Social Network and Relationship Finder (SNARF)

I've been following MS sociologist Marc Smith's work for a few years now. In a nutshell, Marc's task is to mine data for socially-relevant information, and he's written some interesting papers on virtual communities.

Some time ago I was glad to hear that his focus was turning from the analysis of Usenet messages to e-mail in general. In particular, there are characteristics of communities and message threads that indicate quality (signal-to-noise), the sender's role in the community (leaders, learners, flamers, etc.), and the type of communication taking place (announcements, debates, Q&A, etc.). Some of the metrics used are total numbers of subscribers, contributors, and return visitors, frequency of posts, length of posts, length of message threads, and the number of contributors to a thread.

For example, threads with many messages where each contains little new content are easy to categorise as high noise. Threads usually begin with questions, and end with resolutions, so if a single user tends to be among the last 2-3 posters on a variety of threads, that person will tend to be an expert answering questions. If messages in a given community rarely receive replies, you can assume that it's either geared towards announcements or file distribution (if you find repeat contributors), or simply an unhealthy Q&A forum (many first-time contributors, few return customers).

Now point these same principles at e-mail, and you can really save people some time. I often talk about what a black hole e-mail has become for most companies, and how a great advantage of SharePoint is the ability to cut down on e-mail simply by creating more appropriate places to store knowledge. E-mail then becomes something more focused - a notification mechanism, or a place for sidebar conversations; e-mail can stop being this firehose of broadcasts and attachments. When used correctly, e-mail can be read and deleted. If a message needs to survive for any longer, it should probably be placed in a repository.

But back to SNARF. It's the first application designed specifically to infer metadata about e-mail with the goal of saving you time, and it works with any MAPI source. In addition to Outlook's built-in filters to color-code and sort messages, this could start saving people serious time. As Marc mentioned in a presentation, the long-term goal is to embed these techniques into Outlook. It's great to finally see a first step.

[SNARF Home Page]

No Comments