. Word frequencies of a weblog could be compared to overall frequencies in the blogosphere and could be output in word lists which describe the blog’s content. Phrases would do even a better job than single words. A system like Amazon’s SIPs (statistically improbable phrases) could probably describe a blog’s content more “objectively” that a tag cloud based on the author’s personal tagging. A similar idea has been posted by Rageboy (see also David Weinberger.)
Judith Donath, elaborated the Themail concept a step further. She mapped the email conversation between six researchers over a period of 22 days. The conversation took place in preparation of Janet Abram’s and Peter Hall’s book “else/where: mapping” (Donath called her prototype “The Rhythm of Salience” and it was published in the same book.) On the x-axis she distributed the names of the researchers; the y-axis mapped the flow of time. In the whole conversation only 30 messages were exchanged; They are displayed as full text. Visualisation aims to map temporal rhythms and the patterns of interaction between the participants. Each message is represented by a white square; thin white lines between the squares show the reply structure. Important words and phrases are highlighted in a similar manner like in Themail.
As in the case of Themail the usefulness of the map depends on the user’s relation to the mapped conversation. I’m quite sure that it is very useful for the people involved but for outsiders it provides only superficial information and is only useful in a bigger size and when printed out, in order to read the details. I also doubt that the length of the messages allow for reliable content analysis like SIP, in order to highlight important phrases. – The map was handmade and highlights were made by personal estimate. Nevertheless it is an interesting approach to map information diffusion and interaction; probably also useful for the blogosphere. It’s quite obvious that in such a map only a smaller number of messages can be displayed. For a bigger number of messages other forms of visualisations – like e.g. animated maps - might do a better job.
further information:
Rhythms of Social Interaction: Messaging within a Massive Online Network
Sociologically Improbable Phrases