A procedure for field delineation

1
A procedure for field delineation

• with heat maps of bibliographically coupled
  publications using core documents and a cluster
  approach - the case of multiscale simulation and
  modelling

• Edgar Schiebel

Contribution at the ECOST-MEETING-TD1210-290816-07
7233 Identification, location and temporal
evolution of topics - data and algorithm -
comparison of approaches 2016 08 29 2016 08 30
2
Content

• Introduction to the stepwise procedure
• Offline Demonstration of the steps with the
  software BibTechMon

3
Procedure

• Collect a set of publications Download meta data
  from Web of Science
• Not all publications are helpful reduce the
  number of objects
• Calculation of sum similarities (Jaccard index)
  of bibliographically coupled publications
• Selection of a subset of publications with a
  threshold of the sum similarity (beyond the
  expected value)
• Get a heat map of agglomerations of similar
  publications
• Calculation and visualization of a two
  dimensional map of the subset of
  bibliographically coupled publications with a
  spring model (use the second order similarity)
• Filter and visualize the local density of the
  number of publications weighted with the
  similarity to draw agglomerations of publications
  with hot zone.
• Collect the right set of representatives of an
  agglomeration of publications
• Graphically assisted selection of documents in
  the center of a hot red zone
• Selection of additional elements of the community
  by thresholds of similarity (Jaccard index,
  number of common references) (use the first order
  similarity)
• Examine the goodness of the set of
  representatives Visualization of the coupling
  elements (cited references - broad or narrow or
  spread knowledge base)
• Name the subfields Providing lists of TFIDF
  ranked keywords, read highest cited references,
  reviews as well as titles and abstracts of the
  publications with highest sum similarity
• Comparison with cluster analysis
• Cluster analysis (Pearson, Ward or other
  similarities and linkage methods) of the reduced
  set of publications. Visualization with a
  circular dendrogram. Select clusters in different
  hierarchies
• Visualization of selected clusters in the heat
  map
• Comparison of the identified communities of both
  approaches (Concordance matrix)

3
3
4
Data set of publications

• Web of Science (WoS) database. We used a keyword
  based search strategy in the topic search feature
  of WoS. The time span was 1990 to 2014. The
  keyword based search focused on the explicit
  usage of terms formally derived from multiscale
  simulation and modelling (Set 1). In detail we
  got the following number of hits multiscale
  simul (1057 Publ.) multi scale simul (361)
  multiscale model (3664) multi scale model
  (1554). The union of the hits delivered 6326
  records.
• Secondly we collected records on tribology
  research combined with MSSM and enriched the set
  with citing publications. We searched for
  publications in tribology research with the
  following search strings and hits tribo
  (33.755) lubric (40.952) friction (131.023)
  wear (105.615) rheo (81.963) with a union of
  326.066 records (Set 2). The intersection of Set
  1 with Set 2 covered 249 publications (Set 3). To
  have a broader view on the application and
  development of multiscale techniques in tribology
  we enriched Set 3 with citing publications
  without self-citations and got additional 1872
  records (Set 4). The final data set of 8145
  publications was formed by the union of Set 1,
  Set 3 and Set 4. The data was downloaded the 20th
  Oct. 2014.

5
Reduce the number of publicationsStep 1
calculate sum similarities

• Calculation of sum similarities (Jaccard index)
  of bibliographically coupled publications
  (ResearchFronts_aij and ResearchFronts_SumSim)

Mean of Jij 1.9
6
Reduce the number of publicationsStep 2 Select
publications with a threshold for Jij beyond
expectation

• Selection of a subset of publications with a
  threshold for the sum similarity (beyond the
  expected value of 1.9)
• 2325 out of 8145 publications

Jijgt1.9 2325 publications
7
Heat map of agglomerations of bibl. coupled
publications
Noise eliminated
All 7606 publications with cited references
With threshold for sum similarity SJij gt1.9

• Calculation and visualization of a two
  dimensional map of the subset of
  bibliographically coupled publications with a
  spring model (use the second order similarity),
  Filter and visualize the local density of the
  number of publications weighted with the
  similarity to draw agglomerations of publications
  with hot zone.

8
New calculation of the map with the reduced
number of publications
sharply delineated agglomerations
9
Collect the right set of representatives of an
agglomeration of publications (use first order
similarity)
First order similarity to publications nearby
First order similarity to publications nearby
Completed set for subfield 1
Selection of core publicattions in the hot zone
10
Examine the goodness of the set of
representatives Visualization of the coupling
elements
Nicot F, 2006, INT J SOLIDS STRUCT, V43, P3569
Darve F, 2004, COMPUT METHOD APPL M, V193, P3057
CUNDALL PA, 1979, GEOTECHNIQUE, V29, P47
11
Name the subfield granular materials
List of keywords
Cited references, use www.doi.org
List of recent publications
12
Cluster analysis, Pearson colleration, Ward
linkage
Circular tree
Selected levels
13
Cluster structure in the heat map of bibl.
coupled publications
Granular materials
black lines are edges of the similarities and
green lines indicate the cluster category
14
Concordance Example Granular materials 2
clusters
15
List of subfields for multiscale simulation and
modelling
16
Discussion

• Useful for some 10k publications
• It is a procedure to identify research issues not
  large fields
• Semiautomatic and uses bibliometrics as a toolbox
• Different use of second order and first order
  similarities bt documents
• Allows non objective decisions by the person or
  the team who works on the issues for the
  selection of publications
• The purpose is to identify and monitor research
  issues in the scientific community for example
  for expert organizations

17
Thank you for your attention