Research Communities and Networks

To generate a data basis for the models in my simulations, I examined networks of co-authorships from publications on the pre-print repository I analyzed the metadata of all publications in the field of Computer Vision and Pattern Recognition to establish a relationship model of co-authorship. My goal was to reveal the connections between researchers and to check if I could identify communities based on this structure.

These data served as an inspiration for some of the models in my simulation. The fundamental relationships have been published as a dataset (DOI:, names have been anonymized). I used Gephi for the corresponding visualization that reminds me of bacterial cultures, which inspired the ambiguous title 'Science in a Petri Dish'. The data cleaning and interpretation was done with PROLOG and Python Pandas and Numpy.

The data harvesting was done via the OAI REST interface of ArXiv. The network of authorships was then generated by looking at common authorship of single papers of the time periode. I cleand the data (removed minima and maxima) and got some insides of the communities inside of some sections in ArXiv. This section gave me the most interesting result. You can clearly see some of the smaller and larger communities and also stronger connections. Each dot is an author and the common publications are indicated troug green lines.

Project: Research Communities and Networks
Duration 2018
Data (Anonymized Names) Data:
Data (Relations in Gephi) Data:
Data sources Sources:
Publication No paper but mentioned in: Kurzawe, Daniel (2023): Die Dynamik von Forschung und Gesellschaft: Simulationen von Wissenschaftsprozessen. Olms: Hildesheim. ISBN: 978-3-487-16307-9 and open access: DOI: 10.5282/edoc.29687
Code Sources:
Tools SWI-Prolog and Gephi
Poster [DOI: 10.20375/0000-000B-CB4F-9]

Figure: A publication network.