Database Reference
In-Depth Information
interesting by adding context to the network. During an event, a user would want
to analyze the event from different perspectives. For example, in a natural disaster
like a thunderstorm, instead of analyzing all the retweets related to the disaster, a
first responder might be more interested in reports of damage or flooding. Thus,
by filtering the information, we can make the visualization manageable while
enhancing the ability to focus on topics of interest.
To achieve this, we define groups of words called topics. For example, we
create topic 1 with “#zuccotti” (Light) and topic 2 with “#nypd” (Dark). We
can now generate a retweet network consisting of people who retweeted text
matching these topics. Before we visualize the network, we must first extract and
format it. This can be done using the method
ConvertTweetsToDiffusionPath
in class
CreateD3Network
, which is summarized in Listing
5.2
.
The extracted network can be visualized using the method
create_network
, which
is summarized in Listing
5.3
. Figure
5.2
shows the visualization of the top five most
frequently retweeted nodes and those who retweeted them on topics 1 and 2. The
size of a node indicates its importance in the network. Larger nodes have been
retweeted more often than smaller nodes. Nodes are colored according to their
topic preference. The links are directed and show the flow of information. Here,
not only can we identify important information producers (large nodes) as well
as information consumers (nodes with a large number of inlinks). Additionally,
the network shows that people retweet across topics, which is evident from the
connections between the users.
Listing 5.2
Extracting the retweet network
public JSONObject ConvertTweetsToDiffusionPath(String inFilename
,int numNodeClasses, JSONObject hashtags, int num_nodes) {
//Step 1: Read through the file and process Tweets
matching the topics
...
//Step 2: Identify the size of the nodes based on the
number of times they are retweeted
ArrayList<NetworkNode> nodes = ComputeGroupsSqrt(
returnnodes, max, min, numNodeClasses);
...
/
**
Step 3
*
Prune the network to keep only the top |
nodes_to_visit| nodes in the network.
*
Recursively visit all top nodes and retain their
connections.
*
/
for(int k=0;k<nodes_to_visit;k++) {
NetworkNode nd = nodes.get(k);
nd.level = 0;
HashMap<String,NetworkNode> rtnodes =
GetNextHopConnections(userconnections,nd,new
HashMap<String,NetworkNode>());
...
/
**
Step 4: Compact the nodes of the network by removing
*
all nodes who have never been retweeted
*
/
