Visualizing the context of citations referencing papers published by Eugene Garfield: a new type of keyword co-occurrence analysis
Visualizing the context of citations referencing papers published by Eugene Garfield: a new type of keyword co- occurrence analysis
Lutz Bornmann 0 1 2
Robin Haunschild 0 1 2
Sven E. Hug 0 1 2
Eugene Garfield 0 1 2
Robin Haunschild 0 1 2
Sven E. Hug 0 1 2
0 Evaluation Office, University of Zurich , Mu ̈hlegasse 21, 8001 Zurich , Switzerland
1 Max Planck Institute for Solid State Research , Heisenbergstr. 1, 70569 Stuttgart , Germany
2 Division for Science and Innovation Studies, Administrative Headquarters of the Max Planck Society , Hofgartenstr. 8, 80539 Munich , Germany
During Eugene Garfield's (EG's) lengthy career as information scientist, he published about 1500 papers. In this study, we use the impressive oeuvre of EG to introduce a new type of bibliometric networks: keyword co-occurrences networks based on the context of citations, which are referenced in a certain paper set (here: the papers published by EG). The citation context is defined by the words which are located around a specific citation. We retrieved the citation context from Microsoft Academic. To interpret and compare the results of the new network type, we generated two further networks: cooccurrence networks which are based on title and abstract keywords from (1) EG's papers and (2) the papers citing EG's publications. The comparison of the three networks suggests that papers of EG and citation contexts of papers citing EG are semantically more closely related to each other than to titles and abstracts of papers citing EG. This result accords with the use of citations in research evaluation that is based on the premise that citations reflect the cognitive influence of the cited on the citing publication.
Assessments based on bibliometric data is the cornerstone of modern research evaluation
. Research evaluation without bibliometrics seems no longer
imaginable today. The foundation for this importance of bibliometrics was laid by Eugene
Garfield (EG). According to
, EG ‘‘enabled an entire field: scientometrics,
the quantitative study of science and technology’’. EG conceptualized a scientific citation
index similarly designed as Shepard’s Citations—a system which tracked how US court
cases cited former ones—and published this concept in Science
invention of the index ‘‘revolutionized the world of scientific information and made him
one of the most visionary figures in information science and scientometrics’’
(van Raan and
. He founded the Institute for Scientific Information (ISI) in the 1960s (now
Clarivate Analytics, see clarivate.com) which provided later the Science Citation Index,
the Social Sciences Citation Index, and the Arts and Humanities Citation Index—the
predecessors of the Web of Science (WoS). With algorithmic historiography,
Garfield et al.
developed—based on the concept of citations as recursive operations—one of the
first bibliometric network types which finally led to the network program HistCiteTM
During EG’s lengthy career as information scientists, he published more than 1500
publications (see www.researcherid.com/rid/A-1009-2008). In this study, we use the
impressive oeuvre of EG to introduce a new type of bibliometric networks: keyword
cooccurrence networks based on the context of citations, which are referenced in a certain
paper set (here: the papers published by EG). The citation context is defined by the words
which are located around a specific citation. Most of the networks published hitherto are
citation, co-citation, bibliographic coupling, co-authorship networks, or keyword
(Guevara et al. 2016)
. As a rule, previous keyword co-occurrence networks are
based on extracted words from the title and abstract of a publication or from
authorgenerated keyword lists
(van Eck and Waltman 2014)
. To the best of our knowledge, the
citation context was never used in earlier studies as data source for generating keyword
Literature overview and research questions
Bornmann and Daniel (2008)
published an extensive review of studies investigating the
context of citations
(see also Zhang et al. 2013)
. Most of these studies have been published
more than 20 years ago. The studies used the citation context to categorize citations (e.g.
the citer suggests that the cited paper is erroneous in part and offers a correction).
Bertin et al. (2016)
note that ‘‘over the last few year[s], we are witnessing an
increasing body of literature on the citation context of scientific papers’’ (p. 1418). For
Small et al. (2017)
analyzed the context of citations by searching for words that
labeled referenced items as discoveries. The citing sentences which contained ‘‘discovery
words’’ have been called ‘‘discovery citance’’. Hu et al. (2015) investigated the context of
recurring citations in papers published by the Journal of Informetrics. They found that ‘‘for
a specific reference, its first-time citation is usually not as intentional as the succeeding
citations’’ (p. 221).
Jha et al. (2016)
used comprehensive datasets for their citation context
analyses. Their results show—among other things—that the citation context can contain
both evaluative and informative signals.
One of the main challenges of citation context studies is the great effort of producing the
data: The text around the citations must be delimited, extracted, and categorized. In the
past, this had to be done manually or semi-manually and required a lot of effort. As a
consequence, not many studies have been published and most studies were small-scale. In
the age of big scholarly data, producing suitable data is becoming easier. For example,
there exist large text corpora with extensive XML mark-up that facilitate the analysis of
(see Bertin et al. 2016)
. According to
Halevi and Moed (2013)
electronic publishing enables the creation of very large text files that can be analyzed in an
automatic way. Computer scientists, linguistic experts, and information scientists have
begun processing the full texts of research articles in large numbers of journals’’ (p. 1904).
Since a few months, citation contexts can be directly downloaded from a new citation
database, Microsoft Academic
(MA, see Hug and Bra¨ndle 2017; Hug et al. 2017)
. To the
best of our knowledge, MA is the first large-scale scholarly database that allows
downloading citation contexts which are already segmented. In this study, we use citation
context data from MA to produce keyword co-occurrence networks which are based on
extracted keywords from the citation context. To interpret the results of the new network
type, we generated two further networks for comparison: co-occurrence networks which
are based on title and abstract keywords from the (1) cited and (2) citing publications. We
are interested in similarities and differences between the networks. We assume that the
citation context network is similar to the network which is based on title and abstract
keywords from the cited (but not citing) publications. The use of citations in research
evaluation is based on the premise that citations reflect the cognitive influence from the
cited to the citing publication. Thus, the citations in the citing documents should be
similarly contextualized as the cited papers are described in title and abstract.
Microsoft Academic (MA)
MA was publicly released in February 2016
and models ‘‘the real-life
academic communication activities as a heterogeneous graph’’ (Sinha et al. 2015, p. 244).
It gets the majority of its data from web pages indexed by Bing
(Sinha et al. 2015)
updated on a weekly basis
. The database is independent of its
predecessor, Microsoft Academic Search, which has been completely decommissioned towards
the end of 2016. MA data can either be accessed by using the Microsoft Academic search
engine1 or by employing the Academic Knowledge API (AK API).2 In contrast to Google
Scholar, data suitable for bibliometric analyses can be retrieved with relative ease from
(see Hug et al. 2017)
. The database is still under development and evolving quickly
(see Hug and Bra¨ndle 2017)
. For example, a social network for academics has been added
recently. Furthermore, the database has expanded rapidly from 83 million records in 2015,
to 140 million in 2016, and to 168 million in early 2017
(Hug and Bra¨ndle 2017)
1 https://academic.microsoft.com. 2 https://www.aka.ms/AcademicAPI.
All MA data were collected via the AK API using the Evaluate REST endpoint (evaluates
a query expression and returns AK entity results). First, we retrieved all publications by EG
and their MA ID. Only 327 of EG’s 1558 publications were found in MA. Second, we
retrieved all publications citing EG’s papers. Citations contexts were available for 59
papers. Third, for the 59 papers and their citing papers (n = 343 papers), we retrieved the
title and abstract information from the WoS by searching for the Digital Object Identifier
(DOI). These two sets of publications show an overlap of two papers. Finally, we extracted
the citation context (428 citation contexts due to multiple citations in some citing papers)
of the citing papers.
Keyword co-occurrence analysis
We produced three co-occurrences networks using MA data and the VOSviewer software
(www.vosviewer.com; van Eck and Waltman 2010)
: (1) title and abstract network based on
EG’s papers, (2) title and abstract network based on papers citing EG’s papers, and (3)
network based on citation contexts around citations of EG’s papers.
To extract keywords from the titles, abstracts and citation contexts, the text mining
function of the VOSviewer
(van Eck and Waltman 2011)
was used. This function creates a
co-occurrence network of keywords (adjectives and nouns) and displays it on a
twodimensional map. Two keywords are said to co-occur if they both occur in the same title/
abstract or citation context. The distance between two keywords (two nodes) is
approximately inversely proportional to the similarity (relatedness in terms co-occurrence) of the
keywords. Hence, keywords with a higher rate of co-occurrence tend to be found closer to
each other. The VOSviewer provides a clustering function, which assigns keywords to
clusters based on their co-occurrence
(see van Eck and Waltman 2017; Waltman and van
Eck 2013; Waltman et al. 2010)
To generate each of the three maps in this study, identical settings in the VOSviewer
were applied: we used binary counting, a keyword had to occur at least four times, and we
included the 60% most relevant keywords in the network. For each map, the number of
clusters was determined based on interpretability reasons. Keywords not relevant to our
analysis were excluded manually. Words that structure abstracts (e.g. ‘practical
implications’, ‘originality value’) and names of cited authors in citation contexts (e.g. ‘Moed
et al.’) were removed.
In the first step of the analysis, we generated a network based on EG’s papers for which a
citation context was available (n = 59) to get an impression of his papers (included in this
study). From the titles and abstracts of these 59 papers published by EG, 607 keywords
were extracted of which 27 occurred four or more times. Based on the criteria mentioned in
‘‘Keyword co-occurrence analysis’’ section, 15 keywords were included in the map. The
two clusters ‘Journal Impact Factor’ (JIF) and ‘historiographs’ (comprising the HistCite
software package as well as the Science Citation Index) emerged (see Fig. 1). These
entities form an important part of EG’s legacy
(see e.g. Small 2017)
However, comparing Fig. 1 with the 23 broad topics extracted by
publications reveals that the topics in Fig. 1 are not representative for EG’s extensive
oeuvre. Likely, this is due to the very low number of EG’s publications covered by MA,
which form the basis of our analysis. Although 327 of EG’s 1558 publications were found
in MA, citation contexts were available for only 59 publications. However, as our aim is
not to give a complete description of EG’s work, but to explore a new approach to citation
contexts, the scarcity of available data is not a problem for our analysis. All three networks
which we generated in this study are only related to the restricted publication set. Thus, we
expect that also the citing papers of the 59 publications and the corresponding citation
contexts reflect only a part of EG’s influence or impact.
In the second step of the analysis, we produced the co-occurrence network which refers
to the citing papers. From the titles and abstracts of papers citing EG’s 59 papers, 6878
keywords were extracted of which 512 occurred four or more times. After applying the
inclusion criteria (see ‘‘Keyword co-occurrence analysis’’ section), 297 keywords were
used to build the map. These are significantly more keywords than from titles and abstracts
of cited papers. Correspondingly, 18 clusters were identified in contrast to only two clusters
in the cited paper map (see Fig. 1).
The 18 clusters identified by VOSviewer can be characterized with the following
Journal Impact Factor (improvement, citation distribution, percentiles,
disadvantages of JIF);
SNIP (Scopus, new indicator, reference list, research field);
h-index (career, age, validity);
PageRank (information retrieval, machine learning, World Wide Web);
Journal Citation Report (JIF score, prestige, Eigenfactor);
Science Citation Index (research article, patent, text mining);
Publication output of regions or nations (Asia, USA, Western Europe);
Rise of output of nations (last decade, China);
Self-citation (individuals, journals, colleagues);
Coverage of databases (books, humanities, citation practices);
Mapping and visualization (technique, clustering, interdisciplinarity, innovation);
Citation networks and prediction or recommendation (biology, chemistry,
Diffusion of ideas (intellectual structure, knowledge domain, scientific
collaboration, social network analysis);
Scholarly communication (knowledge management, librarian, practitioner, reader);
Publication and citation analysis in medicine (Lancet, New England Journal of
Web links (research productivity and quality, links between universities, search
Digital libraries (future, manuscript, repository);
Peer review and research evaluation (rating, citation frequency, excellence).
The 18 clusters cover a broad range of bibliometric topics, which refer—among other
things—to products of providers of bibliometric data (e.g. JCR), indicators (e.g. h-index),
units of bibliometric analysis (e.g. nations), types of citations (e.g. self-citations),
bibliometric methods (e.g. citation networks), and the use of citations in peer review and
research evaluation. The broad range of topics show that EG’s papers are cited by
publications with a very heterogeneous thematic spectrum.
In the third step of the analysis, we contrast the results on the cited and citing papers
with those of the citation context in the citing papers. From the 428 citation contexts that
enclose references to EG’s publications, a total of 2347 keywords were extracted of which
184 occurred four or more times. Based on the inclusion criteria, 97 keywords were
retained in the map. The following three clusters were identified: ‘Journal Impact Factor’,
‘historiographs’, and ‘Eugene Garfield as founder of the ISI’ (see Fig. 3).
Hence, in Fig. 3, the same two topics as in Fig. 1 can be found (i.e. JIF and
historiographs). However, a closer look at these two clusters reveals that more and different
keywords are associated with these clusters in Fig. 3. For example, the cluster ‘JIF’
additionally comprises keywords such as assessment, research performance, tenure,
normalization, Journal Citation Report (JCR), subject category, and medicine. The differences
between the keywords of the JIF cluster in all three networks will be explored in more
detail at the end of this section. The cluster ‘historiographs’ in Fig. 3 not only comprises
the HistCite software package but also keywords such as idea, structure, citation network,
scientific development, research front
(a term coined by EG according to Sen 2017)
(a term reminiscent of EG’s work on the history of the DNA, see Small et al. 2017)
A third cluster emerges in Fig. 3 that is not present in Figs. 1 and 2: EG as the founder
of the ISI. This cluster comprises keywords such as citation indexing, information retrieval,
the Social Sciences Citation Index (SSCI), Eugene Garfield, founder, database, and ISI.
Hence, this cluster refers to the person EG, whereby similar accomplishments of EG are
mentioned as in his obituaries. For example,
Braun et al. (2017)
praise that EG ‘‘founded
Journal impact factor (17)
Scientific journal (9)
Total number (7)
the Institute for Scientific Information (ISI)’’ (p. 1) and ‘‘created several bibliographic
indexes and databases’’ (p. 1).
The JIF is a recurring topic and ranks among the most frequent keywords in all three
networks. EG ‘‘launched the idea of a journal impact factor, and created the Journal
Citation Reports in which citation analysis is applied as a tool for journal evaluation. Again
in Science he published his most highly cited paper on citation analysis as a tool in journal
(van Raan and Wouters 2017)
Thus, we take a closer look at the keywords associated with this cluster and compare the
results of the three networks. Table 1 lists the most frequent keywords of the JIF cluster in
cited papers, citing papers, and citation contexts. As expected, the lowest number of
keywords can be found for the cited papers. With ‘journal’, ‘impact’, ‘impact factor’, and
‘journal impact factor’, these keywords focus on the indicator itself and do not refer to
related topics. However, the keywords from the citing papers and citation contexts contain
several links to related topics. For example, the JIF is associated in titles and abstracts of
citing papers with keywords such as ‘improvement’, ‘property’, ‘citation distribution’,
‘alternative’, and ‘percentile rank’, indicating a relationship between the JIF and the
development of alternative indicators.
The relationship between the JIF and alternative indicators cannot exactly be found in
the keywords from citation contexts. However, in both networks (citing papers and citation
contexts) the ‘normalization’ is among the most frequent keywords. It refers to the problem
of the JIF that it cannot be used for cross-field comparisons. Citation data have
fieldspecific patterns which can be normalized by different methods
(Bornmann and Marx
Our empirical example has demonstrated that this network type leads to meaningful
results which characterise how cited studies are perceived. Networks based on titles and
abstracts of the citing papers cover a broad range of topics but the overlap with the
networks based on the cited papers and based on citation contexts is marginal. First, no
cluster and no keyword referring to historiography could be identified. Second, the three
clusters representing EG’s achievements (i.e. JIF, JCR, SCI) are associated with different
keywords in the citation context network than in the other two networks. In particular, the
JIF is discussed in relation to other indicators such as SNIP, percentiles, and the h-index,
JCR in relation to JIF scores and the Eigenfactor, and the SCI in relation to patents.
Taken as a whole, the comparison of the three networks suggests that papers of EG and
citation contexts of paper citing EG are semantically more closely related to each other
than to titles and abstracts of papers citing EG. This result accords to our expectation that
the citation context network is related to title and abstract keywords from the cited but not
citing publications. This is in line with the use of citation data in research evaluation
purposes (see ‘‘Introduction’’ section). However, we included only a restricted set of EG’s
papers in this study (see ‘‘Dataset used’’ section). Many papers could not be found in MA.
Furthermore, the citation context was not available for many citing papers. Future studies
investigating the relationship between cited papers, citing papers, and citation contexts
(with co-occurrence networks) should use significantly larger datasets covering broad
ranges of subject areas.
Acknowledgements Open access funding provided by Max Planck Society. We thank Kuansan Wang,
Darrin Eide, and Alvin Chen from the development team of Microsoft Academic for their support and
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution,
and reproduction in any medium, provided you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if changes were made.
Bertin , M. , Atanassova , I. , Sugimoto , C. R. , & Lariviere , V. ( 2016 ). The linguistic patterns and rhetorical structure of citation context: An approach using n-grams . Scientometrics , 109 ( 3 ), 1417 - 1434 . https:// doi.org/10.1007/s11192-016-2134-8.
Best , J. ( 2012 ). Damned lies and statistics: Untangling numbers from the media, politicians, and activists . Berkeley, CA: University of California Press.
Bornmann , L. ( 2017 ). Measuring impact in research evaluations: A thorough discussion of methods for, effects of, and problems with impact measurements . Higher Education , 73 ( 5 ), 775 - 787 .
Bornmann , L. , & Daniel , H. -D. ( 2008 ). What do citation counts measure? A review of studies on citing behavior . Journal of Documentation , 64 ( 1 ), 45 - 80 . https://doi.org/10.1108/00220410810844150.
Bornmann , L. , & Marx , W. ( 2015 ). Methods for the generation of normalized citation impact scores in bibliometrics: Which method best reflects the judgements of experts? Journal of Informetrics , 9 ( 2 ), 408 - 418 . https://doi.org/10.1016/j.joi. 2015 . 01 .006.
Braun , T. , Gla¨nzel, W. , & Schubert , A. ( 2017 ). Obituary. Scientometrics , 111 ( 1 ), 1. https://doi.org/10.1007/ s11192-017-2352-8.
Cobo , M. J. , Lo´pez- Herrera , A. G. , Herrera-Viedma , E. , & Herrera , F. ( 2011 ). Science mapping software tools: Review, analysis, and cooperative study among tools . Journal of the American Society for Information Science and Technology . https://doi.org/10.1002/asi.21525.
Garfield , E. ( 1955 ). Citation indexes for science-New dimension in documentation through association of ideas . Science , 122 ( 3159 ), 108 - 111 .
Garfield , E. , Sher , I. H. , & Torpie , R. J. ( 1964 ). The use of citation data in writing the history of science . Philadelphia: Institute for Scientific Information Inc.
Guevara , M. R. , Hartmann , D. , Aristara´n, M. , Mendoza , M. , & Hidalgo , C. A. ( 2016 ). The Research Space: Using the career paths of scholars to predict the evolution of the research output of individuals, institutions, and nations . Retrieved April 6 , 2016 , from http://arxiv.org/abs/1602.08409.
Halevi , G. , & Moed , H. F. ( 2013 ). The thematic and conceptual flow of disciplinary research: A citation context analysis of the journal of informetrics, 2007 . Journal of the American Society for Information Science and Technology , 64 ( 9 ), 1903 - 1913 . https://doi.org/10.1002/asi.22897.
Hu , Z. , Chen , C. , & Liu , Z. ( 2015 ). The recurrence of citations within a scientific article . In A. A. Salah , Y. Tonta , A. A. A. Salah , C. Sugimoto & U. Al (Eds.), The 15th conference of the international society for scientometrics and informetrics (pp. 221 - 229 ). Istanbul, Turkey: ISSI, Boazic¸i University Printhouse.
Hug , S. E. , & Bra¨ndle, M. P. ( 2017 ). The coverage of Microsoft Academic: Analyzing the publication output of a university . Scientometrics, 113 ( 3 ), 1551 - 1571 . https://doi.org/10.1007/s11192-017-2535-3.
Hug , S. E. , Ochsner , M. , & Bra¨ndle, M. P. ( 2017 ). Citation analysis with microsoft academic . Scientometrics , 111 ( 1 ), 371 - 378 . https://doi.org/10.1007/s11192-017-2247-8.
Jha , R. , Jbara , A.-A. , Qazvinian , V. , & Radev , D. R. ( 2016 ). NLP-driven citation analysis for scientometrics. Natural Language Engineering . https://doi.org/10.1017/S1351324915000443.
Lamont , M. ( 2012 ). Toward a comparative sociology of valuation and evaluation . Annual Review of Sociology , 38 ( 1 ), 201 - 221 . https://doi.org/10.1146/annurev-soc- 070308 -120022.
Leydesdorff , L. ( 2010 ). Eugene Garfield and algorithmic historiography: Co-words, co-authors, and journal names . Annals of Library and Information Studies , 57 ( 3 ), 248 - 260 .
Microsoft. ( 2017a ). Microsoft Academic. Retrieved June 8 , 2017 , from https://www.microsoft.com/en-us/ research/project/academic/.
Microsoft. ( 2017b ). Microsoft Academic Graph. Retrieved June 8 , 2017 , from https://www.microsoft.com/ en-us/research/project/microsoft-academic-graph/.
Mingers , J. , & Leydesdorff , L. ( 2015 ). A review of theory and practice in scientometrics . Retrieved February 2 , 2015 , from http://arxiv.org/abs/1501.05462.
Sen , B. K. ( 2014 ). Eugene Garfield-Glimpses of his writings . Journal of Scientometric Research , 3 ( 1 ), 57 - 60 .
Sen , B. K. ( 2017 ). Eugene Garfield-An obituary . Annals of Library and Information Studies: ALIS , 64 ( 1 ), 7 - 8 .
Sinha , A. , Shen , Z. , Song , Y. , Ma , H., Eide , D. , Hsu , B. P. , & Wang , K. ( 2015 ). An overview of Microsoft Academic Service (MAS) and applications . Paper presented at the proceedings of the 24th international conference on World Wide Web (WWW'15 Companion) , Florence, Italy.
Small , H. ( 2017 ). A tribute to Eugene Garfield: Information innovator and idealist . Journal of Informetrics , 11 ( 3 ), 599 - 612 . https://doi.org/10.1016/j.joi. 2017 . 04 .006.
Small , H. , Tseng , H. , & Patek , M. ( 2017 ). Discovering discoveries: Identifying biomedical discoveries using citation contexts . Journal of Informetrics , 11 ( 1 ), 46 - 62 . https://doi.org/10.1016/j.joi. 2016 . 11 .001.
van Eck , N. J. , & Waltman , L. ( 2010 ). Software survey: VOSviewer, a computer program for bibliometric mapping . Scientometrics , 84 ( 2 ), 523 - 538 . https://doi.org/10.1007/s11192-009-0146-3.
van Eck , N. J. , & Waltman , L. ( 2011 ). Text mining and visualization using VOSviewer . ISSI Newsletter , 7 ( 3 ), 50 - 54 .
van Eck , N. , & Waltman , L. ( 2014 ). Visualizing bibliometric networks . In Y. Ding, R. Rousseau , & D. Wolfram (Eds.), Measuring scholarly impact (pp. 285 - 320 ). Berlin: Springer International Publishing.
van Eck , N. J. , & Waltman , L. ( 2017 ). Citation-based clustering of publications using CitNetExplorer and VOSviewer . Scientometrics. https://doi.org/10.1007/s11192-017-2300-7.
van Raan , A. F. J. , & Wouters , P. ( 2017 ). Eugene Garfield 1925 -2017: Visionary information scientist . Retrieved April 26 , 2017 , from https://www.cwts.nl/news?article= n -q2y2c4& title=eugene-garfield1925-2017-visionary-information-scientist.
Waltman , L. , & van Eck , N. J. ( 2013 ). A smart local moving algorithm for large-scale modularity-based community detection . European Physical Journal B . https://doi.org/10.1140/epjb/e2013-40829-0.
Waltman , L., van Eck , N. J. , & Noyons , E. C. M. ( 2010 ). A unified approach to mapping and clustering of bibliometric networks . Journal of Informetrics , 4 ( 4 ), 629 - 635 . https://doi.org/10.1016/j.joi. 2010 . 07 . 002.
Wouters , P. ( 2017 ). Eugene Garfield ( 1925 - 2017 ). Nature , 543 ( 7646 ), 492 . https://doi.org/10.1038/ 543492a.
Zhang , G. , Ding , Y. , & Milojevic´, S. ( 2013 ). Citation content analysis (CCA): A framework for syntactic and semantic analysis of citation content . Journal of the American Society for Information Science and Technology , 64 ( 7 ), 1490 - 1503 . https://doi.org/10.1002/asi.22850.