Identifying Common Research Areas: A Study Case

Identifying Common Research Areas: A Study Case
V´ıctor Saquicela∗ , Jorge Bermeo∗ , Mauricio Espinoza∗ , Kenneth Palacio-Baus† , Boris Villaz´on-Terrazas‡
∗ Department
of Computer Science, University of Cuenca, Ecuador
{victor.saquicela, jorge.bermeo, mauricio.espinoza}
† Department of Electrical&Electronic Engineering and Telecommunications, University of Cuenca, Ecuador
[email protected]
‡ Intelligent Software Components, iSOCO, Madrid, Espa˜
[email protected]
Abstract—Currently, there is an increasing presence of researchers datasets (services) in the Internet. In this paper,
we present an approach for extracting publications made by
different authors and identifying common research areas among
them. This work makes use of semantic technologies in order
to describe authors and their publications through keywords
clustering techniques involving data mining algorithms.
Keywords—Clustering, Semantic Web, Research Areas, BIBO
The rapidly changing academic environment present in
Universities is characterized by a growing number of researchers or postgraduate students creating new publications
and constantly facing the diversification of their research topics. Identifying similar knowledge research areas has become
a prerequisite in promoting collaboration between researchers
interested in proposing new projects in a given field. In particular, for private companies and public organizations identifying
common interests among their researchers constitute a major
concern. One example of this is the case of the Ecuadorian
Government, which in recent years has promoted a massive
overseas education program aimed to improve the national
human talent. In the domain of research, several types of
datasets or services have been published and accessed through
APIs in order to make researchers information available.
Keeping this in mind, our aim is to create a system that
makes it possible to identify common research areas among a
list of authors as input. To achieve this goal, we have defined a
procedure that follows these steps: (1) the extraction of authors
by the automatic invocation of services using the OAI-PMH1
standard, (2) the extraction of publications of authors, (3) the
semantic description of authors and their publications with
respect to an ontology, and (4) the application of data mining
techniques (clustering) to detect common areas through the
keywords present in publications. We combine the ideas of
several initiatives, and propose a new system focused on the
identification of common research areas.
The remainder of this paper is structured as follows:
First, we present the background and related work done in
the domain of identifying common research areas. Then, we
describe an scenario that shows the problems currently found
in this context. Next, we present the architecture of the system.
Finally, we present some conclusions and identify future lines
of work.
This section provides a brief introduction to the DSpace2
repository and related tools used in searching for publications.
Moreover, the existing approaches related to the identification
of common research areas are described.
A. DSpace Repository of Dissertation Authors
DSpace is an OpenSource platform that allows the management and distribution of digital contents on the Web, using
a workflow based on publication requests and a series of
programmable filters. DSpace acts as a repository for digital
research and educational material produced by a particular
organization or institution [7]. From a technical point of view,
DSpace is implemented in Java and uses PostgreSQL as its
B. Searching for Publications
Currently, several tools found in the Internet allow people
to find scientific publications from specific authors. In general,
these search tools rely on the use of keywords, however, in
recent years many of these tools have started using semantic
technologies to describe the authors’ publications. As an
initial approach, this work proposes the use of tools based
on keywords searching only. Next, we describe the tools that
have been used in this work to find the publications about a
specific author.
There is a wide variety of information sources related to
academic articles. A complete list of these sources can be
found at A disadvantage commonly found among them is that they do not have an API that
allows access for information retrieval, and therefore, the need
of further interpretation of the obtained search results. Next,
three of the most popular academic platforms are analyzed:
Google Scholar3 does not have an API that allows
automatic publications searching, however, there is an
unofficial API that allows searching by title, author or
keyword in order to automatically extract the following fields: title, URL, number of citations, number of
versions, links to citations, and links to versions.
Microsoft Academic4 provides a REST API for publications searching. The results obtained from a query
are in JSON format containing the fields: title, abstract, keywords, authors, number of citations, year,
and URL.
IEEE Explore5 is a search service for publications.
This service has an API aimed to automatically perform searches based on different fields (author, title,
keywords). The obtained results include the fields:
title, abstract, keywords, authors, type of document,
year and URL.
C. Related Work
Karimzadehgan et al. [1] proposed an algorithm to solve
the problem of committee review assignment by modeling the
multi-aspect expertise matching as an integer linear programming problem which can can accommodate any probabilistic
or deterministic method.
“Rules of Selection and Award Programs and/or Projects of
Scientific Research and Development Funded or Co-financed
by the National Secretary of Higher Education, Science, Technology and Innovation (SENESCYT)”, have been proposed
to promote both: academic improvement and the creation of
new knowledge through research. These norms regulate how
the selection and adjudication process of programs and/or
projects of scientific research and technological development
are established, so that, public and private organizations can
access to the funds managed by SENESCYT [4]. This new
regulation system governs the Ecuadorian Higher Education
System and leads the transcendental changes experienced in
the country in at least the last 7 years. In this context, there
are two elements involved in obtaining the best results pointed
out by this challenge:
Atanassova et al. [3] present an Information Retrieval (IR)
system for scientific publications. It provides the possibility of
filtering results according to semantic facets. Semantic annotations are obtained using a rule-based method that identifies
specific linguistic clues organized in a linguistic ontology.
First, the responsibility of a researcher who is part of a
University in Ecuador and that is working within a specific
area, is to publish his/her results and findings. Unfortunately,
this task has not been fully performed yet, mainly because
traditionally it has not been considered by researchers as a
high priority issue and only started being supported and funded
in recent years. The problem is exacerbated by the lack of
knowledge among people and the lack of tools used for this
purpose. When a researcher identified in a given area of action
needs to know about the progress of his/her topic at the local,
national and international level, the common procedure evolves
around literature review. However, we believe that contacting
people who are involved in the same research areas of interest
could highly benefit society.
Osborne et. al [8] proposed the Rexplore system, aimed to
support the exploration and visualization of research trends.
We use a similar ideas for data sources managing and publications enriching, however, we will dynamically add new data
sources to improve author’s information.
Second, SENESCYT, as the government agency in charge
of research management, may require for instance, a list of the
researchers working on a specific area among all universities in
the country to start developing a new research project related
to the country’s needs.
After having analyzed the related work of approaches that
deal with identifying research topics, we can state that the
existing works do not automatically enrich the research topics
obtained by accessing third party research paper repositories,
such as GoogleScholar or the IEEE repository. Furthermore,
we propose the use data mining algorithms (clustering) differently from the aforementioned works.
In order to solve these problems, we propose a searching
process that retrieves potential research works in a specific
area, specifically, applied to the digital repository of the
University of Cuenca.
Dimou et al. [2] present RML, a generic mapping language,
based on R2RML6 , that provides a uniform way used to map
data present in any format to RDF7 . Authors made use of RML
to extract and map data of workshop proceedings published in
HTML to an RDF model, that represents the research topics
of the papers.
To depict an application scenario we introduce the case
of the Ecuadorian Government, which states through Article
350 of the Country Constitution that: “The higher education system has the purpose of academic and professional
training of scientific and humanistic vision, scientific and
technological research, innovation, promotion, development
and dissemination of knowledge and cultures, the construction
of solutions to the problems of the country, in relation to the
objectives of the arrangement of development” [5]. To follow
these purposes, Ecuadorian universities have been investing
important amounts of their resources and efforts in order to
improve their infrastructure and human capital. Particularly, in
the case of scientific and technical research. Initiatives like the
Fig. 1.
System Architecture
Figure 1 shows the process of automating the identification of research areas. Our system consists of five main
components: i) the authors extraction, which retrieves a list of
dissertation authors, ii) publication extraction, which retrieves
a list of publications belonging to the authors, iii) ontology
population, which stores instances, iv) similar research areas,
which makes use of data mining algorithms to detect similar
areas, and v) visualization, which shows the result. Next, we
briefly explain these components by illustrating the description
with some examples.
A. Authors Extraction
Authors data is normally located in the DSpace servers of
the organization that holds them. Authors having dissertations
have consequently registered publications in the institutional
repository. However, these registers might not be necessarily
up to date and could contain incomplete information. There
are different ways to access DSpace. Some of them are:
Database, access through connectors.
OAI-PMH, access through a specific protocol.
After an exhaustive analysis, we opted for the access
through the OAI-PMH protocol. Thus, the first step in our
approach is to take the URL of a OAI-PMH service as an input
and extract a list of authors from it. An example of OAI-PMH
invocation is:
This service retrieves information related to authors. More
specifically, it returns information about the following parameters: contributor, advisor, language, identifier, URI. The results
obtained after the OAI-PMH are shown in the following listing:
Invocation Results Listing
<element name=”dc”> <element name=”contributor”> <element name=”advisor”>
<element name=”es ES”>
<field name=”value”>
Saquicela Galarza, Victor Hugo
<element name=”author”>
<element name=”es ES”>
<field name=”value”>
Haro Valle, Valeria Alexandra
<field name=”value”>
P´erez Rocano, Wilson Rodrigo
<element name=”subject”>
<element name=”es ES”>
<field name=”value”>
<field name=”value”>
<field name=”value”>
<element name=”title”>
<element name=”es ES”>
<field name=”value”>
Data warehouse para el Centro de
Documentaci\’on Regional ”Juan Bautista
The following authors taken from the results, are two
representative examples of authors data, that will serve from
now on, as an illustration for our findings:
V´ıctor Saquicela. This author has a dissertation and
three dissertations as director.
Mauricio Espinoza. This author has a dissertation and
nine dissertations as director.
B. Publication Extraction
Once a relevant author has been discovered, we can extract
its main characteristics such as: publication, keywords, coauthors, etc. The list of authors is the one used to invoke
the different services (mentioned in section II-B) in order to
obtain another list containing the publications associated to
the authors (if this list is not available, our system cannot
continue without further human intervention). Then the system
analyzes the response to obtain a basic description structure
of the publications. This process is performed by the proposed
algorithm 1:
Algorithm 1 Publication Extraction Algorithm
Require: author
publications ← null;
publicationsScholar ← googleScholarSearch(author);
publicationsM icrosof t ← microsof tAcademicSearch(author);
publicationsIEEE ← IEEESearch(author);
for all publication ∈ publicationScholar do
if exist(publication) then
// Publications matching, attributes updating and aggregation
end if
end for
for all publication ∈ publicationM icrosof t do
if exist(publication) then
// Publications matching, attributes updating and aggregation
end if
end for
for all publication ∈ publicationIEEE do
if exist(publication) then
// Publications matching, attributes updating and aggregation
end if
end for
return publications
The services invocation may or may not return a value.
Here we show how we handle responses, which are represented
in a structured manner that can be easily consumed by different
technologies. The result of an invocation of our samples
authors are shown in Figure 2.
Fig. 2.
Result of invocation
The process consists of using the above detailed APIs to
obtain the publications related to the authors extracted in the
previous step. To do this, we have created a model of objects
that represents a scientific article to their respective authors,
keywords and data source. Each data source is processed
according to the heterogeneity of the data they contain.
C. Ontology Population
The result of the list of publications with their corresponding authors, is used to populate an ontology. We store this
result into a triple store using as reference the BIBO8 ontology.
We selected this storage tool in order to increase the discovery
relations between authors, publication and sources. Moreover,
the results established between different authors through both:
keywords and the ontologies, are registered and stored in the
repository, so that they can be used later. We use BIBO to
describe publications and FOAF9 to describe the authors. For
URIs we have defined
and For conversion, we use
the library JENA and the following properties of the ontology
will be considered to map the results. Figure 3 depicts how
the model presented in this approach is mapped to the BIBO
ontology in order to represent author data. The result of this
relation allows the generation and publication of dissertation
data as linked data. Furthermore, it can be noted that we rely
on triple store, which allows access via SPARQL.
instances and clusters. Data, once the algorithm is applied,
will tend to cluster around certain keywords groups, allowing
the user to quickly determine patterns in the data. The results
of applying clustering algorithms associated with keywords
publications, show that the authors found during the process
described in section IV-A have indeed common research areas.
In this paper, we have presented an approach for identifying
common research areas. The goal of this study case is to
analyze existing technologies used to search for publications,
ontologies employed to represent publications, and data mining
algorithms aimed to discover patterns based in keywords.
Additionally, this approach includes the use of different tools
that allow publications search. We describe the process we
followed to demonstrate the potential of this proposal through
an example.
Future work will focus on the addition of new search tools
that could improve the obtained results. Also, we want to carry
out the evaluation of the clustering results obtained at this stage
of development. Furthermore, we pretend to show the results
of linked data and clustering execution in a visual way. Finally,
we plan to create a platform able to integrate data from other
universities. Thus, we aim to discover similar areas of research
between universities.
This work has been supported by the project ” Plataforma
de integraci´on, publicaci´on y consulta integrada de recursos
bibliogr´aficos en la Web Sem´antica”, funded by CEDIA11 .
Fig. 3.
Mapping between BIBO ontology and our model
D. Discovering Similar Areas
To automatically discover similarities, we describe how
clustering algorithms can be used to discover similar research
areas. For clustering algorithm execution, our system uses the
WEKA10 library. WEKA is a collection of machine learning
algorithms used in data mining tasks. It contains tools that can
be employed in: data pre-processing, classification, regression,
clustering, association rules and visualization.
In the discovering process, the keywords of each author’s
publications are extracted, forming a kind of document containing just keywords associated to authors. Before running
the algorithm, we pre-process, normalize and transform the
data using different WEKA filters. In particular, clustering
algorithms for documents. Clustering is the task of uncovering
unanticipated trends by segmenting no predefined clusters.
This approach is used in situations where a training set of
pre-classified records is unavailable [6].
In this matter, we want to cluster keywords related to
a similar area by looking at word weights. We use WEKA
Simple-KMeans clustering. This algorithm is based on the
Euclidean distance measurement to compute distances between
[1] Karimzadehgan Maryam and Zhai ChengXiang, “Constrained multiaspect expertise matching for committee review assignment”, In Proceedings of the 18th ACM conference on Information and knowledge
management (CIKM ’09), ACM, New York, NY, USA, pp. 1697-1700,
[2] Dimou A., Vander Sande M., Colpaert P., De Vocht L., Verborgh R.,
Mannen E and Van de Walle R., “Extraction and Semantic Annotation
of Workshop Proceedings in HTML using RML”, In Proceedings of
The Semantic Publishing Challenge of the 11th Extended Semantic Web
Conference, 2014.
[3] Atanassova Iana and Bertin Marc, “Faceted Semantic Search for Scientific Publications”, 11th ESWC 2014 (ESWC2014), 2014.
[4] SENESCYT, “Reglamento de Selecci´on y Adjudicaci´on de Programas
y/o Proyectos de Investigaci´on Cient´ıfica y Desarrollo Tecnol´ogico (I+D)
Financiados o Cofinanciados por la Secretar´ıa Nacional de Educaci´on
Superior, Ciencia, Tecnolog´ıa e Innovaci´on”, Acuerdo 2012-009, 2012.
[5] Asamblea Constituyente, “Constituci´on de la Rep´ublica del Ecuador”,
[6] Chen, Sherry Y and Liu Xiaohui, “The Contribution of Data Mining to
Information Science”, Journal of Information Science, no. 6, vol. 30, pp.
550-558, 2004.
[7] Tansley, Robert and Bass, Mick and Stuve, David and Branschofsky, Margret and Chudnov, Daniel and McClellan, Greg and Smith,
MacKenzie, “The DSpace institutional digital repository system: current
functionality”, Proceedings of the 3rd ACM/IEEE-CS joint conference
on Digital libraries, IEEE Computer Society, pp. 87-97, 2003.
[8] Osborne, Francesco and Motta, Enrico, “Exploring Research Trends
with Rexplore”, D-Lib Magazine, Corporation for National Research
Initiatives, vol. 19, no. 9, 2013.