You are on page 1of 17

J Comput High Educ (2012) 24:1–17

DOI 10.1007/s12528-011-9044-9

Examining mobile learning trends 2003–2008:


a categorical meta-trend analysis using text mining
techniques

Jui-Long Hung • Ke Zhang

Published online: 10 March 2011


 Springer Science+Business Media, LLC 2011

Abstract This study investigated the longitudinal trends of academic articles in


Mobile Learning (ML) using text mining techniques. One hundred and nineteen
(119) refereed journal articles and proceedings papers from the SCI/SSCI database
were retrieved and analyzed. The taxonomies of ML publications were grouped into
twelve clusters (topics) and four domains, based on abstract analysis using text
mining. Results include basic bibliometric statistics, trends in frequency of each
topic over time, predominance in each topic by country, and preferences for each
topic by journal. Key findings include the following: (a) ML articles increased from
8 in 2003 to 36 in 2008; (b) the most popular domain in current ML is Effectiveness,
Evaluation, and Personalized Systems; (c) Taiwan is most prolific in five of the
twelve ML clusters; (d) ML research is at the Early Adopters stage; and (e) studies
in strategies and framework will likely produce a bigger share of publication in the
field of ML.

Keywords Mobile learning  M-learning  Text mining  Bibliometrics 


Mobile learning trends

J.-L. Hung (&)


Department of Educational Technology, Boise State University,
1910 University Drive, Boise, ID 83725-1747, USA
e-mail: andyhung@boisestate.edu
URL: http://edtech2.boisestate.edu/hungj/web/index.html

K. Zhang
Instructional Technology Program, Wayne State University,
385 Education Building, 5425 Gullen Mall, Detroit, MI 48202, USA
e-mail: ke.zhang@wayne.edu
URL: http://itlab.coe.wayne.edu/kzhang/

123
2 J.-L. Hung, K. Zhang

Introduction

With rapid advances in technology, mobile devices have become widely available
and progressively more affordable. The practice of and research on mobile learning
(ML) are also noticeably increasing worldwide (Attewell 2005; Kim et al. 2008;
Motlik 2008; Zhang 2008). Almost four decades after the first generation of single-
functional mobile technology on the consumer market in 1972 (Polsson 2009),
today all-in-one mobile devices with multiple functions, such as phone, networking,
GPS, camera, music player, digital recorder, and a lot more, are popular. The
diverse functionalities, in addition to high mobility, have naturally led to more
research, development and implementation of ML, or m-learning. Such practices
and research are expected to continuously increase at a rather rapid pace in the years
to come. A large number of studies on ML have been conducted, especially in
recent years. However, there is a severe shortage of shared understanding on some
critical questions regarding ML research. For example, what are the major interests
and foci in ML research? How has research on ML changed, emerged, or diverged
in the past decades? What countries and areas have produced the most academic
articles on ML? What are the observed trends in related research and publication?
This study reported herein, through the use of descriptive research methods (Gall
et al. 1996, pp. 373–380), investigated longitudinal trends of ML research with text
mining techniques. This study aims to answer the following questions:
• What are the ML publication bibliometrics?
• What are the taxonomies in ML research as indicated in abstracts?
• What are the trends of research on ML across time?
This study provides a longitudinal, as well as micro viewpoint on prestigious
publications of ML, and identifies the more frequently researched topics and
themes. More importantly, it helps to discover those yet to be further studied.
Accordingly, this research generates guidance and recommendations for selecting
research topics as well as publication venues on ML. It also provides practical
implications for practitioners in their search for relevant literature on ML. This
study also frames a comprehensive overview on the past and current status of
published research on ML, which may in turn furnish a solid foundation for both
historical and/or meta-analyses of the increasingly published topic of ML.

Literature review

This section starts by defining the research topic and reviewing prior studies, then
discusses the major methods used in the study: bibliometrics and text mining.

Definition of m-learning

Lehner and Nösekabel (2002) defined mobile education as ‘‘any service or facility
that supplies a learner with general electronic information and educational content
that aids in the acquisition of knowledge regardless of location and time’’ (p. 100).

123
TM in m-learning 3

Sharma and Kitchens (2004) defined ML as ‘‘learning that is supported by mobile


devices, ubiquitous communications technology and intelligent user interfaces’’
(p. 205). Attewell and Savill-Smith (2004) emphasized the focus on learners’
mobility, instead of technology itself, and operationalized ML in the 3-year pan-
European m-learning project. Geddes (2004) defined ML as ‘‘the acquisition of any
knowledge and skill through using mobile technology, anywhere, anytime, that
results in an alteration in behavior’’ (p. 1). Similarly, scholars and researchers define
ML with close references to the use of mobile technology to achieve anytime,
anywhere, ubiquitous learning.

Mobile learning research

Due to the relatively short history of ML, the authors could not find any studies
focused on its longitudinal research trends. Since ML can be regarded as a part of
e-learning, this section attempts to build a basic understanding of ML research
trends by summarizing e-learning review research.
In general, e-learning review research was conducted in the following three
ways: (a) author’s own observations (e.g., Lockwood 2007), (b) selective article
review (e.g., Mihalca and Miclea 2007; Winn 2002), and (c) systematic investi-
gation (e.g., Hung 2010). These studies summarized the following conclusions:
First, the evolution of e-learning has been classified into different phases based on
researchers’ varying criteria. For example, Winn (2002) defined the evolution of
e-learning into the age of instructional design, the age of message design, the age
of simulation, and the age of learning environments. Second, the development of
technology is highly correlated to the phase change and its related research. For
example, Mihalca and Miclea (2007) pointed out that the evolutions of learning
theories as well as technological changes have been forging trends in educational
research. Third, the Diffusion of Innovations Model (Rogers 2003) is a suitable
model to explain the relationship between e-learning research and the status of
technology development. For example, Hung (2010) concluded e-learning was at
the Early Majority stage in the Diffusion of Innovations Model, and the articles’ foci
had evolved from effectiveness comparison to developing models and teaching and
learning strategies in various e-learning environments.

Bibliometrics

Bibliometric analysis is a method to evaluate scientific research literature by


measuring certain indicators (Thelwall 2008). In particular, it summarizes publication
information with quantitative statistics. Typical bibliometric analysis yields the
following results: growth of papers by year and citations, rankings of most prolific
contributors, authorship patterns, rankings of geographical distribution of authors,
rankings of most productive institutions, collaboration among institutions, range and
percentage of references per paper, and frequency distribution of subject descriptors
(Keshaval and Gowda 2008). Traditionally, scholars use bibliometric analysis to
assess the impacts of scholarly publications and compare the contributions of authors,

123
4 J.-L. Hung, K. Zhang

institutions, and countries (Okubo 1997). Bibliometrics enable researchers to work on


large amounts of historical data and generate quantitative information.
In response to the criticism for the method’s emphasis on numbers, scholars (e.g.,
Mogil et al. 2009) attempted to remedy this disparity between quantitative and
qualitative approaches by conducting content analysis. In their recent study (Mogil
et al.), a single set of criteria were used to examine 4,525 research papers published
on the journal, Pain during the period of 1975–2007. They were then categorized by
four dimensions: research purpose, the subjects of the study, the types of pain being
studied, and the dependent measures. The categorical data were combined with
bibliometric data obtained from citation databases, and accordingly, the study
revealed the trends in pain research during the 32-year period.
Although content analysis may supplement traditional bibliometrics methodol-
ogies, it may be very time consuming and labor intensive. Thus researchers need a
practical research tool like text mining for speedy content analysis in addition to
categorization.

Text mining

Text mining, also known as text data mining or knowledge discovery from textual
databases (Feldman and Dagan 1995), refers to the process of extracting
meaningful, non-trivial patterns or knowledge from a set of unstructured texts. It
may be considered an extension of data mining or knowledge discovery from
databases (Fayyad et al. 1996). As Fayyad’s Knowledge Discovery in Database
(KDD) Process (Fayyad et al. 1996) suggest, text data mining typically involves five
steps: (1) data selection, (2) data cleaning, (3) data transformation, (4) data mining,
and (5) results evaluation and interpretation. The study reported herein followed the
KDD process to conduct text mining analysis of ML literature.

Method

Data collection

Given the increasing number of publications, especially with free, online publications
and open access resources, it is virtually impossible to have an exclusive search
without well-defined criteria. Considering the range of quality, this investigation was
carefully designed to focus on high quality publications collected in one of the most
widely used web-based databases, the Web of Science (SCI/SSCI). The Web of
Science (SCI/SSCI) was chosen as the source database for the following reasons:
First, the SCI/SSCI database collects journals and proceedings which are included in
both the Science Citation Index (SCI) and the Social Science Citation Index (SSCI).
The database is highly regarded including prestigious journals in science and social
science. Second, the SCI/SSCI is one of the few bibliometric databases (Okubo 1997),
which makes detailed bibliometric analysis possible.
Two terms, ‘‘mobile learning’’ and ‘‘m-learning’’ were selected as keywords for
initial search in this study, using the Web of Science (SCI/SSCI) search engine. The

123
TM in m-learning 5

search was limited to journal articles and proceedings papers, excluding other
resources, such as reviews and editorial materials. A total of 119 articles were retrieved
with the search period of January 1st, 1981-December 31st, 2008. All retrieved data
were preprocessed, or otherwise known as ‘‘cleaned’’, to remove irrelevant information
(e.g., page numbers, issue numbers, and publishers). The processed data were then
aggregated into fields and stored in a database. The following fields were created in the
database for the research purposes of this study: article number, article title, article
abstract, source journal, institution, country, and publication year.

Data transformation and analysis

Bibliometrics

Summary publication bibliometrics on ML were generated from the aggregate


records (Keshaval and Gowda 2008; Narin et al. 1994; Osareh 1996). A set of
bibliometric indictors were applied to describe the characteristics of ML research.
These indicators were the following: the quantitative growth of papers by year,
ranked list of geographical distribution of articles, ranked list of most prolific
institutions, and ranked list of most prolific source journals.

Text mining and clustering analysis

This study utilized abstracts as data sources for text mining for the following two
reasons. First, abstracts provide a comprehensive yet concise overview of an article.
Second, abstracts are much shorter than the full text, which minimizes the influence
of data noise. SAS Enterprise Miner 5.3 (SAS hereafter) was employed to conduct
text mining. More specifically, it was used to group documents based on the
similarity of abstracts. First, a list of stop words was generated by SAS to eliminate
obvious non-technical words (see ‘‘Appendix’’ for a sample list of the stop words).
SAS was used to search each abstract for text similarities of meaningful, content-
related words. As a result, the mining process yielded more accuracy and reduced-
noise in clusters. SAS also agglomerated clusters in a hierarchical tree structure,
which illustrated the taxonomy of ML based on the mining results. Ward’s method
for agglomerative clustering algorithms (Jain and Dubes 1988) was adopted to
generate the taxonomy structure. These clusters were interpreted by two experts in
ML to ensure inter-rater reliability (Gwet 2010).

Results

Publication time trends

Figure 1 summarizes the number of retrieved publications by year in the Web of


Science digital database from January 1, 1981-December 31st, 2008, tracing back to
the oldest time available in the database. The first eight ML articles were published
in 2003. In the following 2 years, publications on ML increased by 18.75%

123
6 J.-L. Hung, K. Zhang

Fig. 1 ML publication chronicle trends 2003–2008

annually, on average. From 2007 to 2008, the number of ML articles almost tripled,
escalating from 13 to 36. Consistent with the booming of mobile technologies,
research and publications on ML has exploded in the past few years. In later
sections, the factors contributing to the two publication explosion phases (i.e.,
2003–2005 and 2007–2008) are examined using text mining techniques.

Prolific countries

Based on author affiliations, text mining results identified the top countries
producing the most publications on ML in the given time frame (Fig. 2). The top
five prolific countries in ML research were Taiwan (27.73%), the USA (15.13%),
South Korea (9.27%), China (7.56%), and the United Kingdom (6.72%).

Prolific universities

In addition, based on author affiliations, the following universities were identified as


the most prolific in ML research and publication (see Fig. 3): National Central
University (Taiwan), Tamkang University (Taiwan), National Chiao Tung Univer-
sity (Taiwan), Athabasca University (Canada), and National University of Tainan
(Taiwan). Noticeably, an overwhelming number of publications on ML were
produced by authors at universities in Taiwan, as indexed and collected in Web of
Science.

Prolific journals

The journals publishing a large number of articles on ML are listed in Table 1.


Journals with the most ML articles were Lecture Notes in Computer Science,
Educational Technology & Society, Journal of Computer Assisted Learning,
Computers & Education, and International Journal of Engineering Education.

123
TM in m-learning 7

Fig. 2 Countries producing the most ML articles

Fig. 3 Institutions producing the most ML articles

These journals together have published 68.06% of the literature on ML in this


database.

Taxonomies

Figure 4 shows the results of the clustering analysis. This method classified articles
to form a hierarchical structure, based on text similarity of the abstracts. In the
retrieved data, only 115 articles included an abstract, and thus 115 abstracts out of
119 articles were analyzed. As a result, a three-level, twelve-cluster hierarchical

123
8 J.-L. Hung, K. Zhang

Table 1 Journals publishing the most ML articles


Source title % of Theme
articles

Lecture Notes in Computer Science 24.37 Computer science, theory and methods
Educational Technology and Society 13.45 Education and educational research
Journal of Computer Assisted Learning 10.92 Education and educational research
Computers and Education 10.08 Computer science in interdisciplinary
applications; education and educational
research
International Journal of Engineering 9.24 Education in scientific disciplines;
Education multidisciplinary

Fig. 4 Hierarchical taxonomy of ML

structure was constructed. These clusters were defined by two domain experts after
inspecting results in each computer-generated cluster.
As illustrated in Fig. 4, at the first level (leftmost column), the 115 articles were
divided into four categories. The first category (Strategies and Frameworks, 22
articles) discussed instructional strategies and frameworks of ML. It included three
clusters, or sub-categories: strategies or frameworks for ML (9 articles), collabo-
rative ML (7 articles), and interactivity of ML (6 articles).
The second category (Acceptance and Issues, 18 articles) discussed the
acceptance of ML and content-related issues. It included two clusters: acceptance

123
TM in m-learning 9

of ML (7 articles) and content protection, transmission, and management (11


articles).
The third category (Effectiveness, Evaluation, and Personalized Systems, 50
articles) was the largest in the taxonomy. This category contained articles on ML
evaluation and personalized ML systems. It included four clusters: effectiveness of
ML (13 articles), adaptive evaluation or intelligent tutoring systems on mobile
devices (15 articles), personalized ML systems (8 articles), and ML tool
development (14 articles).
The fourth category (ML Case Studies, 25 articles) included case studies, with
three clusters: ML applications in K-12 environments (7 articles), ML applications
in training (9 articles), and ML projects in engineering education, language learning,
and music education (9 articles).
As indicated in Fig. 4, the top three clusters were: ML tool development (14
articles, 12.2%), effectiveness of ML (13 articles, 11.3%), and content protection,
transmission, and management (11 articles, 9.6%).

Time trends of article clusters

Growing clusters

Figure 5 shows the results of the time trend analysis of the publications by cluster.
Trends indicated topics of growing or diminishing research interests. Figure 1
revealed two major growth phases of ML publications, 2003–2005 and 2007–2008.
Effectiveness of ML (cluster 6) contributed to the growth in both phases. Since ML

Fig. 5 Time trend of ML publication clusters. Note: CL1: Strategies or frameworks for ML; CL2:
Collaborative ML; CL3: Interactivity of ML; CL4: Acceptance of ML; CL5: Content protection,
transmission, and management; CL6: Effectiveness of ML; CL7: Adaptive evaluation or intelligent
tutoring systems on mobile devices; CL8: Personalized ML systems; CL9: ML tool development; CL10:
ML applications in K-12 environments; CL11: ML applications in training; CL12: ML projects in
engineering education, language learning, and music education

123
10 J.-L. Hung, K. Zhang

was a relatively new carrier of instruction in the first stage, scholars aimed to
understand the effectiveness of ML. Subsequently, the development of new mobile
technology kept the topic hot across both growth stages.
On the other hand, the development of new mobile technology also created new
possibilities for research. The following topics: acceptance of ML (cluster 4),
adaptive evaluation or intelligent tutoring systems on mobile devices (cluster 7),
ML tool development (cluster 9), and ML projects in engineering education,
language learning, and music education (cluster 12) contributed to the growth of
2007–2008.

Decreasing clusters

Only collaborative ML (cluster 2) decreased in 2007–2008.

Clusters without significant trends

Topics such as strategies or frameworks for ML (cluster 1), interactivity of ML


(cluster 3), acceptance of ML (cluster 4), content protection, transmission, and
management (cluster 5), personalized ML systems (cluster 8), ML in K-12
environments (cluster 10), and ML in training (cluster 11) consistently yielded
publications, with no significant fluctuations in the quantity of publications.

Publication clusters by country

Table 2 summarizes the publication clusters by country. The results reflected each
country’s topic clusters or topics in ML publications. In this section, dominance
refers to the country, whose publications exceeded all other countries in the same
cluster by a factor of two or more. For example, Taiwan dominated in publications
on topics in cluster 1, as 55.6% of cluster 1 articles were authored by scholars in
Taiwan, which was 2.5 times the percentage of cluster 1 articles by South Korean
authors (22.2%). Authors of the retrieved articles were from 26 countries; however,
18 countries produced only three or fewer articles each. Therefore, this section
focuses on countries that dominated in one or more clusters.
As indicated in Table 2, Taiwan dominated alone in 5 out of the 12 clusters:
strategies or frameworks for ML (cluster 1), acceptance of ML (cluster 4), adaptive
evaluation or intelligent tutoring systems on mobile devices (cluster 7), personalized
ML systems (cluster 8), and ML tool development (cluster 9). The USA was
dominant in collaborative ML (cluster 2). In addition, both the USA and Taiwan
were dominant in publications on interactivity of ML (cluster 3) and ML
applications in K-12 environments (cluster 10). Also, South Korea and Taiwan
shared dominance in content protection, transmission, and management (cluster 5).
There were no dominant countries in topics of effectiveness of ML (cluster 6), ML
applications in training (cluster 11), and ML projects in engineering education,
language learning, and music education (cluster 12).

123
TM in m-learning

Table 2 ML Publication clusters by country


CL1 (%) CL2 (%) CL3 (%) CL4 (%) CL5 (%) CL6 (%) CL7 (%) CL8 (%) CL9 (%) CL10 (%) CL11 (%) CL12 (%)

China 0.0 14.3 0.0 0.0 9.1 0.0 20.0 0.0 14.3 0.0 0.0 0.0
England 0.0 0.0 16.7 0.0 9.1 23.1 6.7 12.5 7.1 0.0 0.0 0.0
South Korea 22.2 0.0 0.0 0.0 27.3 15.4 0.0 12.5 14.3 0.0 0.0 0.0
Switzerland 0.0 0.0 0.0 0.0 0.0 7.7 0.0 0.0 7.1 0.0 0.0 0.0
Taiwan 55.6 14.3 33.3 42.9 27.3 7.7 40.0 50.0 28.6 28.6 22.2 0.0
Turkey 0.0 0.0 0.0 14.3 0.0 7.7 0.0 0.0 0.0 0.0 0.0 33.3
USA 0.0 28.6 33.3 14.3 0.0 23.1 6.7 0.0 7.1 28.6 0.0 0.0

Note: CL1, Strategies or frameworks for ML; CL2, Collaborative ML; CL3, Interactivity of ML; CL4, Acceptance of ML; CL5, Content protection, transmission, and
management; CL6, Effectiveness of ML; CL7, Adaptive evaluation or intelligent tutoring systems on mobile devices; CL8, Personalized ML systems; CL9, ML tool
development; CL10, ML applications in K-12 environments; CL11, ML applications in training; CL12, ML projects in engineering education, language learning, and
music education
11

123
12 J.-L. Hung, K. Zhang

Table 3 ML Publication categories by journal


Category 1: Category 2: Category 3: Category 4:
Strategies and Acceptance and Effectiveness, ML Case
frameworks (%) issues (%) evaluation, and studies (%)
personalized
systems (%)

Computers and Education 4.5 16.7 16.0 0.0


Educational Technology 18.2 11.1 12.0 16.0
and Society
International Journal 0.0 11.1 4.0 28.0
of Engineering Education
Journal of Computer 13.6 0.0 16.0 8.0
Assisted Learning
Lecture Notes in Computer 31.8 27.8 24.0 20.0
Science

Publication categories by journal

Focusing on the top five prolific journals, Table 3 summarizes the publication
categories. The results revealed that Lecture Notes in Computer Science and
Educational Technology & Society published articles in all four categories.
Computers & Education did not publish articles in ML case studies (category 4).
International Journal of Engineering Education did not publish articles on
strategies and frameworks. However, it published 28% of the articles on ML
projects in engineering education, language learning, and music education (category
4). Journal of Computer Assisted Learning did not publish articles regarding
acceptance and issues (category 2).

Discussion

New technologies and booming ML publications

Whenever a new learning technology is introduced, educational technology articles


tend to focus on its effectiveness (Bernard et al. 2004; Clark 1994; Evans 2008;
Johnson et al. 2000; Kulik and Kulik 1991). Likewise, ML researchers share
interests in ML effectiveness, as evidenced in the findings of this study.
Based on the Diffusion of Innovation Model (Rogers 2003), the diffusion process
can be divided into 5 stages: Innovators, Early Adopters, Early Majority, Late
Majority, and Laggards. Innovators, who represent about 2.5% of the group with
slow growth rate, adopt an innovation very early in the diffusion process. When the
process transitions to the Early Adopters’ stage, the growth rate becomes relatively
dramatic, increasing the percentage of adopters to 13.5% of the group. From Early
Adopter to Early Majority, the growth rate tapers, resulting in an s-shaped curve in
the overall percentage of adopters.

123
TM in m-learning 13

Based on Fig. 1, the period from 2003 to 2007 can be regarded as the Innovators
stage of ML studies and the Early Adopters stage begun after 2007.
Hung (2010) reviewed e-learning articles dated 2000–2008 and concluded that
e-learning research is at the early majority stage and foci have shifted from
comparing the effectiveness of e-learning to developing models for e-learning
environments and for teaching and learning strategies within various e-learning
environments. If ML articles follow a similar path, we may expect more research
studies on strategies and framework (clusters 1–3) in the near future.

Taiwan’s leading status in ML publications

This study revealed that the highest numbers of ML articles in SCI/SSCI journals
were produced by Taiwan-based researchers. Zhang and Hung (2006, 2009) found
that, with strong and aggressive e-learning initiatives, Taiwan’s government has
been providing financial, managerial and legislative support to promote e-learning
development in recent years. They found that Taiwan’s e-learning initiatives were
intended to stimulate the development of IT industries (Executive Yuan of the
Republic of China 2005). E-learning is valued and utilized as a driving force to
speed up the technical, industrial, and economical development of Taiwanese
society. Therefore, the majority of government funding goes to technology-related
research and development. Taiwan’s National E-Learning Project is divided into
seven tracks, of which one is Research and Development on ML Platform and Tools
(Executive Yuan of the Republic of China 2005). That explains why Taiwan’s
scholars have mainly invested their efforts in technology related research such as:
content protection, transmission, and management (cluster 5); adaptive evaluation
or intelligent tutoring systems in mobile devices (cluster 7); personalized ML
systems (cluster 8); and ML tool development (cluster 9).
On the other hand, USA scholars seemed to focus more on instructional aspects
of ML. Therefore, the USA was the most prolific country on topics like
collaborative ML (cluster 2), interactivity of ML (cluster 3), and ML in K-12
environments (cluster 10).
The results also revealed that China was prolific in ML research, yet with no
dominant topics or clusters in particular. It might be attributed to the lack of a strong
mobile technology infrastructure there.

Implications on research and practice

Findings of this study contribute to a deeper understanding of ML, by providing a


broad and longitudinal overview on prestigious publications. It provides a quick,
comprehensive overview for scholars interested in publications on ML. It has also
identified the topics and areas that have been studied more intensively regarding
ML. And more importantly, the findings suggest topics and areas needing additional
research. Based on Table 1, ML articles have appeared in journals of both computer
science and education. Therefore, researchers should pay more attention to
interdisciplinary approaches to research and development of ML in order to
synthesize knowledge from both disciplines.

123
14 J.-L. Hung, K. Zhang

As an emerging research method, text mining enables researchers to obtain


summative information in virtually any given field. This study illustrates the power
and potential of combining bibliometric and text mining techniques to discover
research patterns, themes, and trends. These techniques enable scholars to pay more
attention to data interpretation and pattern analysis, comparing to traditional
information processing or data (content) analysis.
For government decision makers, the findings will provide supportive informa-
tion to enhance understanding of research strengths and weaknesses, which in turn
can influence decision-making and policy change.

Limitations and suggestions for future research

SCI/SSCI journals adopt stringent criteria in reviewing articles. These articles in


general are regarded as having higher impacts in the field. However, they may not
be able to reflect the most recent research trends for two reasons. First, articles
might take 2 years or longer from submission to publication with these top-ranking
journals. Second, ML is a relatively new field and much research was published in
conference proceedings and other, newer journals. SCI/SSCI does not collect
proceedings papers in educational fields and only two proceedings journals in
computer science is included in the database. Therefore, the results and conclusions
of this study reflect the SCI/SSCI prestige as well as limitations, and were not
intended to be exclusive.
In an effort to control the noise of data analysis, this study selected abstracts as
the major data source, and thus the quality of abstracts could have affected the
results. Some key information (such as research methodology) might be absent in
the abstract and thus not reflected in the findings of this study. Future studies might
also use more key words in the initial search to retrieve additional related
publications.

Appendix

See Table 4.

Table 4 A sample list of stop words


a believe going moreover relatively thank
about besides gone my respectively thanks
according beyond got myself s thanx
accordingly by gotten n said that
actually c greetings name say thats
after came h namely saying the
afterwards certain he nd says their
against certainly hello nevertheless secondly theirs
ain changes hence nobody seem them

123
TM in m-learning 15

Table 4 continued
almost co her non seemed themselves
along com here none seeming then
already consequently hereafter noone seems thence
also corresponding hereby nor self there
although course herein nothing selves thereafter
am currently hereupon o sensible thereby
among d hers obviously sent therefore
amongst described herself of seriously therein
an despite hi oh shall theres
and doing him ok she thereupon
another done himself okay since these
any e his or so they
anybody each hither ought some think
anyhow edu hopefully our somebody this
anyone e.g how ours somehow thorough
anything either howbeit ourselves someone thoroughly
anyway et however p something those
anyways etc. i per sometime though
anywhere every i.e perhaps sometimes thru
apart everybody if please somewhat thus
appreciate everyone inasmuch plus somewhere thx
appropriate everything insofar possible soon to
aside everywhere instead presumably sorry together
ask example it probably specified too
asking f its provides specify took
associated for itself q specifying toward
at former j que still towards
available formerly just quite sub tried
away from k qv such tries
b furthermore l r sup truly

References

Attewell, J. (2005). Mobile technologies and learning: A technology update and m-learning project
summary. London, UK: Learning and Skills Development Agency. Retrieved from http://www.m-
learning.org/docs/The%20m-learning%20project%20-%20technology%20update%20and%20project%
20summary.pdf.
Attewell, J., & Savill-Smith, C. (2004). Learning with mobile devices: Research and development.
London, UK: Learning and Skills Development Agency.
Bernard, R., Abrami, P. L., Lou, Y., & Borokhovski, E. (2004). How does distance education compare
with classroom instruction? A meta-analysis of the empirical literature. Review of Educational
Research, 74, 379–439.

123
16 J.-L. Hung, K. Zhang

Clark, R. E. (1994). Media will never influence learning. Educational Technology Research and
Development, 42(2), 21–29.
Evans, C. (2008). The effectiveness of m-learning in the form of podcast revision lectures in higher
education. Computers & Education, 50(2), 491–498.
Executive Yuan of the Republic of China. (2005). National Science and Technology Program for
e-learning [website]. Retrieved from http://elnpweb.ncu.edu.tw/old/english/english1.htm.
Fayyad, U. M., Pitatesky-Shapiro, G., Smyth, P., & Uthurasamy, R. (1996). Advances in knowledge
discovery and data mining, AAAI/MIT Press.
Feldman, R., & Dagan, I. (1995). Knowledge discovery in textual databases (KDT). Proceedings of the
first international conference on knowledge discovery and data mining (KDD-95), 112–117.
Gall, M. D., Borg, W. R., & Gall, J. P. (1996). Educational research: An introduction (6th ed.). White
Plains, NY: Longman.
Geddes, S. (2004). Mobile learning in the 21st century: Benefit for learners. Knowledge Tree e-journal.
Retrieved from https://olt.qut.edu.au/udf/OLTCONFERENCEPAPERS/gen/static/papers/Cobcroft_
OLT2006_paper.pdf.
Keshaval, G. A. G., & Gowda, M. P. (2008). ACM transaction on information systems (1989–2006): A
bibliometric study. Information Studies, 14(4), 223–234.
Gwet, K. L. (2010). Handbook of inter-rater reliability: The definitive guide to measuring the extent of
agreement among raters. Gaithersburg, MD: Advanced Analytics, LLC.
Hung, J. L. (2010). Trends of e-learning research from 2000–2008: use of text mining and bibliometrics.
British Journal of Educational Technology. Retrieved from http://onlinelibrary.wiley.com/
journal/10.1111/%28ISSN%291467-8535/earlyview.
Jain, A. K., & Dubes, R. C. (1988). Algorithms for clustering data. Englewood Cliffs, NJ: Prentice-Hall.
Johnson, S. D., Aragon, S. R., & Shaik, N. (2000). Comparative analysis of learner satisfaction and
learning outcomes in online and face-to-face learning environments. Journal of Interactive Learning
Research, 11(1), 29–49.
Kim, P., Miranda, T., & Olaciregui, C. (2008). Pocket school: Exploring mobile technology as a
sustainable literacy education option for underserved indigenous children in Latin America.
International Journal of Educational Development, 28(4), 435–445.
Kulik, C. C., & Kulik, J. A. (1991). Effectiveness of computer-based instruction: An updated analysis.
Computers in Human Behavior, 7, 75–94.
Lehner, F., & Nösekabel, H. (2002). The role of mobile devices in e-learning: First experiences with a
wireless e-learning environment. Paper presented at IEEE International Workshop on Wireless and
Mobile Technologies in Education, Växjö, Sweden.
Lockwood, F. (2007). Forword. In G. Conole & M. Oliver (Eds.), Contemporary perspectives in
e-learning research: Themes, methods, and impacts on practice (pp. xvi–xvii). New York, NY:
Routledge.
Mihalca, L., & Miclea, M. (2007). Current trends in educational technology research. Cognition, Brain,
Behavior, 11(1), 115–129.
Mogil, S. J., Simmonds, K., & Simmonds, J. M. (2009). Pain research from 1975 to 2007: A categorical
and bibliometric meta-tend analysis of every research paper published in the journal, Pain. Pain,
142, 48–58.
Motlik, S. (2008). Mobile learning in developing nations. International Review of Research in Open and
Distance Learning, 9(2). Retrieved from http://www.irrodl.org/index.php/irrodl/article/view/
564/1071.
Narin, F., Olivastro, D., & Stevens, K. A. (1994). Bibliometrics: Theory, practice and problems.
Evaluation Review, 18(1), 65–76.
Okubo, Y. (1997). Bibliometric indicators and analysis of research systems: Methods and examples, STI
Working Papers 1997/1, OECD Science, Paris.
Osareh, F. (1996). Bibliometrics, citation analysis and co-citation analysis: A review of literature I. Libri,
(46), 149–158.
Polsson, K. (2009). Chronology of handheld computers. Retrieved from http://www.islandnet.com/*kpolsson/
handheld/.
Rogers, E. M. (2003). Diffusion of innovations (5th ed.). New York, NY: Free Press.
Sharma, S., & Kitchens, F. (2004). Web services architecture for m-learning. Electronic Journal of
e-Learning (2), 203–216.
Thelwall, M. (2008). Bibliometrics to webometrics. Journal of Information Science, 34(4), 605–621.

123
TM in m-learning 17

Winn, W. D. (2002). Current trends in educational technology research: The study of learning
environments. Educational Psychology Review, 1(3), 331–351.
Zhang, K. (2008). Ubiquitous technology for language learning: The U-Japan movement in higher
education. Journal of Computing in Higher Education, 20(2), 81–91.
Zhang, K., & Hung, J. L. (2006). E-learning in Taiwan: Policies, practices, and problems. International
Journal of Information and Communication Technology Education, 2(1), 37–52.
Zhang, K., & Hung, J. L. (2009). E-learning in supplemental educational systems in Taiwan: Present
status and future challenges. International Journal on E-Learning, 8(4), 49–64.

123

You might also like