4.3 Interpret Results

Where are the major areas of research based on the input dataset?

This is one of the primary questions that CiteSpace is designed to answer. To answer this question, we will focus on the big picture of the collection of publications represented by your dataset. Let’s make a few adjustments with the sliders in the control panel on the right so that the information of our interest will be shown clearly and information that is less relevant to this question right now will be temporarily hidden from the view.

1. Node Size

At this level, we don’t really need to see the size of a node, although it provides rich information about the history of a node. Use the slider under Article Labeling ►Node Size ►[Slide to 0] (marked by the pointer #1 in the following figure).

2. Cluster Label Size

The font size of the cluster labels are controlled by a slider with two controls: one control the threshold for showing or hiding a label based on the size of the cluster (i.e. to make sure large-enough clusters are always labeled), and the other control the font size of the cluster labels (marked by the pointer #2 in the screenshot).

3. Transparency of Links

Detailed links would be useful later, but we can ignore them for now using the transparency slider to set all the links’ transparency to the lowest level, i.e. invisible. In hindsight, a more accurate term would be completely transparent.

After making these minor adjustments, it will be straightforward to answer the question: Where are the major areas of research? Evidently, the largest area (cluster #0 with the largest number of member references) is biological terrorism. The second largest is posttraumatic stress (cluster #1), i.e. PTSD. The third one is ocular injury (cluster #2). The fourth one is blast (cluster #3). And there are a few smaller clusters. So now we have a general idea what constituted terrorism research during the period of 1996 and 2003. You can repeat the process on a current dataset to get an up-to-date big picture.

How are these major areas connected?

To answer this question, we need to bring back the lines connecting nodes. Adjust the transparency slider to make the lines visible.

A useful indicator of how different clusters are connected is a type of nodes that have high betweenness centrality scores. In CiteSpace, betweenness centrality scores are normalized to the unit interval of [0, 1]. A node of high betweenness centrality is usually one that connects two or more large groups of nodes with the node itself in-between, hence the term betweenness. CiteSpace highlights nodes with high betweenness centrality with purple trims. The thickness of a purple betweenness centrality trim indicates how strong its betweenness centrality is. The thicker the stronger. Occasionally, a node with high betweenness centrality may appear at the center of a network component, but our interest is in the nodes that are truly in between.

To make see the purple rings, switch the node rendering mode to tree rings, which is the first icon shown in the following figure, i.e. concentric citation rings represent how many citations were made to the node in corresponding years. Remember that colors represent when citations were actually made.

Where are the most active areas?

Burst Detection

Citation burst is an indicator of a most active area of research. Citation burst is a detection of a burst event, which can last for multiple years as well as a single year. A citation burst provides evidence that a particular publication is associated with a surge of citations. In other words, the publication evidently has attracted an extraordinary degree of attention from its scientific community. Furthermore, if a cluster contains numerous nodes with strong citation bursts, then the cluster as a whole captures an active area of research, or an emerging trend.

The burst detection in CiteSpace is based on Kleinberg’s algorithm (Kleinberg, 2002).

Using View ►Citation Burst History can generate a summary list of articles that are associated with citation bursts. This visualization shows which references have the strongest citation bursts and which periods of time the strongest bursts took place. For example, from the list, we can tell that Schuster et al. (2001) has the strongest bursts among articles published since terrorist attacks in 2001. It is also interesting to note that North et al. (1999) has the second strongest citation burst in the period of 2002 and 2003.

Burst detection and visualization can be applied to other types of nodes. For the node type of author, it will show you those authors who have rapidly increased the number of publications. Similarly, institutions will identify universities that are particularly active in the relevant research areas. For keywords, it will show you fast growing topics.

The general procedure is the same for different types of nodes. Here we illustrate the procedure with an example of detecting the burstness of keywords in publications of Drexel University between 2000 and 2014.

1. Select the node type: Keyword

2. Generate a network as usual: 2000-2014; Slice length: 1; Top N=100; GO

o (N=392, E=3033)

3. Run the burst detection function: Citation Burst

4. Visualize the entities, i.e. nodes, that have bursts: View > Citation Burst History

What is each major area about? Which/where are the key papers for a given area?

Cluster labels can tell us the context in which they are most cited because the label terms are extracted from citing articles’ titles, keywords, or abstracts.

To explore these clusters in more depth, you should use the Cluster Explorer:

Clusters ►Cluster Explorer

The initial appearance of the Cluster Explorer shows four windows: 1) Clusters, 2) Citing Articles, 3) Cited References, and 4) Representative Sentences. Windows 2-3 are blank until you select a cluster in the Clusters window by checking the checkbox in front of each row of cluster information.

The following figure shows a screenshot after Cluster #0 was selected in the checkbox. As you can see, the Citing Articles window and the Cited References window are both populated accordingly. In the Citing Articles window, each entry is a citing paper, i.e. a paper that cites members of the cluster. The number in front of each entry shows the portion of the references cited by this particular article out of all the references in total. For example, Bak, SJ (2000) has a coverage of 0.28, i.e. 28% of the total 65 references in this cluster (you can find the 65 listed in the Clusters window’s third column – Size).

The phrase biological terrorism was highlighted in yellow in the Citing Articles window. Note that the phrase is also the label of this cluster in the visualization. Furthermore, the phrase also appears in the Clusters windows’ 7^th column – Top Terms (log-likelihood ratio). For technical details, see (C. Chen et al., 2010).

The Cited References window shows the member references of this cluster. Each reference is listed with the number of citations, burstness if any, its centrality score, along with the name of the first author, the year of publication, source (i.e. journal or conference), volume number, and page number.

In the Summary Sentences window, if you click on the Start button, CiteSpace will extract the most representative sentences from the abstracts of the citing articles to each cluster. A sentence is considered representative if it is either a sentence with a high degree centrality or a sentence with a high PageRank score.

Timeline View

You can switch to a timeline view of the network by choosing the Timeline radio button in the Layout panel on the right (as pointed by the red arrow in the following figure). In a timeline view, each cluster is arranged on a horizontal timeline. The direction of time points to the right.

You have seen some of the basic moves. CiteSpace has many other features. We will introduce other features at more advanced levels.