On 18 May 2008, 37 people had tagged this page on boston.com.
This is a graph of all delicious users within 2 degrees of separation of those users who tagged the page. The center (root) of the graph is the URL. The first ring shows the 37 delicious users who tagged the page (the URL's taggers). The second ring shows the URL's taggers' 1st degree friends. The third ring shows 2nd degree friends.
This is the isolated backgraph (people in the main graph who also know each other) with edges at 99% transparency. Nodes are hidden. Where end points overlap the most we see obvious "hot spots." These hot spots show relative amount of inbound connections, potentially suggesting where information flows the most in this network. Revealing the nodes allows the viewer to isolate the people having the most inbound connections.
This is the backgraph at much higher opacity. Other patterns clearly emerge in this view but they are not as clearly understood (by me, anyway).
Networks of people tagging http://www.army.mil/-news/2007/03/22/2359-army-reaches-public-through-youtube-flickr-delicious/ This is a top-down view of a cone tree. Tilting the tree on its side shows connections that cross tiers.
Networks of people tagging the URL: http://pajhome.org.uk/crypt/md5/
Same as the previous but at fewer degrees of separation (1).
Networks of people tagging a page called "CIA Psychology Intel Analysis," which happened to be popular on the day I was testing the URL feature. Not sure what it means, just looks pretty. There is one highly connected backgraph.
Why did this happen?
This is a single level radial tree of my sent mail (Sent Items in Outlook). Recipients are radiated around me in order of the number of mails I sent to them and scaled accordingly. This particular layout doesn't accomodate for node size and so large nodes are crowded into the same space. We're working on it. It's a start...
Cone tree seems to be the best layout for my sent mail. This simple hierarchy breaks people into groups of 50 from 0 to 250. The majority of recipients to whom I send mail has received less than 50 emails from me. All of the remaining groups have less than 10 recipients. The largest two groups have only two recipients each. NOTE: There's a typo in the group labels; two 1-50 labels. The smaller group is actually 101-150.
Close-up of the 1-50 group
Getting a little crazy with a rough cut at a radial tree for my received Bridgeborn mail (Inbox in Outlook). Nodes = 3197, Edges =2071. The most-outlying tiers are not work related. These are friends from a fraternity alumni association of which I was chairman of the board of directors. Also, I appear in here twice. This is because I didn't normalize any instances of email addresses in the first run. That's needed. I'm in the center of the white inner circle and at the root of the two large orange "fans"
The radial tree shows degrees of separation, over which we lay information about sharing behavior and indirect relationships.
This force-directed graph with accented edges clearly shows two separate communities-of-interest held together by a common person (me). The translucent white edges represent the relationships that hold the members of each COI together.
This cone tree is a different way to highlight the "hierarchical" nature of relationships. Most information and relationships in this network are found in the first tier below the root. We are also showing isolation of subtrees by selecting a person in the graph.
Evidence of 2 communities of interest (CoI) in a del.icio.us network shown by the saturated backgraphs.
A very dense del.icio.us network
The same network as previous, but with increased saturation of the backgraph.
Radial Tree visualization of Cisco router NetFlow connections. Shows inbound "hot spots" at 3 nodes (the routers) and the IPs that are connecting to those 3 nodes.
Close-up of Radial Tree for an IP with more connections. Here we see at least 5 hot spots. We can manipulate the attributes of the graph and use filters to isolate the source IPs
Here we re-centered the previous graph on one of the hot spots. This pushed 3 nodes to the 4th degree of separation and you can see a path from the root to these nodes.
What does this mean? The two sub-trees are routers.
Web service endpoints
Backgraph of physics.ao-ph at 3 levels in radial tree layout. From the arxiv API release days ago. This was one hour of thought and that ultimately led to a 2 line quick and dirty mod to Social DNA to point it to a new service and to expect XML instead of JSON. You can see I haven't even updated the form prompts.
Same graph with the foregraph visible. Everything from SocialDNA still works except for the tag cloud. It's a stand-alone page and I just haven't gotten around to it yet. It's actually more work!
all:electron in a cone tree at 2 levels. Cone tree is a great layout for term association.
all:electron at 3 levels
all:electron at 3 levels with backgraph visible. The search term is the node connected by 3 green edges. It is loosely connectected to the rest of the graph compared to other terms. I thought it might be the "all:" prefix, but "electron" returns the same graph.
The search term is not connected to the backgraph. I need a linguist: what does that mean besides the obvious? (That "electron" is directly related to 3 other categories.)
all:electron at a force directed graph. Here the category "gr-qc" node is selected because it seems to have the most categories referring to it.
Terms related only through foregraph are not connected when the foregraph is hidden.
Foregraph only. In the foregraph a different node, astro-ph, has the most references.
This is an DoDAF OV-6c (sequence diagram) in Flare
