Now, I just introduced last week the idea of looking at author relationsships by a graph with a short example. In this post I want to further explain my approach to create such kind of figures. The first requirement is a valid base of data in forms of bibliographic information.
As a bibliography file in form a bibtex database was readily available, the first version of my graph display software is working with bibtex files as source. In a later post, it will be shown how databases can be easily created by automatic parsing of google scholar or other sources.
The created python software reads the bibtexfile, creates a list of authors and a list of relationsships. From these two lists, it creates a graph with nodes and edges and invokes the graphviz software to draw the graph and export it as a png file. The python lib pygraphviz is used as interface for graphviz which includes classes for creation and analysis of graphs, nodes, and edges.
An example of a graph for my bibtex database from my phd thesis is shown below. We can see many different authors and a very complex structure of relationships (along with some bugs as authornames are written differently in publications, e.g. Jager and Jaeger).
We see most authors are in a cloud and only some authors have no connection to others at all. Now, in order to get a better visibility of the graph, the following simple filters are available:
2. Authors with number of relations lower than a threshold are removed
3. Edges with lower weight than a threshold are removed.
These filters are very powerful to reduce complexity of the figure. In order to find main contributors to an area, authors with small amount of publications can be removed, which are mostly PhD students with only a short time in research.
Additionaly, a specific author can be marked as main author and only authors with specific neighbourhood level are printed. In this way, only direct neighbors, or authors over two edges are printed.
Here I wanted to study the network around professor S. Voinigescu from University of Toronto and his direct network and removed all authors with only one publication.
What kind of filters do you want to see for the graph display software? In what kind of research about research are you interested? I’m curious about your feedback.