BlogHow to analyse a co-authorship network in Linkage

Let's see how you can explore the papers and their authors in Linkage.

Importing the graph

First, you must have an account, you can signup here.

After selecting "New Job", you should have "co-authorship network" selected. Let's type in a subject: "Bayesian Networks". There, we can choose between 3 sources: arXiv, HAL or PubMed. I'm gonna continue with HAL.

Here, Linkage is gonna fetch papers from the selected source which match you search term. Each node is an author and each link is a co-publication between 2 authors.

Then Linkage execute a clustering on the freshly obtained graph. This can take some time and you can make this faster by restricting the search range in the import page by select Manual in the Clustering section

When the clustering is finished, you will be able to download the results and view them them.

Exploring the graph and the clustering

Linkage job is to find two things: The topics of discussion and the clusters of people that discuss together.

So, when going to the result page, you're gonna be welcomed with a network: The clusters linked between each others by the topics.


You can adjust the number of topics and clusters to simplify or complexify the view. The default one shown is the one with the best score based on a likelyhood score, you can see the score and a graph of this score for all the topics at the bottom of the page.

Thanks for reading and feel free to send us your feedback at

When we had no computers, we had no programming problem either. When we had a few computers, we had a mild programming problem. Confronted with machines a million times as powerful, we are faced with a gigantic programming problem.

Dijkstra (1986) Visuals for BP's Venture Research Conference

Published June 2, 2017


Terms and conditions         Credits