See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259288466

Lee.M.R, and T.T. Chen.( 2012).Revealing research themes and trends in

knowledge management: From 1995 to 2010 Knowledge-Based
Systems.Volume 28, April 2012, Pages 47–58

Article in Knowledge-Based Systems · April 2012

DOI: 10.1016/j.knosys.2011.11.016

CITATIONS READS

113 4,142

2 authors:

Maria R. Lee Tsung Teng Chen

Shih Chien University, Taipei, Taiwan National Taipei University
61 PUBLICATIONS 920 CITATIONS 33 PUBLICATIONS 704 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Tsung Teng Chen on 18 November 2019.

The user has requested enhancement of the downloaded file.

 Knowledge-Based Systems xxx (2011) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Knowledge-Based Systems
journal homepage: www.elsevier.com/locate/knosys

Revealing research themes and trends in knowledge management:

From 1995 to 2010
Maria R. Lee a,⇑, Tsung Teng Chen b
a
Department of Information Technology and Management, Shih Chien University, 70, Ta-Chih Street, Taipei 104, Taiwan
b
Graduate Institute of Information Management, National Taipei University, 151, University Rd., San Shia, Taipei 237, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: Visualizing the entire domain of knowledge and tracking the latest developments of an important disci-
Received 5 July 2011 pline are challenging tasks for researchers. This study builds an intellectual structure by examining a total
Received in revised form 2 October 2011 of 10,974 publications in the knowledge management (KM) field from 1995 to 2010. Document co-citation
Accepted 15 November 2011
analysis, pathfinder network and strategic diagram techniques are applied to provide a dynamic view of
Available online xxxx
the evolution of knowledge management research trends. This study provides a systematic and objective
means in exploring the development of the KM discipline. This paper not only drew its finding from a large
Keywords:
data set but also presented a longitudinal analysis of the development of the KM related studies. The
Knowledge management
Research trends
results of this study reflect that the coverage of key KM papers has expanded into a broad spectrum of
Visualization disciplines. A discussion of the future of KM research is also provided.
Intellectual structure Ó 2011 Elsevier B.V. All rights reserved.
Literature review

1. Introduction The objectives of this study are to map the structure of the KM
field to identify the interconnections of its subfields, to visualize
Because of the enormous number of research papers published, the current research trends to better understand and foresee their
it is important to be able to visualize the growth of scientific impact on the development of the field, and to use the analysis of
knowledge, to reveal the evolution of research themes, and, hence, past research to provide insight into the future direction and/or
to identify inter-relationships within a knowledge domain. How- trends of KM research. This paper analyzes KM research published
ever, it is a challenging task to discover the existence of a scientific from 1995 to 2010 with a total of 10,974 papers. To obtain a dy-
paradigm and the corresponding movements of such a paradigm. namic review of the evolution of KM research trends and to spec-
In addition, the main scientific research themes are very difficult ulate about its future development, the authors subdivided the
to analyze and grasp using traditional methodologies. publications into three time periods: 1995–2000, 2001–2005 and
This study aims to capture and reveal insightful patterns of 2006–2010.
intellectual structures that are shared by researchers in the knowl- We provide a systematic and objective way to capture and sum-
edge management (KM) field. With the advent of the era of the marize the content of those publications, which facilitates visuali-
knowledge economy, the field of KM has received increased atten- zation and which enhances the understanding of research themes
tion from academics and corporate sectors [33,36,48,58–60,66,67]. and related trends in KM. A factor analysis technique is applied as a
Much literature has been published in the KM area to date data reduction and structure detection method [62,63,38]. The
[23,26,25,44,37]. However, most prior studies primarily focus on Pearson correlation coefficients between items (papers) are used
subjectively categorized topic domains, rather than focusing on as the basis for pathfinder network (PFNET) scaling [61]. Specifi-
providing an objective and comprehensive review of the evolution cally, the intellectual structure map of KM can be revealed by
of these domains. In fact, the KM literature is still in its formative applying PFNET to the co-citation graph. The major research
stage, but it has developed substantial gravitas [26]. Consequently, themes and their interrelationships can thus be easily identified
it is crucial to identify the breadth and diversity of its content as via the intellectual structure map. A strategic diagram is used to
well as the magnitude of its progress and evolution. represent the KM themes’ evolution showing the mainstream re-
search themes, under-explored directions, isolated subject areas,
⇑ Corresponding author. Tel.: +886 2 2538 1111x8000; fax: +886 2 25381 and potential areas for future investigation [7,57,17]. This study
111x8003. aims to provide scholars a broad spectrum of interrelated concepts,
E-mail addresses: [email protected] (M.R. Lee), [email protected]. which provide a source for anyone interested KM research and
edu.tw (T.T. Chen). help simulate further interest. In addition, this study also intends

0950-7051/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.knosys.2011.11.016

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
2 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx

to be a quick reference for novice scholars to become familiar with these prior studies, this study uses an updated version of Cocite-
this field of study. Seer, a system to visualize large co-citation networks, to derive
The remainder of this article is structured as follows. In Section the intellectual structure of the domain of KM [13]. A strategic dia-
2, we examine the survey studies on visualizing knowledge struc- gram is also used to provide a global visualization representation of
tures and then present a brief discussion of the existing literature, the structure map of any scientific field or related subfield obtained
reviewing KM research specifically. A discussion of the method from publications appearing in different time periods on the basis
employed in our analysis of the trends of KM research is provided of a co-word analysis [7,57,17].
in Section 3. The next section presents the results. Discussion of the
analysis and the limitations of the approach are presented in Sec- 2.2. Knowledge management review
tion 5. Finally, the conclusion drawn from this study is presented
in Section 6. Research within the field of KM has consistently grown and accu-
mulated great importance in both academia and business, particu-
2. Literature review larly within the last decade. Many KM literature reviews have
been published. Table 1 below summarizes a major list of KM re-
2.1. Visualizing knowledge structure views for the past 10 years. The list shows the author, year of publi-
cation, source of the data, review period, number of reviewed
The study of the intellectual structure of scientific disciplines articles and research methodology. The methodologies taken by
has been receiving increased attention in the academic community the researchers describe the research procedure employed. The ma-
since early eighties. A pioneering bibliometric technique, author jor research methods used in KM reviews are described in Table 2.
co-citation analysis (ACA), is used to discover how scientists in a Dwivedi et al. [19] find organizational and systems context-
particular subject field are intellectually interrelated as perceived based KM research are the most widely published topics. Shannon
by authors in their scientific publications [62]. The basis for the et al. [50] results show country, institutional and individual pro-
visualization of knowledge structures is formed by the interrela- ductivity, co-operation patterns, publication frequency, and favor-
tionships between these elements. Visualizing the knowledge ite inquiry methods in KM and IC studies. Nie et al. [41] explore six
structure is the art of making maps and shares some intrinsic essential issues regarding KM research field, which include: why
semantic characteristics with cartography [10,11]. Visualization the research field is necessary, what enables its birth or triggers ac-
mapping is used to explore large amounts of data and to derive tions on it, what it deals with, how to implement it, how to support
new insights by identifying trends, or clusters, in the data associ- it, and where it has been applied. It concludes that KM could be
ated with a field of study. The map created through citation divided into general, strategy-oriented, information-oriented, hu-
analysis provides a series of historical data, which cover the litera- man-oriented, and process oriented perspectives. Guo and Shef-
ture year by year [20,55]. field [25] study KM theoretical perspectives, research paradigms
The unit of analysis in ACA is constructed by authors who have and research methods and the results show that KM research from
made influential contributions to a subject of interest. An author is positivist, interpretivist, and critical pluralist paradigms. Lee and
recognized by their work, which represents a resource for the con- Chen [35] address the topical content is knowledge engineering,
cept. Co-citation analysis refers to the interrelationship between semantic web and AI related sub-areas. The purpose of [44] is to re-
key concepts. According to the study of [56], authors contributing view and position 20 of the most frequently cited KM articles in
to concepts viewed as overlapping or closely related are more management journals. KM publications focus on knowledge in
likely to be cited together by other researchers than authors con- organizations, knowledge-based, theory of the firm, strategy, and
tributing to concepts viewed as specific or different. Visualizing a knowledge creation. Serenko et al. [51,52] conduct citation analysis
co-citation network is an important technique for illustrating the of individuals, institutions, and countries in KM and intellectual
intellectual structure of a knowledge domain [14]. capital fields. Results indicate the publications from several leading
One commonly used method in author co-citation analysis is Table 1
factor analysis, which has been applied to analyze the essential A list of KM reviews conducted in 2000–2010.
dimensionality of the given co-citation data in a subject domain.
Article Source of data Time Size Research method
The study of [63] demonstrates the author co-citation analysis of period
the information science field and shows that some authors do in-
[19] Web of Science 1974– 1043 Meta analysis
deed simultaneously belong to several specialties. The co-citation 2008
relationships between authors are usually represented by a co- [50] KM/IC journals 1994– 2175 Scientometric
citation matrix, which serves as an input of the factor analysis. 2008 analysis
The co-citation matrix computes a correlation matrix of Pearson [41] Various journals 1987– 1870 Domain analysis
2008
correlation coefficients, which can be used as a measure of similar-
[25] Various journals 2000– 160 Literature review
ity between pairs of authors. Pathfinder network (PFNET) scaling is 2004
used to prune the network defined by the correlation matrix [61]. [35] CiteSeer 1998– 2405 ACA & PFNET
The study by Chen and Steven [12] applies PFNET scaling to extract 2005
the most important relationships from the correlation matrix. [44] Management 1990– 20 Literature review
journals 2002
However, most of the visualizing approaches suffer from a major [23] WoS 1975– 1407 PFNET
shortcoming caused by the limitations of large datasets. If the 2004
datasets are too large, then the size of the corresponding author [51] KM/IC journals 1993– 450 Meta analysis
co-citation matrix could be large, and the analysis becomes compu- 2003
[55] SCI & SSCI 1990– 58 ACA
tationally complicated, expensive, and difficult to formalize [65].
2002
Recent research techniques have been developed to address the [37] Elsevier DB 1995– 234 Literature review
inadequacies of existing co-citation visualization tools. Knowledge 2002
domain visualization (KDV) detects and visualizes emerging trends [45] Various 1994– 180 ACA
and transient patterns in the scientific literature [9,5]. Some recent 1998
[8] Various 1997– 23 Literature review
works in knowledge discovery and data mining systems perform
2001
analysis of the engineering domain [39,40]. After examining all

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx 3

Table 2
Major research methods used in KM reviews.

Meta analysis The process of synthesizing research results by using statistical methods to find the general trend for results across the studies
Scientometric analysis Scientometrics uses the bibliometrics technique to measure the impact of (scientific) publications
Domain analysis The process of identifying, collecting, organizing and representing the relevant information of a domain
Author co-citation Analysis ACA is a specific form of co-citation analysis based on counting highly co-cited authors
(ACA)
Factor analysis Factor analysis is a statistical method to examine how variables may be consolidated and represented by fewer underlying constructs
(factors)
Pathfinder Network (PFNet) Pathfinder networks based on graph theory derive from proximities data for pairs of entities. It keeps only the strongest links between
nodes

authors and foundations are referenced regularly. Liao [37] shows and resources. This database initially was primarily used in the com-
KM technologies are classified as: KM framework, knowledge- puter science field, but now it has included many other fields
based systems, data mining, information and communication tech- [42,43]. The database includes the bibliographic information (meta-
nology, artificial intelligence/expert systems, database technology, data) for research articles published in journals, conferences pro-
and modeling. Chauvel and Despres [8] examine KM research are ceedings, and citations. The proposed procedure leverages the
in six dimensions: phenomena, action, level, knowledge, technol- Microsoft Academic Search citation index by using key phrases to
ogy and outcome. query the index and retrieve all discovered research papers. The doc-
Considering the literature summarized above, the studies of uments retrieved by the query are then used as the initial seed set to
[56,45] apply the ACA methodology in KM studies, whereas the search for papers that cite them or for papers that are cited by them
study of [23] applies PFNET. Furthermore, the study of [45] inves- [14,16].
tigates the intellectual structure and interdisciplinary breadth of The full citation graph is built by linking all papers retrieved,
KM in its early stage of development, searching for the phrase which provides a unified approach to reveal a large data set of
‘‘knowledge management’’ within the research articles from 1994 co-citation relationships. The resulting citation graph was built
to 1998. Four factors were found in the research: knowledge man- from the research papers and citation information retrieved by
agement; organizational learning; knowledge-based theories; and querying the term ‘‘knowledge management’’ from Microsoft Aca-
the role of tacit knowledge in organizations. demic Search in March 2011. The query of ‘‘knowledge manage-
The study of [56] examines the intellectual structure of KM, ment’’ existing in all domains resulted in a total of 10,974
1990–2002, by using the ACA methodology. Eight factors were re- papers. To obtain a dynamic view of the evolution of KM literature
vealed by the research: knowledge as firm capability; organiza- and to be able to predict its future developments, we then ran the
tional information processing and IT support for KM; knowledge publications to generate co-citation graphs in three different time
communication, transfer and replication; situated learning and periods. The first period runs from 1995 to 2000, the second period
communities of practice; practice of KM; innovation and change; runs from 2001 to 2005, and the third period runs from 2006 to
philosophy of knowledge; and organizational learning and learning 2010.
organizations.
The study of [23] concludes that the KM research field has not
yet developed as its own area of study because the KM subject 3.1. The intellectual structure method
highly interacts with other disciplines. Noting the research by
the study of [26], the field of KM is expanding, and it has the poten- The foundation of the intellectual structure method is the cita-
tial to offer a uniform basis for many other disciplines. However, it tion and co-citation analyses. The citation analysis method was
can be argued that there is a major missing link: the need for an in- pioneered by Garfield [20]. It is one of the informatics methods
depth understanding of crucial subjects within the field of KM. This mainly applied in analyzing the citation relationships between
article contributes to the analysis of a large dataset of existing KM documents. It is an analytic tool that can be used to reveal latent
publications by disclosing trends in KM research. information from voluminous literatures [16]. The concept of
Comparing to afore-mentioned reviews, this article applies co-citation was proposed in 1973 by Small, who developed the
three research methods to conduct the KM visualization trend re- citation map as an analytic tool for interpretation of the results
search: document co-citation analysis (DCA), PFNet, and strategic of literature analysis [56]. The citing and cited relation between
diagram. The DCA is to identify major research themes in KM field. the articles of related field may be in the formation of chain, tree,
The PFNet is used to reveal the interrelationships between themes. and network structure. These relationships may be abstracted into
The strategic diagram is applied to visualize the changes, evolution a citation network represented by a graph.
and differences of themes. This article contributes to not only the Citation network reveals the citation relationships (links) be-
analysis of a large dataset of existing KM publications by disclosing tween articles (nodes), it may also expose important nodes via
trends in KM research. It also identifies the major research themes the structure of the network. Two kinds of citation relationships
and their interrelationships in the domain. Further, it reveals are commonly used: the direct citation and the induced co-citation
changes and shows the mainstream research themes, under- relationship. When an article refers to two articles at the same
explored directions, isolated subject areas and potential subject time, a indirect co-citation relationship is established between
areas of knowledge management domain. these two referred articles. The co-citation is an induced relation-
ship derived from the action of citation. It is usually depicted as
an un-directed line between nodes in a graph. Graph based rela-
3. Methodology tionships between multiple nodes are conveniently represented
by a matrix. A 1 in a matrix’s cell indicates the presence of the rela-
The authors constructed a full citation graph from the data tionship. A 0 in a cell denotes the absence of relationship between
drawn from an online citation database—Microsoft Academic the nodes in the corresponding column and row of the matrix. We
Search1—which is a free search engine for academic research papers may have the adjacent matrix A to represent the relationship of di-
rect citation, and the co-citation relationship can be derived by the
1
http://academic.research.microsoft.com/. matrix of ATA (the transpose of the matrix A multiplies with itself).

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
4 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx

The co-citation matrix serves as the input to the factor analysis two-dimensional strategic diagram to represent the concepts of re-
(FA) procedure, which detects the structure of the data and reduces search themes. The horizontal axis represents centrality (% of var-
numerous paper nodes into 10–20 consolidated components. The iance explained), and the vertical axis represents density (number
FA procedure also generates a Pearson correlation coefficients’ of papers within a factor theme).
matrix. The correlation coefficients in the matrix are used as a rel- The themes presented in the quadrant 1consist of mainstream
ativeness measurement. PFNET [49] scaling is then applied to research, which includes both well-developed and important stud-
extract the most important relationships from the correlation ies for structuring a research field. The themes in the quadrant 2
matrix [12]. The PFNET algorithm regards papers as nodes and consist of isolated subject areas, which are both weakly developed
Pearson correlation coefficients as links’ strength; it assumes a and marginal. The themes in the quadrant 3 are peripheral subject
graph in which all nodes are fully connected by weighted links. areas, which have well-developed internal ties but unimportant
The weights are represented by the value of correlation coefficients external ties. Therefore, the studies in the quadrant 3 are only of
for each pair of articles. The topology of a PFNET is determined by a marginal importance for the field. The themes in the quadrant
two parameters q and r, and the corresponding network is denoted 4 are important for the research field but are not yet developed,
as PFNET(r, q). The parameter q constrains the scope of minimum which consequently represents under-explored subject areas.
cost paths to be calculated. The parameter r defines the Minkowski
metric used for computing the distance of a path. The result of
PFNET scaling is shown spatially as a sparsely connected graph 4. Result analysis
whereas the strongest links between nodes in the graph are
preserved. 4.1. Major knowledge management research themes
The steps we applied in building an intellectual structure is
listed as follow: A factor analysis is applied as both a data reduction method (to
reduce the number of variables) and a structure detection method
1. Apply a keyword (‘‘Knowledge Management’’ in this case) (to discover the structure in the relationships between variables).
search into a designated citation database such as Microsoft The factor analysis is also used to merge correlated variables (pa-
Academic Search or CiteSeer. pers) into one factor (theme). As discussed in the previous section,
2. The articles found in the first step are used as seed papers to the co-citation matrix is derived from the citation graph and is fed
expand the searches, whereas the references of the seed papers into the factor analysis. Twenty top-ranked factors are selected as
and the articles cited the seed papers are retrieved. the representative sub-areas of KM. The nature of the factor is
3. Use the citation count as the threshold to filter out less cited determined by papers with an absolute loading of over 0.6 for a
(usually less important) papers. factor. The research described by this factor is given a proper
4. The result from step 3 is transformed into an adjacent matrix A, descriptive name to represent the theme. The review analysis used
the co-citation matrix is derived by ATA. here is based on documents rather than authors, because a scho-
5. Apply factor analysis and derive the Pearson correlation coeffi- lar’s specialty may change over time or evolve [15,35,34]. Table 3
cient matrix. The top 20 components are selected as the main shows the top 20 factors with the percentage of variance explained
research themes. Papers with a factor loading over 0.6 to a com- and the descriptive names of the themes within the three time
ponent are recorded for later use. periods (1995–2000, 2001–2005 and 2006–2010).
6. Use the Pathfinder scaling algorithm to keep only the strongest Note: Because of space constraints in Table 2, CSCW represents
links in the Pearson correlation coefficient matrix, a pathfinder Computer Supported Cooperative Work, K. stands for knowledge,
network (PFNET) graph is derived from the Pathfinder Org. stands for organizational, KMS for Knowledge Management
computation. Systems, SW for software, DB for database, CBR for case-based rea-
7. Draw the PFNET graph. soning, and competitive adv. for competitive advantage.
8. Retrieve papers with significant factor loading recorded in step 5. Each factor reflects a subfield represented by the set of concep-
9. Review, analyze, and synthesize the papers retrieved in step 8. tual ideas contributed by the papers loading significantly on it. The
grouping of concepts into distinct factors refers to the correspond-
Since the factor analysis takes the co-citation matrix as the in- ing features between KM publications building on these ideas. A
put, the value in the matrix cell represents the number of incidence factor is given by a descriptive theme name, which is based on
that two papers are cited together by another paper. A relatively an interpretation of the areas represented collectively by the pa-
high value in the co-citation matrix will result in a higher value pers (concepts) loading on each factor.
of the corresponding Pearson correlation coefficient, which tells For example, factor 1 in the 1995–2000 period represents the
these two papers are frequently selected together by authors. research theme of ontology. A total of 42 ontology-related papers
The variance and loading are metrics produced by the factor anal- were included in the factor, which contains paper title, author, year
ysis, which is applied to consolidate many data items (papers) into of publication, published journal/conference names, and number of
fewer components (research themes). The percentage of variance citations. Within the ontology theme, Gruber’s study [22] has more
of a factor represents the proportion of the total variance of the than 1294 citations, followed by Guarino’s study [24], which has
data set (paper collections) explained by this particular compo- more than 894 citations. Such information can be used by novice
nent. Since a component represents a collection of highly corre- researchers to gain useful insights about the themes of the re-
lated papers, each paper ascribed to a component has a loading search field. Factor 3–14 in 2006–2010 represents the studies con-
value to the component, which is equivalent to the relative mea- ducted on the semantic web. Again, Gruber’s study [21] has 69
surement of papers and their assigned component. A paper, in fact, citations to date.
is attributed to all the components with a varied loading value but A factor analysis not only helps someone new to a particular re-
is assigned to the factor which it has the highest loading. search field find the most cited papers but also helps uncover the
root of the research. For example, factor 1–12, knowledge sharing,
3.2. Theme evolution model in the 1995–2000 period can be dated back as early as a 1972 pa-
per [46]. Factor 3–9, data mining, in the 2006–2010 period may
A modified strategic diagram is applied to reveal the dynamic make reference to a paper by Shannon et al. [50]. Factor 2–7, net-
evolution of KM fields. Following in [57], Fig. 1 shows a proposed work society, in the 2001–2005 period refers to a study by [53].

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx 5

Some themes consistently appear in the three different time 4.3. Interrelationships between research themes
periods. We regard these themes, such as organizational memory,
knowledge creation and KMS, as potential research trends that will The basis for PFNET scaling is utilizing the Pearson correlation
continue to gather interest, evolving and lasting for sometime. coefficients between items (papers). The nodes located close to
Some themes occur only in one period, and this phenomenon the center of a PFNET graph represent papers that contribute to a
may imply a transient nature. However, these themes may merely core concept. The intellectual structure map of KM can be revealed
converge with other themes; for example, text mining and query by applying PFNET to the main research themes. To explicitly visu-
processing approach may converge with data mining. Organiza- alize the relationships between factors in PFNET, Fig. 5 is derived
tional intelligence and software engineering are shown once in from Figs. 2–4 by combining nodes in the same factor into a node
the later period with a high variance, which may suggest that these block to highlight the intellectual structure relationships between
are momentum-gathering emerging studies. themes in the 1995–2000, 2001–2005 and 2006–2010 periods. The
The result of Table 3 indicates the KM themes have expanded into nodes located close to the center of the graph represent papers that
a broad spectrum of disciplines, including business, information contribute to a core concept. The lines indicate a connection rela-
retrieval, database, software engineering, machine learning, etc. tionship. The number inside the box refers to factor number de-
scribed in Table 3.
The intellectual structure map facilitates the identification of
4.2. Visualization of intellectual structure
the research themes and their interrelationships. The position of
the themes in the map also helps to derive the themes evolution
When the factor analysis is applied, Pearson correlation coeffi-
in the strategic diagram shown in Fig. 6.The number inside the
cients between items (papers) are also calculated. Pathfinder net-
box represents the factor number. For example, in 1995–2000, fac-
work (PFNET) scaling is used to derive the essential relationships
tor 1–1 research (ontology) is related to factor 1–13 (software
from the correlation matrix [12]. The value of the Pearson correla-
process), factor 1–14 (distributed database), and factor 1–15 (orga-
tion coefficient should be between 1 and 1. When the coefficient
nizational memory) research. In 2001–2005, factor 2–7 (network
is approaching 1, it means two items (papers) simply correlate
society) research is related to factor 2–11 (knowledge transfer),
completely. Items that are closely related represent the fact that
factor 2–14 (knowledge sharing), and factor 2–16 (innovation)
the papers are highly correlated and should be placed close to-
research. In 2006–2010, factor 3–1 (organizational intelligence) re-
gether graphically. The distance between items is inversely propor-
search is co-related to factor 3–2 (software engineering) and factor
tional to the correlation coefficient, which depicts less correlated
3–5 (organizational memory) research.
items apart and highly correlated items close together.
The interrelationship provides a truly novel interpretation
Figs. 2–4 shows PFNET scaling of KM drawn from Microsoft Aca-
whereby the dependency of variables can be further analyzed. This
demic Search for the following three time periods: 1995–2000,
approach should be useful to study the causal structure within the
2001–2005, and 2006–2010. Articles under the same factor are
situation being investigated.
actually painted with the same color. Each article is assigned with
a unique number. The number in the parentheses is the factor
number that the article belongs to. The factor number is assigned 4.4. Research themes evolution representation
based on the number in Table 3, apart from 1995–2000 period
without a prefix (1-), 2001–2005 period without a prefix (2-), A strategic diagram is used to represent a dynamic view of the
and 2006–2010 period without a prefix (3-). Cyan nodes with (0) evolution of knowledge management research trends. Fig. 6
represent articles that are not assigned to any factor due to insuf- presents the KM themes’ evolution over these different time peri-
ficient factor loading. For example, 0(405) means the article #405 ods in the proposed strategic diagram. The diagram situates each
does not assign to any factor due to low factor loading. The nodes theme area within a two-dimensional space divided into four
located close to the center of a PFNET graph represent papers quadrants. Quadrant 1 represents mainstream research themes,
contributing to a fundamental concept and in the mainstream of quadrant 2 stands for isolated subject areas, quadrant 3 is periph-
a research domain. eral subject areas, and quadrant 4 represents under-explored

Fig. 1. A modified strategic diagram.

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
6 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx

Table 3
Top 20 KM research themes for three time periods.

1995–2000 Factor # Themes (% of variance) 2001–2005 Factor # Themes (% of variance) 2006–2010 Factor # Themes (% of variance)
1–1 Ontology (8.984) 2–1Contextual reasoning (5.983) 3–1Org. intelligence (11.313)
1–2 Concept map (7.277) 2–2 K. integration (5.402) 3–2 SW engineering (9.266)
1–3 CSCW (6.085) 2–3 Firm resource (5.16) 3–3 KM model (5.265)
1–4 K. creation (6.019) 2–4 SW engineering (4.911) 3–4 Multi-agent (3.893)
1–5 Text mining (5.914) 2–5 Company failures (4.865) 3–5 Org. memory (3.607)
1–6 Temporal model (5.37) 2–6 K. structure (4.166) 3–6 KM problems (3.455)
1–7 K. discovery (5.195) 2–7 Network society (4.038) 3–7 Information systems (3.305)
1–8 Query processing (4.299) 2–8Competitive adv. (3.871) 3–8 Intellectual capital (3.278)
1–9 KMS (4.108) 2–9 Situated learning (3.452) 3–9 Data mining (2.719)
1–10 Machine learning (3.84) 2–10 Org. memory (3.311) 3–10 KMS (2.542)
1–11 KM techniques (3.207) 2–11K. transfer (2.876) 3–11 Firm resources (2.505)
1–12 K. sharing (3.157) 2–12 Distributed org. (2.834) 3–12 K. creation (2.297)
1–13 SW engineering (3.006) 2–13 KM model (2.743) 3–13 Innovation (2.292)
1–14 Distributed DB (2.997) 2–14 K. sharing (2.63) 3–14 Semantic web (2.27)
1–15 Org. memory (2.986) 2–15 K. creation (2.559) 3–15 K. Reuse (2.265)
1–16 User model (2.928) 2–16 Innovation (2.367) 3–16 Situated learning (2.216)
1–17 WWW (2.488) 2–17 K. engineering (2.34) 3–17 Task-based KM (2.097)
1–18 Business process (2.427) 2–18 KMS (2.289) 3–18 SW development (1.98)
1–19 Internet access (2.017) 2–19 Intellectual capital (2.052) 3–19 K. level (1.888)
1–20 CBR (1.894) 3–10 KM diagnostic (1.957) 3–20 KM Challenge (1.871)

subject areas [56]. The circle corresponds to the position of the processes that have vital implications for organizational longevity.
theme from 1995 to 2000, the square represents the position of Wisdom can be defined as a mode of symbolic processing by a
the theme from 2001 to 2005, and the diamond shows the position highly developed will [28,2].
of the theme from 2006 to 2010. The position of each theme in the Table 4 shows a list of the KM themes’ evolution over different
strategic diagram is computed based on the themes position in periods based on the strategy diagram in Fig. 7. The unique theme
Fig. 5, % of variance in Table 3, and number of cited papers. The re- number shown before the theme name is based on the description
sults enable us to evaluate the evolution of the interrelationships in Table 3. The green broken line represents the theme movement
between the themes. from 1995–2000 to 2001–2005 time periods. The red line shows
The four quadrants are defined as above. The number repre- the theme movement from 2001–2005 to 2006–2010 time periods.
sents the theme name related in Table 3. The higher the percentage The orange line represents the movement from 1995–2000 to
of variance explained by a factor, the more papers belong to this 2006–2010 time periods. The list demonstrates the theme move-
factor. Therefore, a factor with a high percentage of variance that ment in each time period, which will subsequently be discussed
is explained may appear as a popular or mainstream theme (center in detail. The list facilitates the visualization of the research themes
of a structure map, high percentage of variance). A peripheral area trend.
may be uncovered by investigating the intellectual structure map. Software engineering moves from an under-explored subject
Peripheral areas are usually positioned at the edge of a structure area (factor 1–13 in 1995–2000) toward a mainstream subject area
map (edge, low percentage of variance). The number of citations (factor 2–4 in 2001–2005 and factor 3–2 in 2005–2010). This is in
a paper received signals the important role played by the highly ci- line with [30], which argues that information science (IS) is not
ted paper. Under-explored subject areas may be identified with the contributing to the advancement of KM as much as it should be
structure map by finding a factor with low explained variance but in the past due to different professions are contributed to and
with a positioned closed to the center of the structure map (center, influenced the development of KM in their own ways. [30] sums
low percentage of variance). Isolated subject areas may be identi- up that KM is a natural development within IS.
fied by the structure chart if a branch is broken away from the Some of the core issues of organizational memory include organi-
main trunk of the structure chart or if papers belonging to an zational context, retention structure, organizational learning and so
edge-positioned factor connect with papers belonging to one other forth. The research focus of organizational memory moves from an
factor (edge, high percentage of variance). under-explored subject area (center, low variance) (factor 1–15 in
The themes evolving toward the upper right corner of the map 1995–2000) to an isolated subject area (edge, high variance) (factor
represent papers contributing to a fundamental concept. The main- 10 in 2001–2005), and then, it advances to a mainstream research
stream research areas in 2006–2010 are factors 3–1, 3–2, 3–3, 3–5, area (center, high variance) (factor 3–5 in 2006–2010). These trends
and 3–7, which represent research in the areas of organizational show an increased interest in organizational memory research. The
intelligence, software engineering, KM models, organizational study of [6] presents the analysis of the use of organizational mem-
memory, and information systems. The mainstream research areas ory from 1991 to 2001 to assess whether progress has been made in
in 1995–2000 are ontology, computer support cooperative work theoretical development and empirical research. The study of [6]
(CSCW), knowledge creation, temporal model, and KMS (factors concludes that despite progress in organizational memory related
1–1, 1–3, 1–4, 1–6, 1–9, respectively), whereas in 2001–2005, the research, there is still evidence of fragmentation in the literature.
areas are knowledge integration, software engineering, company Our result reflects the finding that the organizational memory is
failures, knowledge structure, network society, and competitive an under-specified and multidisciplinary construct during the
advantage research (factors 2–2, 2–4, 2–5, 2–6, 2–7, 2–8, 1995–2005 period. Over the years, the important organizational
respectively). memory attributes are management of total quality and decision
The mainstream of 2006–2010 is organizational intelligence, making process improvements within organizations. The modern
which consistent with the chain of knowledge flow (data–informa- workplace with experiential learning or evidence-based practice is
tion–knowledge–wisdom), also known as knowledge pyramid the current focus in organizational memory [32].
[1,28]. According to [1], wisdom is the ability to increase effective- To add value with knowledge management, knowledge man-
ness, especially in strategic decision making and visioning agement systems (KMS) facilitate the generation and sharing of

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx 7

Fig. 2. PFNET scaling of KM drawn from 1995 to 2000. Each node in the graph represents an article, whereas the number by the node is a unique article number, and the
number in the parenthesis is the factor number this article belongs.

knowledge. Bonner [4] explains that approximately half of US com- practice and application. Effective knowledge management imple-
panies worked on building knowledge management systems mentation has become increasing important to enhance innovation
(KMS) in early 1999. This finding is consistent with the KMS theme [18]. Knowledge reuse has been considering as a major justification
revealed in factor 1–9 in 1995–2000 (a mainstream research). Be- for KM, for example, no need to reinvent solutions, but offer pro-
cause there is no single model for KMS, various forms and formats ductivity gains [29].
appear in different industries [47]. The KMS theme (factor 2–18)
appears to become a peripheral subject area in 2001–2005. KMS
plays a central role in managing knowledge from various sources, 5. Discussion
such as employees, customers, business partners, etc. KMS facili-
tates knowledge acquisition, creation, integration, transfer and The review result is critical affected by the data set adopted. As
applications [2]. The position of the KMS theme (factor 3–10, in June of 2011, the Microsoft Academic Search (MAS) includes ele-
2006–2010) shows the research is mature enough for this role. ven domains and one ‘‘other’’ domain. The other domain brings in
The positions of the themes enable us to identify those which top journals such as PNAS, Nature, Science, Synthese, Lingua etc.
are of central importance to the region initially, but shift later to The other category contains high-impact articles that may appeal
become of secondary interest. For example, Ontology (factor 1–1) to the broad scientific community as well as the general public.
in 1995–2000 represents a mainstream research area (center, high The MAS has had fourteen domains at the beginning of September.
variance). However, in 2006–2010, semantic web (ontology) (fac- A query made at the 4th of September to the MAS using ‘‘knowl-
tor 3–14) appears to be a peripheral subject area (edge, low edge management’’ as the key phrase returns 12,519 articles. The
variance). This fact may imply that these aforementioned studies top cited paper from these articles is authored by Alavi and Leidner
may be well developed internally but that they are only of a [2]. If we carry out the same query but limit our search to the com-
marginal importance for the field. puter science domain, 9222 articles would have been returned. The
The positions of the themes also indicate potentially fruitful re- ratio between the number of computer science and non-computer
search areas such as KM challenge (3–20), knowledge reuse (3–15) science related articles is approximately 73.7%. In contrast with
and innovation (3–13) in 2005–2010 KM studies. The three themes MAS, the CiteSeer database is primarily indexing articles in the
show they are important research themes, but under-explored, fields of computer and information science and engineering. A
immature yet. While the field of knowledge management has long query into the Scopus and Google Scholar database will also return
been studied by scholars, but many challenges remain. These chal- Alavi and Leidner’s paper as the top cited article. At the date of 4th
lenges reside in both theoretical and conceptual studies as well as of September, MAS indexes 27,170,758 articles from fourteen

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
8 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx

Fig. 3. PFNET scaling of KM drawn from 2001 to 2005.

Fig. 4. PFNET scaling of KM drawn from 2006 to 2010.

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx 9

index the major KM journals such as Journal of Knowledge Man-

agement, and KM World Magazine. Furthermore, many studies
conducted in the area of KM use WoS, and this fact may limit the
search of relevant articles.
This article searched the research indexed by Microsoft Aca-
demic Search, which is a free academic search engine for object-
level vertical searches [41]. The database contains broad domains
with citation contexts. The search results for ‘‘knowledge manage-
ment’’ in all domains, including computer science, engineering,
physics [13,3,31,54], chemistry, mathematics and other domains,
listed a total of 10,974 publications as of March 2011.
The major output of scientific research is indeed publications.
We have identified several key trends by studying the KM research
analyzed in this article. From the analysis of the publications and
their citations retrieved from Microsoft Academic Search on
knowledge management, we have found that there is a fragmented
hypothesis in terms of the range of subject research topics in KM.
Similar to the prior studies conducted by Gu [23] and Holsapple
and Wu [26], this study highlights that KM is an expanding field
that has the potential to offer a unifying basis for many other dis-
ciplines. The cross-disciplinary subject areas identified in our find-
ings include information retrieval, software engineering, machine
learning, distributed databases, multi-agent systems, and data
mining.
In terms of the methodological and theoretical dimensions of
the disciplines, the study of [19] addresses the finding that there
is need for a broader review and analysis of the KM literature in
general. We have demonstrated how a method or a theory evolves
and how the evolution of these themes can influence the field of
KM. For example, research themes found in 1995–2000 seem pre-
dominately related to the Internet, such as search engine and web
information categorization. In 2006–2010, semantic web promises
to provide a common framework that allows data to be shared and
reused across different applications, which is consistent with the
current web 2.0 evolutions.
The study of [56] addresses the way that the dynamics of cita-
tions and citation patterns [38] provide useful insights into the
conceptual structure of a field and allow researchers to envision re-
search streams and the linkages among them. The study of [45] ap-
plies the ACA derived from the 1994–1998 publications in the KM
field and summarizes that the interdisciplinary breadth surround-
ing KM was discovered mainly in the discipline of management.
This study validates the hypothesis with empirical evidence that
the discipline of Computer Science is not a key contributor in KM
during that time. However, computing technology presents a par-
adigm shift where information technology has become a rapidly
advancing and expanding research and development field over
the last decade. Our study provides factual confirmation that the
discipline of computer science currently plays a key role in KM.
Fig. 5. Intellectual structure map of KM from three time periods. The study of [56] examines KM literature from 1990 to 2002 by
using ACA and highlights the intellectual structure of management
domains. It would be a fair assertion that MAS is a representative related publications in the field. The sub-fields reflect the influence
multi-domain database. of KM on a wide range of basic disciplines such as philosophy and
In contrast with subjectively paper selection process, our meth- economics.
od makes an important paper being selected naturally. A paper will Small [55] discusses that map the structure of sciences provides
not be selected unless it is (1) indexed by the academic search en- a snapshot of evolution of knowledge in micro-level and macro-le-
gine, (2) passed a citation count threshold, and (3) loaded signifi- vel. The micro-level helps to deal with histories of individual scien-
cantly to a component. tific evolving whereas macro-level reveals the changes between
Lee and Chen [34] compare publication extracts from Web of interrelationships in the entire bodies of knowledge. The article
Science (WoS) and the CiteSeer databases in the ubiquitous com- visualizes the structure of KM knowledge in both macro and micro
puting field. The study concludes that differences exist between levels, which facilitates the tracking of structural evolution of sci-
the two databases: the CiteSeer citation index is primarily a com- entific knowledge [11]. This paper not only drew its finding from
puter, information science, and engineering citation database, a large data set but also presented a longitudinal analysis of the
whereas WoS reveals application and business themes as well as development of the KM related studies. We report the phenomenal
technical themes. The study of [23] points out that WoS does not evolution of the KM researches.

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
10 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx

Fig. 6. Themes evolution over three time periods in the proposed strategic diagram.

Table 4
A list of the KM themes’ evolution across different time periods.

Themes 1995–2000 2001–2005 2006–2010

Software engineering (1–13) Under-explored subject area (2–4) Mainstream research (3–2) Mainstream research
Organizational memory (1–15) Under-explored subject area (2–10) Isolated subject (3–5) Mainstream research
Knowledge management systems (1–9) Mainstream research (2–18) Peripheral subject (3–10) Isolated subject
Knowledge creation (1–4) Mainstream research (2–15) Peripheral subject (3–12) Peripheral subject
Ontology (1–1) Mainstream research (Not in the top 20 list) (3–14) Peripheral subject

Fig. 7. Themes changes across different time periods.

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx 11

The motivation behind our investigation is to adopt a past- research themes, but under-explored, immature yet. This finding
future orientation by analyzing the past to prepare for the future indicates they are potentially fruitful research areas, which may
in the KM field [64]. The analysis creates a firm foundation for facilitate researchers targeting their KM studies.
advancing knowledge, which facilitates KM theory development, We believe this study could be useful for a wide range of users,
closes areas where a plethora of research exists, and uncovers notably scientists, researchers and librarians. It can also facilitate
areas where research is needed [55]. The outcome of the analysis the novice researcher in gaining useful and interesting insights into
provides an objective assessment of a large number of research the exciting field of KM study. However, the research method could
papers. not exclude some identified limitations. Future research is encour-
We acknowledge that this study has a number of limitations. aged to overcome the limitations.
The research method could not exclude the phenomenon of
self-citation and self-reference [27]. Secondly, the use of Microsoft
Academic Search as the data source might be a limitation, but it did References
result more data than any previous studies. We have done some
[1] R. Ackoff, From data to wisdom, Journal of Applies Systems Analysis 16 (1989)
preliminary evaluation between several data sources, but an 3–9.
exhaustive comparison is not feasible. Finally, the limitation of [2] M. Alavi, D. Leidner, Knowledge management and knowledge management
the co-citation method is that the newly published paper may systems: conceptual foundations and research issues, MIS Quarterly 25 (1)
(2001) 107–136.
not have enough time to make an impact. A future study utilizes [3] O. Aljayyous, Greywater reuse: knowledge management for sustainability,
the bibliographical coupling method for the latest period may Journal of Desalination 167 (2004) 27–37.
overcome this limitation. [4] D. Bonner, Leading Knowledge Management and Learning, American Society of
Training and Development, Virginia, 2000.
[5] K. Borner, C. Chen, K. Boyack, Visualizing knowledge domains, Annual Review
of Information Science and Technology 37 (2002) 179–255.
6. Conclusion [6] A. Casey, F. Olivera, Learning from the past: are view of the organizational
memory literature, in: 5th International Conference on Organizational
Learning and Knowledge, 2003, pp. 1–27.
In this article, we have provided a visualization overview of the [7] T. Cahlik, Comparison of the maps of science, Scientometric 49 (3) (2000) 373–
wide distribution of KM publications by presenting a systematic 387.
[8] D. Chauvel, C. Despres, A review of survey research in knowledge
and analytical study during 1995–2010 period. We have visualized management: 1997–2001, Journal of Knowledge Management 6 (3) (2002)
large co-citation networks and synthesized the variety of concepts 207–223.
derived from the intellectual structure of the domain of KM. To [9] C. Chen, CiteSpace II: Detecting and Visualizing Emerging Trends and Transient
Patterns in Scientific Literature, JASIST, 2005.
capture the dynamic change within the evolution of KM, we have
[10] C. Chen, Visualization of Knowledge Structures, Handbook of Software
sub-divided the fifteen-years of publications into three time peri- Engineering and Knowledge Engineering, vol. 2, 2002, pp. 700–744.
ods, and the position of each theme in the proposed strategic dia- [11] C. Chen, R.J. Paul, Visualizing a knowledge domain’s intellectual structure,
gram was computed. The results of our investigation reflect that Computer 34 (2001) 65–71.
[12] C. Chen, M. Steven, Visualizing evolving networks: minimum spanning trees
the key contributors to KM papers spanned across a broad spec- versus pathfinder networks, in: IEEE Symposium on Information Visualization,
trum of disciplines. The evolution and development of KM depends 2003, pp. 67–74.
largely on the extensions and enrichment of the concepts charac- [13] H.H. Chen, C. Pang, Organizational forms for knowledge management in
photovoltaic solar energy industry, Knowledge-Based Systems 23 (8) (2010)
terizing these research streams. 924–933.
The results highlight important issues for future research in the [14] T.T. Chen, D. Yen, CociteSeer: a system to visualize large cocitation networks,
KM field. To the extent that the research themes reflect the evolu- Electronic Library 28 (4) (2010) 477–491.
[15] T.T. Chen, M. Lee, Revealing themes and trends in the knowledge domain’s
tion of the KM field as a whole, the research trends that we have intellectual structure, Lecture Notes in Computer Science 4303 (2006) 99–107.
observed raise an important question about the future of the KM [16] T.T. Chen, L. Xie, Identifying critical focuses in research domains, in:
field: ‘‘Is the KM field likely to focus on a dominant paradigm or Proceedings of 9th International Conference on the Information Visualization
(IV’05), 2005, pp. 135–142.
to fragment itself into a myriad of subfields in the future?’’ The re- [17] J.P. Courtial, J. Law, A co-word study of artificial intelligence, Social Studies of
sults of our analysis using the proposed strategic diagram show Science 19 (1988) 301–311.
that most of the themes are spreading around the center of the [18] J. Darroch, Knowledge management, innovation and organization
performance, Journal of Knowledge Management 9 (3) (2005) 101–115.
map. This finding may imply that the movement is supporting nei-
[19] Y. Dwivedi, K. Venkitachalam, A. Sharif, W. Al-Karaghouli, V. Weerakkody,
ther the fragmentation hypothesis nor the dominant hypothesis. Research trends in knowledge management: analyzing the past and predicting
This finding may also imply that the research trend in KM is still the future, Information Systems Management 28 (2011) 43–56.
evolving and that it has not yet reached its maturity. However, this [20] E. Garfield, Citation indexes for science: a new dimension in documentation
through association of ideas, Science 122 (1975) 108–111.
interpretation of our findings needs more support through further [21] T. Gruber, Ontology of folksonomy: a mash-up of apples and oranges,
investigation. International Journal on Semantic Web and Information Systems (2005).
We have examined the intellectual structure of KM by sketching [22] T. Gruber, Toward principles for the design of ontologies used for knowledge
sharing, International Journal of Human–Computer Studies 43 (5) (1995) 907–
the knowledge management intellectual landscape through visual- 928.
izations of document co-citation behaviors. Our finding provides a [23] Y. Gu, Global knowledge management research: a bibliometric analysis,
review of prior concept [55], which facilitates us gaining a better Scientometrics 61 (2) (2004) 171–190.
[24] N. Guarino, Formal ontology and information systems, in: International
understanding of the themes of KM by tracing the research path Conference on Formal Ontology in Information Systems, 1998.
and mapping the paradigm shift. The proposed methodologies [25] R. Guo, J. Sheffield, Paradigmatic and methodological examination of
can be easily applied to other disciplines and provides a powerful knowledge management research: 2000–2004, Decision Support Systems 44
(3) (2007) 673–688.
research tool for understanding the epistemology of a field as it [26] C.W. Holsapple, C.W.J. Wu, In search of a missing link, Knowledge
evolves. Management Research and Practice 6 (1) (2008) 31–40.
Although this study cannot claim to be exhaustive, it does pro- [27] K. Hyland, Self-citation and self-reference: credibility and promotion in
academic publication, Journal of the American Society for Information
vide reasonable insights into the state-of-the-art in KM research.
Science and Technology 54 (3) (2003) 251–259.
The results presented in this article have several important impli- [28] N. Kakabadse, A. Kakabadse, A. Kourmin, Reviewing the knowledge
cations. There is no doubt that KM research will continue growing management literature: towards a taxonomy, Journal of Knowledge
significantly in the future. New methods should be developed for Management 7 (4) (2003) 75–91.
[29] A. Kankanhalli, O.K. Lee, K.H. Lim, Knowledge reuse through electronic
coping with the topic of KM challenge, knowledge reuse and inno- repositories: a study in the context of customer service support, Information
vation in KM studies. The three themes show they are important & Management 48 (2011) 106–113.

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
12 M.R. Lee, T.T. Chen / Knowledge-Based Systems xxx (2011) xxx–xxx

[30] G. Kebede, Knowledge management: an information science perspective, [49] R. Schvaneveldt, Pathfinder Associative Networks: Studies in Knowledge
International Journal of Information Management 30 (2010) 416–424. Organization, Ablex Publishing Corp., Norwood, New Jersey, 1990.
[31] A. Kosilov, M. Mandl, Z. Pasztory, M. Gladyshev, M. Idrissova, T. Zhantikin, A [50] C. Shannon, W. Weaver, The Mathematical Theory of Communication,
knowledge management and information portal for the Kazakhstan Atomic University of Illinois Press, Urbana, 1998.
Energy Committee, International Journal of Nuclear Knowledge Management 3 [51] A. Serenko, N. Bontis, L. Booker, K. Sadeddin, T. Hardie, A scientometric analysis
(4) (2009). of knowledge management and intellectual capital academic literature (1994–
[32] A. Kransdorff, Corporate DNS, Gower Publishing, 2006. 2008), Journal of Knowledge Management 14 (1) (2010) 3–23.
[33] G.B. Laleci, G. Aluc, A. Dogac, A. Sinaci, O. Kilic, F. Tuncer, A semantic backend [52] A. Serenko, N. Bontis, Meta-view of knowledge management and intellectual
for content management systems, Knowledge-Based Systems 23 (8) (2010) capital literature: citation impact and research productivity rankings,
832–843. Knowledge Process Management 11 (3) (2004) 185–198.
[34] M. Lee, T.T. Chen, Visualizing intellectual structure in ubiquitous computing, [53] H. Simon, The Sciences of the Artificial, third ed., MIT press, Cambridge, Mass,
Lecture Notes in Artificial Intelligence 6232 (2010) 261–272. 1996.
[35] M. Lee, T.T. Chen, Visualizing trends in knowledge management, Lecture Notes [54] N. Singpurwalla, Knowledge management and information superiority, Journal
in Computer Science 4798 (2007) 362–371. of Statistical Planning and Inference 115 (2) (2003) 361–364.
[36] B.M. Li, S.Q. Xie, X. Xu, Recent development of knowledge-based systems, [55] H. Small, Macro-level changes in the structure of co-citation clusters: 1983–
methods and tools for one-of-a-kind production, Knowledge-Based Systems 1989, Scientometrics 26 (1) (1993) 5–20.
24 (7) (2011) 1108–1109. [56] M. Subramani, S.P. Nerur, R. Mahapatra, Examining The Intellectual Structure
[37] S.H. Liao, Knowledge management technologies and applications: literature of Knowledge Management, 1990–2002 – An Author Co-citation Analysis,
review from 1995 to 2002, Expert Systems with Applications 25 (2003) 155– Management Information Systems Research Center, University of Minnesota,
164. 2003, p. 23.
[38] K. McCain, Mapping authors in intellectual space: a technical overview, Journal [57] W.A. Turner, F. Rojouan, Evaluating input/output relationships in a regional
of the American Society for Information Science 41 (6) (1990) 433–443. research network using co-word analysis, Scientometrics 22 (1) (1991) 139–
[39] J. Mothe, B. Dousset, Mining document contents in order to analyze a scientific 154.
domain, in: Sixth International Conference on Social Science Methodology, [58] R. Vandaie, The role of organizational knowledge management in successful
2004. ERP implementation projects, Knowledge-based Systems 21 (2008) 920–926.
[40] J. Mothe, C. Chrisment, T. Dkaki, B. Dousset, S. Karouach, Combining mining [59] S. Wan, T.C. Lei, A knowledge-based decision support system to analyze the
and visualization tools to discover the geographic structure of a domain, debris-flow problems at Chen-Yu-Lan River, Taiwan, Knowledge-Based
Computers, Environment and Urban Systems 30 (2006) 460–484. Systems 22 (8) (2009) 580–588.
[41] K. Nie, T. Ma, Y. Nakamori, An approach to aid understanding emerging [60] H. Wei, Q.X. Xu, X.S. Tang, A knowledge-based problem solving method in GIS
research fields – the case of knowledge management, Systems Research and application, Knowledge-Based Systems 24 (4) (2011) 542–553.
Behavioral Science 26 (2009) 629–643. [61] H.D. White, Pathfinder networks and author co-citation analysis: a remapping
[42] Z. Nie, J. Wen, W.Y. Ma, Object-level vertical search, in: Conference on of paradigmatic information scientists, Journal of American Society of
Innovative Data Systems Research (CIDR), 2007, pp. 235–246. Information Science & Technology 54 (2003) 423–434.
[43] Z. Nie, Y. Zhang, J. Wen, Y. Ma, Object-level ranking: bring order to web objects, [62] H.D. White, B.C. Griffith, Author cocitation: a literature measure of intellectual
in: Proceedings of the 14th International World Wide Web Conference, 2005. structure, Journal of American Society for Information Science 32 (1981) 163–
[44] I. Nonaka, V. Peltokorpi, Objectivity and subjectivity in knowledge 171.
management: a review of 20 top articles, Knowledge and Process [63] H. White, H.K. McCain, Visualizing a discipline: an author co-citation analysis
Management 13 (2) (2006) 3–82. of information science, Journal of American Society for Information Science 49
[45] L.J. Ponzi, The intellectual structure and interdisciplinary breadth of (4) (1995) 327–356.
knowledge management: a bibliometric study of its early stage of [64] J. Webster, R. Watson, Analyzing the past to prepare for the future: writing a
development, Scientometrics 55 (2002) 259–272. literature review, MIS Quarterly 26 (2) (2002) 8–13.
[46] K.R. Popper, Objective Knowledge: An Evolutionary Approach, Clarendon [65] F. Wei, C. Zhiming, Research on knowledge management visualization of deep
Press, Oxford, 1972. web, Key Engineering Materials 439 (2009) 189–194.
[47] M. Quaddus, J. Xu, Adoption and diffusion of knowledge management systems: [66] Y.F. Wen, An effectiveness measurement model for knowledge management,
field studies of factors and variables, Knowledge-based Systems 18 (2005) Knowledge-Based Systems 22 (5) (2009) 363–367.
107–115. [67] Y. Xu, A. Bernard, Quantifying the value of knowledge within the context of
[48] J.T. Rodríguez, B. Vitoriano, J. Montero, A natural-disaster management DSS for product development, Knowledge-Based Systems 24 (1) (2011) 166–175.
humanitarian non-governmental organizations, Knowledge-Based Systems 23
(1) (2010) 17–22.

Please cite this article in press as: M.R. Lee, T.T. Chen, Revealing research themes and trends in knowledge management: From 1995 to 2010, Knowl. Based
Syst. (2011), doi:10.1016/j.knosys.2011.11.016
View publication stats