Borner Maru Goldstone The simultaneous evolution of author and paper networks 2004

Citation

Borner, K. Maru J. Goldstone R. The simultaneous evolution of author and paper networks. PNAS. April 6, 2004.

Online version

[1]

Summary

This paper investigates the overtime evolution of citation and author networks and discusses the two processes that mainly govern the growth/evolution of such networks: aging and growth. The authors propose a model that can fit the systematic deviations from power distribution of citation networks well while accounting for the inter-related nature of the paper citation and co-authorship networks and importance of topic distributions. The model aims for integrating the following properties of citation networks:

• Authors have a bias to cite recent papers. Even highly cited papers stop receiving citation after a certain amount of time passed. This feature works against the richer get richer phenomenon enforced by aging, and frequently prevents a scale-free distribution of connectivity.
• Authors have a tendency to cite papers from the reference list of papers they have read, which is a recursive follow up of links in the network.

Brief description of the method

The model proposed in this paper is called TARL which encompasses Topics, Aging, and Recursive Linking. The model attempts to capture different roles of the authors in (i) producing, (ii) storing, and (iii) disseminating information directly by (co)authorship or indirectly by reading/citing others’ papers.

Algorithm

• Initialization

generate #_papers papers and assign a random topic to each paper;

generate #_authors authors and assign a random topic to each author;

randomly assign #_coauthos+1 authors to papers of the same topic;

• Simulation

for each year do {
 add #_new_authors new authors, deactivate authors older than #_author_age;
 for each topic do {
   randomly partition set of authors into author_groups of size #_co-authors+1;
   for each author_group do {
     for each new_paper to be produced, do {
       generate new_paper;
       randomly select #_read_ papers from existing papers;
       get all references of read_ papers up to #_reference_path_length;
       for each new_paper_reference do {
         select a time_slice from (start year to curr_year-1) with probability given in aging_function;
         randomly select a paper published or cited in this time_slice; as a new_paper_reference;
         add the new_paper_reference to new_paper;
       }
     }
   }
 }
 add all new papers to the set of existing papers;
 add new links to author and paper information;
}

Datasets

This paper uses PNAS Dataset for detailed results. However, to examine the general properties of the networks, the authors use datasets from other papers listed in related papers section (primarily Mark Newman's papers):

Co-authorship Networks: LANL, MEDLINE, SPIRES, NCSTRL, Math., Neurosci., PNAS.

Paper-citation Networks: ISI, PhysRev, PNAS, SIM.

Related Papers

1. Newman, M. E. J. (2001) Physical Review E 64, 016131.
2. Newman, M. E. J. (2001) Physical Review E 64, 016132.
3. Newman, M. E. J. (2001) in Proceedings of the National Academy of Sciences, pp. 404 - 409.
4. Redner, S. (1998) European Physical Journal B 4, 131-134.

There are also other papers that we read for our project, and topic-wise they are also related to this paper:

1. Steve Hanneke and Eric Xing, Discrete Temporal Models of Social Networks. Workshop on Statistical Network Analysis, held at the 23 rd International Conference on Machine Learning, 2006.

1. Vladimir Ouzienko, Prediction of Attributes and Links in Temporal Social Networks.