Cocitation Networks and Random Walk

1,307 views
1,128 views

Published on

How to manage a Research assessment exercise (scientific publications) with Random Walk protocols and Cocitation networks

Published in: Education, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,307
On SlideShare
0
From Embeds
0
Number of Embeds
21
Actions
Shares
0
Downloads
2
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Cocitation Networks and Random Walk

  1. 1. Academic online resources: : assessment & usage Researchers, field activists in the context of Bibliometric analyses Univ. LILLE (France) 26/27 NOV 2009 Manuel Durand-Barthez / URFIST Paris
  2. 2. Problematics • Issued from a basic model: Web of Science (ISI) • Dominating classical Model: (Social) Science Citation Index Journal Citation Reports (JCR)
  3. 3. Two types of alternative approaches • Starting from Journals Corpus of the Journal Citation Reports DB (JCR) => Eigenfactor Corpus of the Scopus DB (Elsevier) ⇒SJR (Scimago Journal Rank) • Starting from Articles Links generated by Citebase and CiteseerX Numerous variants of the H-Index suggested by Harzing Publish or Perish
  4. 4. Two link-structured analysis models « Two important types of techniques in link-structure analysis are co-citation based schemes, and random-walk based schemes » (Lempel ; Soffer, 2002) Co-citation Random Walk
  5. 5. Profile of the ISI Model • Scope type: starting from Journals Serials as Document type including Articles Journal Citation Reports and its Impact Factors
  6. 6. Profile of the ISI Model Approximate Definition • Citation count is raw, « first ring » type • Encompasses equally all citing publications issued from the global corpus of the Journal Citation Reports • Whatever their intrinsic « quality » (?) and their disciplinary origin may be
  7. 7. Beyond the first ring • Execution of the Weighted Page Rank, Lawrence Page’s algorithm adapted to Google • Simulating the steps of a researcher browsing the references cited by an article, consulting forwards other articles • Random Walk protocol
  8. 8. Eigenfactor and SJR • Eigenfactor of C.Bergstrom applied to JCR + Scimago Journal Rank (SJR) • Iterative execution of the Page Rank • Cartography of subject areas, traceability of the links • Sequencing and linking between disciplines (STM / Soc Sci)
  9. 9. Eigenfactor and SJR • Extended analysis, from 3 up to 5 years • Automatic adjustment of means of citations inherent to each discipline • Inspired by the distinction Popular vs. Prestigious journals of the Bollen’s « Journal Status »
  10. 10. Preeminence of classical Journals • Eigenfactor and SJR apply to corpuses giving priority to classical journals implying a subscription or a strong embargo • On the contrary, co-citation networks conceptors refer to Open Access corpuses
  11. 11. Impact of OAI type Publications • Citebase (Univ. Southampton, S.Harnad & T.Brody) and CiteseerX (Univ. Pennsylvania) suggest three citations levels: References cited by a « parent » article Shared references References citing simultaneously the « parent » item and other items
  12. 12. Aftermath of co-citations • Cartography of Subject Areas • Amplification and optimal visibility generated by Open Access • Disciplinary interlinking
  13. 13. Aftermath of co-citations • Thin granularity, nominative, at least at the Lab level • Setting aside the popularity of Journals, which are just globally anonymous document types • Focusing on the actors of the research
  14. 14. Aftermath of co-citations • Author, circumscribed in a Lab team, at a period p • Lab team which may be included in one or several communities, appearing through graphs or maps • Analogy with the Research Front Maps of ISI’s ScienceWatch
  15. 15. Data about Research « actors » • Moderate the raw effects of the basic H-Index. See chiefly Harzing’s suggestions: Consider articles positioned under the H level, improve on the H-index by giving more weight to recent articles, thus rewarding academics who maintain a steady level of activity. Adjust the H according to the number of authors contributing to an article
  16. 16. Weighting factors • According to disciplines, variation of The mean citation rate The mean annual number of publications • Unawareness of the effective contribution of each author to a publication
  17. 17. Combination of both models • Random walk + • Co-citation How to adapt an H to a Hub, a cluster of authors and publications ? How to dispatch that new H on a scale adapted to each author who contributes to its calculation ?
  18. 18. By way of conclusion: an hypothesis • To be assessed: Inside a node of international and transdisciplinary co-citations Making reference to the OAI net of articles as such And not to closed corpuses of journals
  19. 19. I.F., H and… Copernic • Instead of beeing assessed In the precinct of a research unit or of an institution In relation to a journal as such • Combine both alternative models • For a copernician ®evolution of evaluation

×