An Architecture to Enhance a Reference Management System with

Recommendations from Open Linked Data

María Hallo

and Sergio Luján-Mora

Department of Informatics and Computer Science, Escuela Politécnica Nacional, Quito, Ecuador

Department of Software and Computing Systems, University of Alicante, Alicante, Spain

Keywords: Content-based Recommender Systems, Open Linked Data, Reference Management Software.

Abstract: Reference management software helps students and researchers to store and to cite publications in different

citing formats. Common features of reference management software include advanced searching, reference

libraries and generation of citations. Some implementations help users to be connected to digital libraries to

get the article metadata of interest. Some researchers need to manage references of publications from open

access journals because of the elimination of barriers such as the price to get publications. However, the

number of publications grows each year and the researchers devote so much time to the retrieval, analysis and

management of bibliographic information. To solve this problem, in this work, we present a framework to

support the search, download and management of bibliographic information. A content-based recommender

module based on Open Linked Data is included into the framework. The metadata of the research publications

and the corresponding PDF files links are extracted using the recommender module and the Application

Program Interface from the Directory of Open Access Journals (DOAJ). The results are presented to the user

for the selection process. The metadata of the selected publications are stored in a local database integrated in

a bibliographic management system. A prototype was developed and was tested with information from open

access journals managed by the DOAJ.

1 INTRODUCTION

Reference management is a hard task for researchers.

At present, the researchers have access to more

information than they can consume and most of the

retrieved publications are not so relevant. The search

for related work is a hard consuming task for

researchers. The number of publications grows each

year and the open access initiative helps to have better

diffusion to some of those publications.

Some programs have been used to collect research

literature helping to organize communities for sharing

research literature (Ayers & Priedhorsky, 2011), such

as: Wikindx, Zotero, Mendeley, Reworks,

Referencer, JabRef, Kbibtex, EndNote, Readcube,

etc. (Kaur & Dhindsa, 2016). However, most of these

tools do not have an associated recommender system.

Recommender systems for research publications

are useful applications to help researchers to know the

state of the research in a specific topic. A good

recommender system is one that recommends the

most relevant items considering the user preferences

and goals (Beel, Langer & Genzmehr, 2013).

There are several types of recommender systems.

Content-based filtering has been used in several

recommender systems for reference management

such as: Dblib, Rec4LRW and TechLens (Torres et

al, 2004). Mendeley Suggest combine content based

recommendation with collaborative filtering (Jack et

al, 2016). The main problem in those solutions is

matching domain specific schemas. In collaborative

recommendations the main problem is the lack of

standards for data portability to integrate data from

different sources (Passant & Vojta, 2010).

Generally, recommender systems have been built

based on user preferences, interactions and resource

descriptions. A new approach of recommender

systems based on interlinked data has been reported

in several studies, enabling better integration of

resources between applications (Peska, 2013). The

common representation formats for Linked data,

typically is Resource Description Framework (RDF),

a language to represent machine-readable statements

in the form of triples <subject> <predicate> <object>.

The common semantics of this data is represented

using ontologies and related languages, i.e. RDFS

Hallo, M. and Luján-Mora, S.

An Architecture to Enhance a Reference Management System with Recommendations from Open Linked Data.

DOI: 10.5220/0006804104890496

In Proceedings of the 10th International Conference on Computer Supported Education (CSEDU 2018), pages 489-496

ISBN: 978-989-758-291-2

489

(W3C, 2014), OWL (W3C, 2004); and SPARQL

(W3C, 2008). Some research has been reported using

Linked Data on content-based recommender systems

(Di Noia, 2012). Linked Data is a set of best practices

to publish an interlinked data on the Web, and it is the

basis of an interconnected global data space where

data providers publish their content publicly

(Berners-Lee, 2006).

According to (Figueroa et al, 2015) the main

research contributions in content-based recommender

systems using Linked Data are resumed on: a) the

definition or extension of a similarity measures, b) the

definition or extension of an ontology, c) the

definition or extension of recommender algorithms,

d) the information enrichment.

In this paper we present a framework for a

reference management system including a content-

based filtering recommender approach based on Open

Linked Data for open access to scientific

publications. Using this proposed framework, a

prototype system has been developed. This system

has been tested with information from the Directory

of Open Access Journals (DOAJ) for the initial search

combined with information from ACM publication

metadata for the recommendations.

The rest of the paper is structured as follow: In

Section 2, we introduce related work. In Section 3, we

present the architecture of the proposed solution. In

Section 4, the results of a system implementation

using the proposed framework is described. Finally,

in Section 5, we show the conclusions and future

work.

2 RELATED WORK

Recommender systems are used to increase the user

satisfaction and precision of the retrieved

information. They help users to find their items of

interest (Ricci, 2011). There are several

classifications of recommender systems, one of them

(Lops,

De Gemmis & Semeraro, 2011) presents three

groups of recommender systems: Content-based,

collaborative filtering and knowledge-based

recommender systems.

Content-based recommender systems match up

the attributes of a user profile, constructed by user’s

past interest, with the information extracted from the

item in order to recommend to the user new

interesting items. Collaborative filtering (CF)

recommender systems makes recommendations to a

user based on items that other users liked in the past

(Goldberg et al, 1999). The main problem is to obtain

sufficient number of ratings to have a useful system.

Knowledge recommender systems are based on

specific domain knowledge about item features and

their corresponding matching with user preferences

(Ricci et al, 2011). Ontologies and Linked Data can

be used to improve the search based on domain

concepts and open data structure as Linked Data

(Hallo et al, 2016). There are also hybrid

recommender systems which use the combination of

the advantages of the previous cited techniques based

on the better features of each one. Additional methods

such as citation analysis (Vellino, 2015), taxonomic

topic expansion (Zarrinkalam & Kahani, 2012), query

expansion (Lüke et al, 2012) or term recommender

(Lüke et al, 2013) were also studied to enhance CF

algorithms.

There are reported recommender systems for

reference management enabling researchers to

control their research paper metadata, annotations and

PDF files. A Research Paper Recommender System

(RPRS) suggests research papers to the users

according to their personal preferences.

Most of the RPRS use content-based

recommender systems; few of them use collaborative

filtering or knowledge-based recommendations (Pohl

et al, 2007). Open Linked Data has been proposed to

improve the relevance of the retrieved items.

In Linked Open Data cloud there are millions of

RDF triples distributed in several datasets (Yang,

2010). This data, in RDF format, could be used to

enhance research paper recommender systems.

Following, we describe some of the studies related to

Linked Data-enable recommender systems, dataset

used and algorithms to produce recommendations.

In 2017, a SKOS Recommender Prototype which

produce scalable recommendations through

SPARQL-like query from Linked Data repositories

was proposed (Wenige & Rughland, 2017). This

system offers a combination of similar resource

retrieval and graph pattern matching. In other study.

Ammami et al (2016) developed a Latent Dirichlet

allocation (LDA)-based approach to scientific paper

recommendation. In this proposal the core idea is to

exploit the topics related to the authored articles to

formally define the author’s profile using topic

modelling and language modelling to represent the

recommender papers. The recommendation

technique is based on the closeness of the language

used in the research papers and the one used in the

recommender papers. This proposal alleviates the

cold start problem typical of collaborative filtering

techniques. In other study, Meymandpour et al (2013)

report a hybrid recommender system based on Linked

Open Data which could be also used for scientific

paper recommendation. In this study the

CSEDU 2018 - 10th International Conference on Computer Supported Education

490

recommendation method is based on the measure of

the semantic similarity of the resources combined

with user ratings. The features of each item are

extracted based on the relations with their neighbours.

Each link has a weight which could be assigned by

the experts. This system also work in absence of

ratings for items.

Beel et al. (2013) recommend to consider the

impact of the labelling in research paper

recommendations. In that work the nature of the

recommendation was analysed finding that the

institutional recommendations have better acceptance

than the sponsored recommendations. For all the

proposals it was necessary to have open RDF data sets

to improve the recommendations.

There are some bibliographic data sets available

on the Linked Open Data. RKBexplorer is a Semantic

Web application containing information from several

sources such as: DBLP, ACM, IEEE, Citeseer.

(Glaser et al, 2007). The datasets are available

through an SQL endpoints and resolvable

URIs. Semantic Web Dog Food is other large

structured dataset that focuses on the publications

from the Semantic Web community (Hu et al, 2015).

For the recommendations, Passant (2010) propo-

sed a semantic distance measure and collaborative

filtering approach based on Linked Data.

In relation to similarity measures, the studies

selected applied a variety of similarity measures

(Cheniki et al, 2016). These include pairwise cosine

function for vector similarity computation between

items, feature-based similarity to evaluate semantic

distance on different datasets, rating-based similarity

to compute the popularity of items among users.

Semantic similarity is also used based on diverse

relationships that can be found between concepts of

Linked Data datasets. Such relationships can be paths,

links or shared topics among a set of items. In

addition topic-based similarity has been used

capturing the relatedness between items based on the

categories they belong to (Figueroa et al., 2015).

Recommendations of a document x are based on the

most related to x in descending order (Hajra et al,

2014).

3 TECHNICAL ARCHITECTURE

In this section we describe the proposed technical

architecture to integrate the open source reference

management system Wikindex with a recommender

module based on Open Linked Data. Open Linked

Data from ACM metadata has been used for the test.

In the Figure 1, we present the technical architecture,

where each component and the information flow

between them are shown. Following, we describe the

components of the architecture: Integration Module,

Extraction Module, Reference Management Module,

and Recommender Module.

3.1 Integration Module

The integration module receives the query terms

(keywords, title, abstract or author) from the user and

it calls the extraction module to get the searched

publications. This module also presents the retrieved

publications to the user.

Figure 1: Technical Architecture.

An Architecture to Enhance a Reference Management System with Recommendations from Open Linked Data

491

3.2 Extraction Module

This module extracts the searched publications from

DOAJ database. The retrieved publications are

selected to be stored in a Wikindx local database. The

user could also add notes and additional files.

3.3 Recommender Module

An Open Linked Data content-based recommender

module is used to retrieve the article metadata that

better fits the search keywords from the DOAJ

selected publications. A scientific research paper

dataset in RDF is used to get the recommendations.

This module, adapted from the proposal of

Meymandpour (2013), has two components: semantic

information retrieval, and resource ranking.

3.3.1 Semantic Information Retrieval

In this step, facts and relations related to the selected

resources are extracted from several open linked data

sources after checking the quality of them. In our

proposal we consider for the recommendation the

elements: Resource (R), Category (C) and the

relations between them: R - R, R - C, C - C. (Figueroa

et al., 2105).

a) R-R are relations between resources based on

similar features. i.e. similar shared authors or

key-words. We considered that two articles are

similar in an RDF graph if they are the subjects

of two RDF triples having the same property and

the same object. (i.e. objects from the same

author).

b) R-C are relationships between a resource and a

category represented by the RDF property

rdf:type and the SKOS properties skos:subject ,

skos:isSubjectOf or dcterms:subject from the

Dublin Core vocabulary.

c) C-C are hierarchical relationships between

categories. They can be represented by using

RDFS property rdfs:subClassOf or the SKOS

property skos:narrower or skos:broader.

Thesaurus terms are linked with each other with

semantic relations such as “broader”, “narrower” or

“related”. The Thesaurus is an instrument to index

and retrieve subject-specific information.

Given an initial resource (or a set of initial ones)

a set of candidate resources linked to them located at

a predefined distance are generated. The resources are

retrieved based on the links R-R, R-C and C-C.

Following we present some SPARQL queries

used to find the resources for recommendation.

Find categories (?catRURI) of the resource

<RURI>

PREFIX dcterms:

<http://purl.org/dc/terms>

SELECT ?catRURI WHERE {

<RURI> dcterms:subject ?catRURI}.

• Find candidate resources (?cRURI) of the same

category <catRURI>:

PREFIX dcterms : <http :// purl.org

/dc/ terms />

SELECT DISTINCT ?cRURI WHERE {

?cRURI dcterms : subject <catRURI

>} .

•

Find candidate resources ?cr1URI linked to the

resource < RURI>.

SELECT ?cr1URI WHERE {

#output links

{<RURI> ?link ?cr1URI.}

UNION

#input links

{?cr1URI ?link <RURI>.}

}

3.3.2 Resources Ranking

For ranking the candidate resources, Passant propose

to calculate the Linked Data semantic distance

(LDSD) based on the number of direct input and

output links between two resources. In this proposal

the links came from a specific domain (Passant,

2010).

The SPARQL query that counts input and output

direct links between an initial resource (<inUrI>) and

a resource (<cRURI>) from the set of candidate

resources is:

SELECT DISTINCT count (?links) WHERE

{

# output links

{ <inURI > ?links <cRURI >. }

# input links

UNION

{ <cRURI > ?links <inURI >.}

The similarity of two resources (r1, r2) is

measured by:

LDSD (r1, r2) =1/(1+Cdout +Cdin) (1)

CSEDU 2018 - 10th International Conference on Computer Supported Education

492

Cdout is the number of direct output links (from r1 to

r2), Cdin is the number of direct input links. Using

these SPARQL queries, the ranking algorithm

calculates the LDSD for each pair of resources

composed of an initial resource and each of the

resources obtained from the previous step. The same

author test also another measure of similarity

considering direct and indirect links but the results are

not so much different.

3.4 Reference Management Module

Wikindx has been selected to be integrated with the

extraction and the recommender modules. Wikindx is

a free open source bibliographic management system.

Its main characteristics are (

Mucnjak, D. 2008):.

• The system allows editing or creating

bibliographic styles through a graphical

interface.

• The articles can be reformatted to another

citation style.

• The user can import and export other

bibliographic formats including BibTeX or

Endnote format.

• The users can export the bibliography in various

bibliographic styles (APA, Chicago, IEEE).

• The system can be used by one or multiple users

in a networked web server for a collaborative

work.

• The controller manages the commands for the

model or view.

4 RESULTS

A system prototype based on the proposed framework

is been developed.

The user interface is developed applying the

model, view, controller (MVC) software architecture

pattern. It divides the application in three

components:

• The model stores data to be presented in views.

• The view generates outputs to the user.

The sails.js framework and the Bootstrap library

are used for the interface development. Sails.js is a

framework used to easily build web applications.

Bootstrap is an open source toolkit containing HTML

and CCS design templates to develop front end

interfaces. For the metadata extraction, an API from

DOAJ was used to get the search metadata

publications. The metadata were retrieved in a JSON

file. Ajax and JQuery have been used to process the

JSON metadata displaying them on the webpage.

The recommender module is developed based on

information from Open Linked Data and tested with

an ACM scientific research paper dataset, in RDF

format, stored in RKBexplorer. The publication

metadata obtained for the retrieval publications from

DOAJ’s database are title, journal, abstract,

keywords, publisher, category and authors. The users

could record the metadata from the selected

publications in their own local system.

Following we present some SPARQL query tests to

get publications for recommendation:

• To find links to a publication.

• To find the authors of a publication.

• To find other publications of the same authors.

• To find other publications of the same area of

interest.

Figure 2: Prefix for SPARQL queries.

All the queries use the prefix shown in the Figure 2

and the selected publication identified by the URI

http://acm.rkbexplorer.com/id/100233.

• To find links to a publication.

In the Figure 3, we show the results of the indicated

SPARQL query execution to present some links to

the selected publication:

http://acm.rkbexplorer.com/id/100233.

Some properties are found such as rdf:type, akt:has-

title, akt:has-author, akt:addresses-generic-area-of-

interest.

SELECT DISTINCT ?links ?o WHERE

{<http://acm.rkbexplorer.com/id/1002

33> ?links ?o} Limit 20

PREFIX id:

<http://acm.rkbexplorer.com/id/>

PREFIX rdf:

<http://www.w3.org/1999/02/22-rdf-syntax-

ns#>

PREFIX rdfs:

<http://www.w3.org/2000/01/rdf-schema#>

PREFIX akt:

<http://www.aktors.org/ontology/portal#>

PREFIX owl:

An Architecture to Enhance a Reference Management System with Recommendations from Open Linked Data

493

Figure 3: Links to a publication.

• To find the authors of a publication.

In the Figure 4, we show the results of the execution

of the indicated SPARQL query to present the names

of the authors of the selected publication and the

results of the execution. The properties used in the

query are akt:has-author and akt:full-name. Different

names for the same author in the publications are also

presented.

SELECT DISTINCT ?name WHERE

<http://acm.rkbexplorer.com/id/10023

3> akt:has-author ?a.?a akt:full-

name ?name}

Figure 4: Authors of a publication.

• To find other publications of the same authors.

In the Figure 5, we show the results of the indicated

SPARQL query execution to present the titles of the

publications from the same authors of the selected

publication. The properties used in the query are

akt:has-author and akt:has-title.

SELECT ?p ?t WHERE

{<http://acm.rkbexplorer.com/id/1002

33> akt:has-author ?a. ?p akt:has-

author ?a. ?p akt:has-title ?t}

Figure 5: Publications from the same author.

• To find other publications of the same area of

interest.

In the Figure 6, we show the results of the execution

of the SPARQL query to display the titles of the

publications from the same area of interest of the

selected publication. The properties used in the query

are akt:addresses-generic-area-of-interest and

akt:has-title.

SELECT DISTINCT ?t WHERE

{?p akt:addresses-generic-area-of-

interest ?o.

<http://acm.rkbexplorer.com/id/10023

3> akt:addresses-generic-area-of-

interest ?o. ?p akt:has-title

?t}Limit 20

Figure 6: Publications of the same area of interest.

5 CONCLUSIONS AND FUTURE

WORK

We demonstrated how Linked Open Data can be used

to get research publications for recommendations.

The proposal architecture has been used to build a

new recommender systems that can operate for a

CSEDU 2018 - 10th International Conference on Computer Supported Education

494

particular domain data. We carried out a preliminary

evaluation using the data and categories from Open

ACM Linked Data resources. It is important to link

the resources of different datasets considering that

each one use different URIs for the resources i.e.

publications or authors. In the future we will make

relevance tests for the rankings and we will also study

the possibilities to scale this solution to manage big

quantities of information. Finally, we will work in:

analyzing a term based recommender module,

integrating other scientific databases and developing

a method to control de quality of the data. We need

also to define criteria for the final system evaluation

of the recommender system.

ACKNOWLEDGEMENTS

This publication comes from research conducted in

the project PII-16-06, with the financial support of

Escuela Politécnica Nacional from Quito, Ecuador.

REFERENCES

Amami, M., Pasi, G., Stella, F. and Faiz, R. 2016. An LDA-

Based Approach to Scientific Paper Recommendation.

In: Natural Language Processing and Information

Systems: 21st International Conference on Applications

of Natural Language to Information Systems, Springer

International Publishing, Switzeland, Cham. pp..200-

210.

Ayers, P. and Priedhorsky, R. 2011. WikiLit: collecting the

wiki and wikipedia literature. In:Proceedings of the 7th

international symposium on wikis and open

collaboration - WikiSym '11, 3/5 october 2011, Montain

View, USA. NewYork:ACM, pp.229-230.

Beel, J., Langer, S. and Genzmehr, M. 2013. Sponsored vs.

organic (research paper) recommendations and the

impact of labeling. In: Aalberg T., Papatheodorou C.,

Dobreva M., Tsakonas G., Farrugia C.J. (eds) Research

and Advanced Technology for Digital Libraries. TPDL

2013. Lecture Notes in Computer Science, vol 8092.

Springer, Berlin, Heidelberg pp.391-395.

Berners-Lee, T. 2006. Linked Data - Design Issues. [online]

W3C Available at: http://www.w3.org/DesignIssues/

LinkedData.html [Accessed 29 Nov. 2017].

Cheniki, N., Belkhir, A., Sam, Y. and Messai, N. 2016.

LODS: A linked open data based similarity measure. In

2016 IEEE 25th International Conference on Enabling

Technologies: Infrastructure for Collaborative

Enterprises (WETICE), IEEE, Paris, pp. 229-234.

Di Noia, T., Mirizzi, R., Ostuni, V., Romito, D. and Zanker,

M. 2012.. Linked open data to support content-based

recommender systems.In: Proceedings of the 8th

International Conference on Semantic Systems - I-

SEMANTICS '12., ACM, New York, USA, pp. 1-8.

Figueroa, C., Vagliano, I., Rocha, O. R., and Morisio, M.

2015. A systematic literature review of Linked Data

based recommender systems. Concurrency and

Computation: Practice and Experience. 27(17), pp.

4659-4684.

Glaser, H. and Millard, I. 2007. RKB explorer: application

and infrastructure. SWC'07 In: Jennifer Golbeck and

Peter Mika (eds.), Proceedings of the 2007

International Conference on Semantic Web Challenge,

CEUR-WS.org, Aachen, Germany, pp.97-104.

Goldberg, D. Nichols, D., Oki B.M. and Terry, D. 1992.

Using collaborative filtering to weave an information

tapestry. Commun. ACM. 35 (12), pp. 61–70.

Hajra, A., Latif, A. and Tochtermann, K. 2014. Retrieving

and ranking scientific publications from linked open

data repositories. In: Proceedings of the 14th

International Conference on Knowledge Technologies

and Data-driven Business - i-KNOW '14. ACM,

NewYork, USA.

Hallo, M., Luján-Mora, S., Maté, A and Trujillo, J. 2016.

Current state of Linked Data in digital libraries. Journal

of Information Science. 42(2), pp.117-127.

Hu, Y., Janowicz, K., Hitzler, P. and Sengupta, K. (2015).

The semantic web journal as linked data. In:

Proceedings of 2015 International Semantic Web

Conference (demos and posters track).

Bethlehem.USA.

Jack, K., Ingold, E. and Hristakeva, M. 2016. Mendeley

Suggest Architecture. A Practical Guide to Building

Recommender Systems. Available at: https://building

recommenders.wordpress.com/2016/10/10/mendeley-

suggest-architecture/ [Accessed 29 Nov. 2017].

Kaur, S. and Dhindsa, K. 2016. Comparative study of

citation and reference management tools: Mendeley,

Zotero and ReadCube.. In: International Conference on

ICT in Business Industry & Government (ICTBIG),

IEEE, pp.1-5.

Lops, P., De Gemmis, M. and Semeraro, G. 2011. Content-

based recommender systems: State of the art and trends.

In: Ricci., R. Okach, L. Shapira, B. Kantor, (eds).

Recommender systems handbook, F.., Springer, pp. 73-

105.

Lüke, T., Schaer, P. and Mayr, P. 2013. A framework for

specific term recommendation systems. In:

Proceedings of the 36th international ACM SIGIR

conference on Research and development in

information retrieval - SIGIR '13.ACM, New York,

USA, pp.1093-1094.

Lüke, T., Schaer, P. and Mayr, P. 2012. Improving retrieval

results with discipline-specific query expansion.In:

Zaphiris, P., Buchanan, G., Rasmussen, E. and

Loizides, F. (eds) Theory and Practice of Digital

Libraries, Springer Berlin Heidelberg, Berlin,

Heidelberg, pp.408-413.

Beel, J., Genzmehr, M. and Langer, S. 2013. Sponsored vs.

organic (research paper) recommendations and the

impact of labeling. In: Aalberg T., Papatheodorou C.,

Dobreva M., Tsakonas G., Farrugia C.J. (eds) Research

and Advanced Technology for Digital Libraries. TPDL

An Architecture to Enhance a Reference Management System with Recommendations from Open Linked Data

495

2013. Lecture Notes in Computer Science, vol 8092.

Springer, Berlin, Heidelberg pp.391-395.

Meymandpour, R. and Davis, J. 2013. Linked data

informativeness .In: Ishikawa, Y., Jianzhong, L.,Wang,

W. and Zhang, W.(eds). Web Technologies and Appli-

cations, Springer, Berlin, Heidelberg, pp. 629-637.

Mucnjak, D. 2008. Citation manager softwares freely

available to U of Zagreb students (EndNote, Zotero,

Biblioexpress, Wikindx)-analysis and comparison. In

BOBCATSSS Symposium: Providing Access to

Information for Everyone, Zadar, Hrvatska.

Passant, A. 2010. Measuring semantic distance on linking

data and using it for resources recommendations. In

AAAI spring symposium: linked data meets artificial

intelligence, AAAI Publications, p.123.

Peska, L. and Vojtas, P. 2013. Enhancing recommender

system with linked open data. In: Henrik Larsen, Maria

Martin-Bautista, María Vila, Troels Andreasen, and

Henning Christiansen (eds.). In Proceedings of the 10th

International Conference on Flexible Query Answering

Systems,Springer-Verlag, New York, USA. pp.483-

494.

Pohl, S., Radlinski, F.and Joachims, T. (2007).

Recommending Related Papers Based on Digital

Library Access Records. In Proceedings of the 7th

ACM/IEEE-CS Joint Conference on Digital Libraries,

ACM, New York, USA, pp. 417–418.

Ricci, F,. Rokach , L. and Shapira, B. 2011. Introduction to

Recommender Systems Handbook, In: F. Ricci., R.

Okach, L. Shapira, B. Kantor, (eds). Recommender

Systems Handbook, Springer US, Boston, USA, pp. 1-

35.

Torres, R. McNee, S. M. Abel, M. Konstan, J. A. and Riedl,

J. (2004) Enhancing Digital Libraries with Techlens+.

In: Proceedings of the 4th ACM/IEEE-CS Joint

Conference on Digital Libraries, Tucson, AZ, USA,

pp. 228-236.

Vellino, A. 2015. Recommending research articles using

citation data. Library Hi Tech. 33(4), pp. 597-609.

Wenige, L., and Ruhland, J. 2017. Retrieval by

recommendation: using LOD technologies to improve

digital library search. International Journal on Digital

Libraries.pp. 1-17.

W3C. 2014. RDF Schema 1.1. Available at:

http://www.w3.org/TR/rdf-schema/ (Accessed 29 Nov.

2017).

W3C. 2008. SPARQL Query Language for RDF.

Available at: https://www.w3.org/TR/rdf-sparql-query/

(Accessed 29 Nov. 2017).

W3C. 2004. OWL Web Ontology Language Overview.

Available at: https://www.w3.org/TR/owl-features/

(Accessed 29 Nov. 2017).

W3C. 2004. Resource Description Framework (RDF):

Concepts and Abstract Syntax. Available at:

https://www.w3.org/TR/rdf-concepts/ (Accessed 29

Nov. 2017).

Yang, S. Q. and Wagner, K. 2010. Evaluating and

comparing discovery tools: how close are we towards

next generation catalog?. Lib. Hi Tech. 28, pp.690–709.

Zarrinkalam, F. and Kahani, M. 2012. A multi-criteria

hybrid citation recommendation system based on linked

data. In: 2nd International eConference on Computer

and Knowledge Engineering (ICCKE), Mashad, pp.

283–288.

CSEDU 2018 - 10th International Conference on Computer Supported Education

496