collect_bibliography() creates a network from bibliographic records, such as an export from the Web of Science, Scopus, or OpenAlex. The result is one multimodal network of the works, their authors, their keywords, and their sources, in which the works also cite one another. The classic bibliometric networks, such as co-citation, bibliographic coupling, co-authorship, and co-word networks, are then projections or proximities of this network (see Recipes).

collect_bibliography(records, keywords = c("author", "indexed", "both"))

Arguments

records

The bibliographic records. A data frame from {bibliometrix} or {openalexR}, or the path to a BibTeX or BibLaTeX file (see Details).

keywords

Which keywords to collect. One of "author" (the default), the keywords that the authors give ('DE', or 'keywords' in BibTeX and OpenAlex), "indexed", the keywords that the database adds ('ID', 'keywords-plus' in BibTeX, or the concepts or topics of OpenAlex), or "both". Keywords are lowercased.

Value

A stocnet object. A multimodal, multiplex network of works, authors, keywords, and sources.

Details

Only base R is used, so no bibliometric package is needed. The records can come from:

  • bibliometrix::convert2df(), a data frame with one record in each row and the fields of the Web of Science as its columns: 'AU' (authors), 'CR' (cited references), 'DE' (author keywords), 'ID' (indexed keywords), 'SO' (source), 'PY' (year), 'DI' (DOI), 'TI' (title), and 'SR' (a short label);

  • openalexR::oa_fetch(), a data frame of works with the columns 'id', 'display_name', 'authorships' (or 'author'), 'publication_year', 'doi', 'source_display_name' (or 'so'), 'referenced_works', 'keywords', and 'concepts' or 'topics';

  • a BibTeX or BibLaTeX file, such as the BibTeX export of the Web of Science, which holds the cited references of each record in its 'cited-references' field. A file that Zotero exports holds no cited references, so it gives works, authors, keywords, and sources, but no citations. The editors of a work are its authors where it has no authors. Records in other forms can be given as a data frame with the columns of {bibliometrix}.

The network has four modes and four layers:

LayerFromToDirected
"cites"worksworksyes
"authored"authorsworksno
"tagged"workskeywordsno
"published"workssourcesno

A mode or layer with no nodes or ties is left out. The works have a 'citations' column with the number of times the database counts them as cited, in all the works it holds ('TC' in {bibliometrix}, 'cited_by_count' in OpenAlex, or 'times-cited' in BibTeX), where it gives one. This is the global citation count. The local citation count, how often the records cite a work, is the indegree of the work in the "cites" layer. Each "authored" tie has the position of its author among the authors of the work in a 'position' column, 1 for the first author, since some counts weight the first or last authors more. Each tie has the publication year of its work as its 'time', and for "cites" this is the year of the citing work. Since a work that is published stays published, and a citation stays made, the ties are events that increment the network, so to_time(.data, 2010) gives the literature up to 2010.

Cited works

A cited reference becomes a tie to a work. The reference is matched to a work in the records where possible:

  1. by a normalised string of its first author, year, and source, as the Web of Science writes references (e.g. "GLANZEL W, 1996, SCIENTOMETRICS"), where this string identifies one record (the 'SR' column of {bibliometrix}, or else the same string made from the first author, year, and source of the record);

  2. else by its DOI, where both the reference and the record have one;

  3. else by its OpenAlex identifier, for OpenAlex records.

A reference that matches no record is a work too, since co-citation is about the works cited, most of which are not in the records. These works have FALSE in the 'recorded' column of the nodes, so filter_nodes(.data, recorded | is.na(recorded)) keeps only the works in the records (and the other modes). They are not inactive, since they hold ties. References that are not matched to a record are one work where they have the same normalised string of first author, year, and source, as in {bibliometrix}, whether or not they give a DOI. The 'doi' of such a work is the first DOI that its references give. Two works of an author in the same year and source are therefore one work, even where their DOIs differ, and a work that is written differently in two references, such as with another spelling of its author, is two works. A reference that matches the record that cites it is another work, since a work does not cite itself. Scopus writes references in another order, so its references are matched to records only by DOI.

Recipes

The classic networks come from the existing functions:

NetworkRecipe
citationto_layer(.data, "cites")
co-citationto_proximity(to_layer(.data, "cites"), across = "columns")
bibliographic couplingto_proximity(to_layer(.data, "cites"), across = "rows")
co-authorshipto_mode(.data, "authors", through = "works", similarity = "fractional")
co-wordto_mode(.data, "keywords", through = "works")
authors and keywordsto_twomode(.data, through = c("works", "sources"))

Co-citation ties two works that are cited by the same work, and coupling ties two works that cite the same work, in both cases weighted by the number of such works. "fractional" counting shares each work among its authors, so that a work of many authors does not dominate (Perianes-Rodriguez, Waltman, and van Eck 2016); use "count" for full counting. See to_mode() for other similarities, such as "association" strength.

References

Perianes-Rodriguez, Antonio, Ludo Waltman, and Nees Jan van Eck. 2016. "Constructing bibliometric networks: A comparison between full and fractional counting". Journal of Informetrics, 10(4): 1178-1195. doi:10.1016/j.joi.2016.10.006

Examples

records <- data.frame(
  SR = c("SMITH J, 2001, SCIENTOMETRICS", "JONES K, 2003, SCIENTOMETRICS",
         "LI X, 2005, J INFORMETR"),
  AU = c("SMITH J;JONES K", "JONES K", "LI X;SMITH J"),
  TI = c("Citations", "Coupling", "Co-words"),
  PY = c(2001, 2003, 2005),
  SO = c("SCIENTOMETRICS", "SCIENTOMETRICS", "J INFORMETR"),
  DE = c("citation; network", "coupling; network", "network"),
  CR = c("PRICE DJD, 1965, SCIENCE, V149, P510",
         "SMITH J, 2001, SCIENTOMETRICS, V50, P1; PRICE DJD, 1965, SCIENCE",
         "SMITH J, 2001, SCIENTOMETRICS; JONES K, 2003, SCIENTOMETRICS"))
(bib <- collect_bibliography(records))
#> 
#> ── # Bibliographic network ─────────────────────────────────────────────────────
#> # A dynamic, labelled, multiplex, multilevel network of 4 works, 3 authors, 3
#> keywords, and 2 sources and 5 cite arcs, 5 authored ties, 5 tagged ties, and 3
#> published ties from 2001 to 2005
#> 
#> ── Nodes 
#> # A tibble: 12 × 7
#>   label                         mode    title      year doi   reference recorded
#>   <chr>                         <chr>   <chr>     <int> <chr> <chr>     <lgl>   
#> 1 SMITH J, 2001, SCIENTOMETRICS works   Citations  2001 NA    NA        TRUE    
#> 2 JONES K, 2003, SCIENTOMETRICS works   Coupling   2003 NA    NA        TRUE    
#> 3 LI X, 2005, J INFORMETR       works   Co-words   2005 NA    NA        TRUE    
#> 4 PRICE DJD, 1965, SCIENCE      works   NA         1965 NA    PRICE DJ… FALSE   
#> 5 SMITH J                       authors NA           NA NA    NA        NA      
#> 6 JONES K                       authors NA           NA NA    NA        NA      
#> # ℹ 6 more rows
#> 
#> ── Ties 
#> # A tibble: 18 × 5
#>    from    to layer     time position
#>   <int> <int> <chr>    <int>    <int>
#> 1     1     4 cites     2001       NA
#> 2     2     1 cites     2003       NA
#> 3     2     4 cites     2003       NA
#> 4     3     1 cites     2005       NA
#> 5     3     2 cites     2005       NA
#> 6     5     1 authored  2001        1
#> # ℹ 12 more rows
#> 
to_layer(bib, "cites") |> to_proximity(across = "columns")
#> ── # Proximities of Bibliographic network ──────────────────────────────────────
#> # A labelled, undirected network of 4 works, 3 authors, 3 keywords, and 2
#> sources and 2 ties
#> # Transformed by projection: proximity (count)
#> 
#> ── Nodes 
#> # A tibble: 12 × 7
#>   label                         mode    title      year doi   reference recorded
#>   <chr>                         <chr>   <chr>     <int> <chr> <chr>     <lgl>   
#> 1 SMITH J, 2001, SCIENTOMETRICS works   Citations  2001 NA    NA        TRUE    
#> 2 JONES K, 2003, SCIENTOMETRICS works   Coupling   2003 NA    NA        TRUE    
#> 3 LI X, 2005, J INFORMETR       works   Co-words   2005 NA    NA        TRUE    
#> 4 PRICE DJD, 1965, SCIENCE      works   NA         1965 NA    PRICE DJ… FALSE   
#> 5 SMITH J                       authors NA           NA NA    NA        NA      
#> 6 JONES K                       authors NA           NA NA    NA        NA      
#> # ℹ 6 more rows
#> 
#> ── Ties 
#> # A tibble: 2 × 2
#>    from    to
#>   <int> <int>
#> 1     1     2
#> 2     1     4
#> 
to_mode(bib, "authors", through = "works", similarity = "fractional")
#> ── # Projection of Bibliographic network ───────────────────────────────────────
#> # A labelled, undirected network of 3 nodes and 2 ties
#> # Transformed by projection: mode 2 (fractional)
#> 
#> ── Nodes 
#> # A tibble: 3 × 1
#>   label  
#>   <chr>  
#> 1 SMITH J
#> 2 JONES K
#> 3 LI X   
#> 
#> ── Ties 
#> # A tibble: 2 × 2
#>    from    to
#>   <int> <int>
#> 1     1     2
#> 2     1     3
#>