=Paper=
{{Paper
|id=None
|storemode=property
|title=Cite4Me: A Semantic Search and Retrieval Web Application for Scientific Publications
|pdfUrl=https://ceur-ws.org/Vol-1035/iswc2013_demo_7.pdf
|volume=Vol-1035
|dblpUrl=https://dblp.org/rec/conf/semweb/NunesFDC13
}}
==Cite4Me: A Semantic Search and Retrieval Web Application for Scientific Publications==
https://ceur-ws.org/Vol-1035/iswc2013_demo_7.pdf
Cite4Me: A Semantic Search and Retrieval Web
Application for Scientific Publications
Bernardo Pereira Nunes1,2 , Besnik Fetahu1 , Stefan Dietze1 , and Marco A. Casanova2
1
L3S Research Center, Leibniz University Hannover, Appelstr. 9a, 30167 Hannover, Germany
{nunes, fetahu, dietze}@L3S.de
2
Department of Informatics, Pontifical Catholic University of Rio de Janeiro,
Rio de Janeiro/RJ – Brazil, CEP 22451-900
{bnunes, casanova}@inf.puc-rio.br
Abstract. Cite4Me is a Web application that leverages Semantic Web technolo-
gies to provide a new perspective on search and retrieval of bibliographical data.
The Web application presented in this work focuses on: (i) semantic recommen-
dation of papers; (ii) novel semantic search & retrieval of papers; (iii) data inter-
linking of bibliographical data with related data sources from LOD; (iv) innova-
tive user interface design; and (v) sentiment analysis of extracted paper citations.
Finally, as this work also targets some educational aspects, our application pro-
vides an in-depth analysis of the data that guides a user on his research field.
1 Introduction
The huge amount of Web data and resources, particularly in the academic area, calls for
strategies to analyse and explore resources and data.
While scientific disciplines are very data- and knowledge-intensive, the lack of se-
mantic tools hampers information management and decision making. This includes sci-
entific data as well as unstructured academic publications as one of the key outcome
of scientific work. This is due to information access offered by digital library providers
such as ACM Digital Library1 and Elsevier2 being mostly based on free text search and
hierarchical classification3 .
Thus, we present a novel Web application for exploratory search, retrieval and visu-
alization of scientific publications. Cite4Me aims at providing a single access point for
accessing papers and, therefore, assisting searchers on finding relevant topics, papers,
and unveiling new nomenclature more efficiently. For this, we use reference datasets
such as DBpedia4 to explore semantic relationships between scientific papers and user
queries. We also perform a topic coverage analysis to provide an overview of different
bibliographic datasets. Cite4Me is a Web application which exploits results of previous
research works [2–5].
1
http://dl.acm.org
2
http://www.elsevier.com
3
http://www.acm.org/about/class/
4
http://dbpedia.org
2 Cite4Me - The Application
Cite4Me implements semantic and co-occurrence-based methods to search and retrieve
academic papers and suggest related work in a user-friendly interface that assists users
in exploring relationships between authors, institutions, papers and query terms. Due to
space restrictions, we present in this paper the most relevant features of Cite4Me to the
Semantic Web field.
2.1 Search and Retrieval
Cite4Me implements standard techniques, such as free text search, to search and re-
trieve scientific publications. In this section, we emphasize the semantic and exploratory
search mechanisms.
Exploratory Search. The exploratory search or graph search component assists users
to discover related work, people and institutions that are working on a specific topic.
A crucial step to provide this type of search is the annotation of the publications’ con-
tent. For this, we used DBpedia Spotlight API5 for extracting entities, entity types and
their categories. For instance, the categories of the extracted concepts are used to inter-
link publications through the topics they cover. In cases where two publications share
the same category (dcterms:subject property), then a link between both publications is
created. Figure 1 shows an example of topically related publications.
Fig. 1. Preview of the exploratory search funcionality.
Semantic Search. The semantic search component of Cite4Me is similar to the explicit
semantic analysis (ESA) technique [1]. After running the annotation process aforemen-
tioned, the relatedness score between the enriched concepts (DBpedia entities) found
5
http://dbpedia.org/spotlight
in the user query terms and the publications’ content are computed and ranked. The
relatedness score is computed based on the tf-idf score for the entities found in the pub-
lications’ content. The ranking of the retrieved documents is based on the sum of the
tf-idf scores of the matching concepts.
Figure 2 illustrates the semantic search functionality. Alongside the results of the
semantic search a tag cloud shows the most prominent terms for a given user query.
The tag cloud is updated while browsing through the list of results. The tags are selected
based on the tf-idf score for the entities found in the abstract of the retrieved papers.
Fig. 2. Preview of the semantic search funcionality.
Paper recommendation. Another important feature of Cite4Me and which differen-
tiates it from similar tools is the semantic paper recommendation. Given a scientific
publication, the tool recommends a related paper based on a score calculated according
to direct and lateral relationships between the publication of interest and the remaining
papers in our corpus.
To compute the relatedness score, we rely on previous work by Nunes et al. [2, 4],
where the paths connecting two enriched concepts in the scientific publications are anal-
ysed using a variation of the Katz index, a measure based on Social Network Theory,
and quantifying the weight of the connectivity between two concepts given a knowledge
graph (in our case DBpedia graph).
After computing the relatedness scored between enriched concepts, the paper rec-
ommendation relies on an aggregated measure that takes into account the relatedness
inter-documents. Finally, we generate a ranked list of pairwise publications according
to the overall score (see [5] for more details). Thus, the top-ranked publication is rec-
ommended to the user, as shown in Figure 3.
3 Datasets
Currently, Cite4Me is linked to a dataset (LAK Dataset6 ) which contains semi-
structured research publications from the ACM Digital Library (under a special license)
6
http://www.solaresearch.org/resources/lak-dataset/
Fig. 3. An example of paper recommendation based on S CS w .
and other public datasets (see also [6] for details). The dataset contains 315 full papers
along with their descriptive metadata while new publications are added continuously.
Metadata as well as the full text body are freely available in a variety of formats, includ-
ing RDF accessible via a public SPARQL endpoint. We are currently working on ex-
panding the number of papers available in Cite4Me. However, due to copyright reasons,
the process to expose scientific publications from publishers is still under discussion.
4 Conclusion
This paper presented the application of previous works in the Semantic Web field within
Cite4Me, a Web application that assists users in finding relevant scientific papers by ex-
ploring semantic relationships between them. For more information about the Cite4Me
Web application please refer to http://www.cite4me.com.
References
1. E. Gabrilovich and S. Markovitch. Computing semantic relatedness using wikipedia-based
explicit semantic analysis. In Proceedings of IJCAI’07, pages 1606–1611, San Francisco,
CA, USA, 2007.
2. B. Pereira Nunes, S. Dietze, M. A. Casanova, R. Kawase, B. Fetahu, and W. Nejdl. Combining
a co-occurrence-based and a semantic measure for entity linking. In ESWC, 2013 (to appear).
3. B. Pereira Nunes, B. Fetahu, and M. A. Casanova. Cite4me: Semantic retrieval and analysis of
scientific publications. In M. d’Aquin, S. Dietze, H. Drachsler, E. Herder, and D. Taibi, editors,
LAK (Data Challenge), volume 974 of CEUR Workshop Proceedings. CEUR-WS.org, 2013.
4. B. Pereira Nunes, R. Kawase, S. Dietze, D. Taibi, M. A. Casanova, and W. Nejdl. Can entities
be friends? In Proceedings of WOLE, in conjuction with the ISWC’12, volume 906 of CEUR-
WS.org, pages 45–57, Nov. 2012.
5. B. Pereira Nunes, R. Kawase, B. Fetahu, S. Dietze, M. A. Casanova, and D. Maynard. Inter-
linking documents based on semantic graphs. In Proceedings of KES’13, 2013 (to appear).
6. D. Taibi and S. Dietze. Fostering analytics on learning analytics research: the lak dataset. In
Proceedings of the LAK Data Challenge, held at LAK2013, April 2013.