<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Exploring Linked Classical Music Catalogs with OVERTURE</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Pasquale Lisena</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Manel Achichi</string-name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Eva Fernandez</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Konstantin Todorov</string-name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Raphael Troncy</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>EURECOM</institution>
          ,
          <addr-line>Sophia Antipolis</addr-line>
          ,
          <country country="FR">France</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>University of Montpellier</institution>
          ,
          <country country="FR">France</country>
        </aff>
      </contrib-group>
      <abstract>
        <p>In this paper, we introduce OVERTURE|a web application allowing to explore the interlinked catalogs of major music libraries including the French National Library, Radio France and the Philharmonie de Paris. We have rst developed the DOREMUS ontology which is an extension of the well-known FRBRoo model for describing works and expressions as well as the creation process. We have implemented a so-called marc2rdf tool allowing for the conversion and linking of bibliographical entries about music works, interpretations and expressions from their original MARC-format to RDF following this DOREMUS ontology. We present an exploratory search engine prototype that enables to browse through the reconciled collection of bibliographical records of classical music and to highlight the various interpretations of a work, its derivative, its performance casting as well as other rich metadata.</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>Music Metadata to Linked RDF Graphs</title>
      <p>
        The DOREMUS project1 develops methods to describe, publish, connect
and contextualize music catalogs on the web of data. Our focus is on the
description of classical and traditional musical works as well as their
interpretations (events) [
        <xref ref-type="bibr" rid="ref1">1</xref>
        ]. We rely therefore on the FRBRoo model, which combines the
strength of the FRBR practices in distinguishing a work from its interpretation
and derivatives and the CIDOC-CRM model that emphasizes the role of events
for modeling the process of creation of creative works. We extend FRBRoo by
adding classes and properties speci c to the music domain (e.g. for
representing the intended casting for interpreting a work). The research challenges we
are tackling are related to the population of this rich ontological model from
raw library metadata and on its visualization that should enable users to search
for classical music but also see in uencers and derivative works, or the many
interpretations of a famous work.
      </p>
      <p>Data Conversion. The data collected from the BnF and the Philharmonie de
Paris (PP) describing music works are represented in the MARC format (UNI{</p>
      <sec id="sec-1-1">
        <title>1 http://www.doremus.org/</title>
        <p>
          and INTERMARC variants). A MARC le is a succession of elds, each
carrying a 3-digit label, and sub elds, delimited by the $ symbol (Fig. 1). The
semantics of these elds and sub elds is described in documents issued by The
International Federation of Library Associations and Institutions (IFLA)2. Note
that a sub eld tag can change its meaning depending on the eld, under which
it is found. In our example (Fig. 1), the $a tag under the eld 500 stands for
the musical genre (sonata), whereas the same tag under the eld 700 stands
for the composer name (Beethoven). We have developed an open source
prototype that allows for the automatic conversion of UNI- or INTERMARC
bibliographic records to RDF3, implementing the DOREMUS model [
          <xref ref-type="bibr" rid="ref2">2</xref>
          ]. The
conversion process relies on explicit expert-de ned transfer rules (or mappings) that
indicate where in the MARC le to look for what kind of information,
providing the corresponding property path in the model as well as useful examples
that illustrate each transfer rule, as shown in Fig. 2. We have used the
DOREMUS properties to name the extracted relations (e.g., mus:U12 has genre4 is
the property describing the genre of a work). Resources are identi ed by URIs
that use the corresponding DOREMUS class labels in their names (as for
example http://data/doremus.org/expression/UUID identifying an instance of the
FRBRoo class Expression). Often, the RDF data that we extract will contain
string literals, such as genres (sonata) or instruments (piano). These values are
2 http://www.ifla.org/publications/ifla-series-on-bibliographic-control-36.
3 https://github.com/DOREMUS-ANR/marc2rdf
4 The mus pre x refers to http://data.doremus.org/ontology/
        </p>
        <p>
          Exploring Linked Classical Music Catalogs with OVERTURE
controlled by speci c vocabularies, such as MIMO5 describing musical
instruments. Therefore, our data conversion methods include an automatic mapping
of string literals to URIs coming from controlled vocabularies. Inspired by the
Datalift platform [
          <xref ref-type="bibr" rid="ref5">5</xref>
          ], we have developed a generic string2uri component
enabling to lookup strings in a SKOS controlled vocabulary and to propose best
candidate match.
        </p>
        <p>Among the several existing MARC to RDF conversion tools, such as the
Catamandu toolkit6 or marcmods2rdf7, the one that is closest in spirit to
ours is Marimba8, developed in the context of the Linked Data at the BNE
(Biblioteca Nacional de Espan~a) project. Just like our converter, and in contrast
to Catmandu and marcmods2rdf, Marimba is based on the IFLA ontologies,
including FRBR, and is designed by the help of librarians. However, Marimba
focuses on literature library entries, while we implement the DOREMUS model,
which is better suited to describe music.</p>
        <p>
          Data Heterogeneities and Linking. Due to the high data heterogeneity,
link discovery in the musical eld becomes a challenging task. To train and test
data linking tools, we have collected benchmark data from the BnF and the
PP, now part of the OAEI instance matching evaluation campagne9. Our tests
with SILK [
          <xref ref-type="bibr" rid="ref3">3</xref>
          ] and LIMES [
          <xref ref-type="bibr" rid="ref4">4</xref>
          ], achieving an average F-measure of around 55%,
con rm that new methods, speci c to the music eld, need to be proposed,
in order to handle speci c heterogeneity types of these datasets. In
particular, special attention has to be payed to multilingual descriptions of works, as
well as to signi cant lexical di erences in work titles. Descriptive heterogeneities
(level of detail, number of properties) also hinder the performance of general
purpose tools. Furthermore, certain properties, although mapped correctly
according to the transfer rules, contain highly heterogeneous information, as for
example cidoc-crm:P3 has note { a eld of free text in the form of comments.
We are currently developing a linking approach that combines expert heuristics
with indexing techniques that is able to propose a solution according to the
heterogeneity type manifested by the data.
2
        </p>
        <p>OVERTURE: an Exploratory Search Engine for Music
OVERTURE (Ontology-driVen Exploration and Recommendation of
mUsical REcords) is a modern web application that relies on a Node.JS server sitting
in front of a Virtuoso triple store. The web application performs SPARQL queries
to access the data via the Node.JS server that could be a Linked Data Fragment
server. The user interface is realized with the AngularJS2 library. At the top, a</p>
      </sec>
      <sec id="sec-1-2">
        <title>5 http://www.mimo-international.com/vocabulary.html</title>
        <p>6 http://librecat.org
7 http://simile.mit.edu/repository/RDFizers/marcmods2rdf/
8 http://marimba4lib.com
9 http://islab.di.unimi.it/content/im_oaei/2016/#doremus</p>
        <p>Lisena et al.
navigation bar directs the user towards one of the main concepts in the
DOREMUS model (Work, Expression, Performance, Recording, Score, Person), that
become distinct sections of the application. Inside each section, the user can
lter the list of records shown in a class-speci c and advanced search form. In this
way, (s)he can search a performance by choosing a particular location, a
performer, or an expression by selecting the musical key, the genre or by searching
by title. From these parameters, the server generates the SPARQL query that
matches the user information need.</p>
        <p>SELECT ?expressions ?title
WHERE {
?expressions a frbroo:F22_Self-Contained_Expression ;
cidoc:P102_has_title ?title .</p>
        <p>mus:U12_has_genre &lt;http://data.doremus.org/vocabulary/genre/si&gt;
}</p>
        <p>Clicking on a result, more detailed information is shown such as the
information about the composer, the casting, a longer descriptive note or di
erent expressions. In the future, this selection will contain related expressions,
selected through a recommendation algorithm. The application is available at
http://overture.doremus.org10.</p>
      </sec>
    </sec>
    <sec id="sec-2">
      <title>Acknowledgments</title>
      <p>This work has been partially supported by the French National Research
Agency (ANR) within the DOREMUS Project, under grant number
ANR-14CE24-0020.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          1.
          <string-name>
            <given-names>M.</given-names>
            <surname>Achichi</surname>
          </string-name>
          ,
          <string-name>
            <given-names>R.</given-names>
            <surname>Bailly</surname>
          </string-name>
          ,
          <string-name>
            <given-names>C.</given-names>
            <surname>Cecconi</surname>
          </string-name>
          ,
          <string-name>
            <given-names>M.</given-names>
            <surname>Destandau</surname>
          </string-name>
          ,
          <string-name>
            <given-names>K.</given-names>
            <surname>Todorov</surname>
          </string-name>
          , and
          <string-name>
            <given-names>R.</given-names>
            <surname>Troncy</surname>
          </string-name>
          . DOREMUS:
          <article-title>Doing Reusable Musical Data</article-title>
          .
          <source>In 14th International Semantic Web Conference (ISWC)</source>
          ,
          <year>2015</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          2. P. Cho e and
          <string-name>
            <given-names>F.</given-names>
            <surname>Leresche</surname>
          </string-name>
          . DOREMUS:
          <article-title>Connecting Sources, Enriching Catalogues and User Experience</article-title>
          .
          <source>In 24th IFLA World Library and Information Congress</source>
          ,
          <year>2016</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          3.
          <string-name>
            <given-names>A.</given-names>
            <surname>Jentzsch</surname>
          </string-name>
          ,
          <string-name>
            <given-names>R.</given-names>
            <surname>Isele</surname>
          </string-name>
          , and
          <string-name>
            <given-names>C.</given-names>
            <surname>Bizer</surname>
          </string-name>
          .
          <article-title>Silk: generating RDF links while publishing or consuming linked data</article-title>
          .
          <source>In 9th International Semantic Web Conference (ISWC)</source>
          ,
          <year>2010</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          4.
          <string-name>
            <given-names>A. N.</given-names>
            <surname>Ngomo</surname>
          </string-name>
          and
          <string-name>
            <given-names>S.</given-names>
            <surname>Auer</surname>
          </string-name>
          .
          <article-title>LIMES: A Time-E cient Approach for Large-Scale Link Discovery on the Web of Data</article-title>
          .
          <source>In 22nd International Joint Conference on Arti cial Intelligence (IJCAI)</source>
          , pages
          <fpage>2312</fpage>
          {
          <fpage>2317</fpage>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          5. F. Schar e, G. Atemezing,
          <string-name>
            <given-names>R.</given-names>
            <surname>Troncy</surname>
          </string-name>
          ,
          <string-name>
            <given-names>F.</given-names>
            <surname>Gandon</surname>
          </string-name>
          ,
          <string-name>
            <given-names>S.</given-names>
            <surname>Villata</surname>
          </string-name>
          ,
          <string-name>
            <given-names>B.</given-names>
            <surname>Bucher</surname>
          </string-name>
          ,
          <string-name>
            <given-names>F.</given-names>
            <surname>Hamdi</surname>
          </string-name>
          ,
          <string-name>
            <given-names>L.</given-names>
            <surname>Bihanic</surname>
          </string-name>
          ,
          <string-name>
            <given-names>G.</given-names>
            <surname>Kpklian</surname>
          </string-name>
          ,
          <string-name>
            <given-names>F.</given-names>
            <surname>Cotton</surname>
          </string-name>
          ,
          <string-name>
            <given-names>J.</given-names>
            <surname>Euzenat</surname>
          </string-name>
          ,
          <string-name>
            <given-names>Z.</given-names>
            <surname>Fan</surname>
          </string-name>
          , P.-Y. Vandenbussche, and
          <string-name>
            <given-names>B.</given-names>
            <surname>Vatant</surname>
          </string-name>
          .
          <article-title>Enabling linked-data publication with the Datalift Platform</article-title>
          .
          <source>In (AAAI) Workshop on Semantic Cities</source>
          , pages
          <volume>25</volume>
          {
          <fpage>30</fpage>
          ,
          <year>2012</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          10 A video is also available at https://www.dropbox.com/s/4e7n5a4qgyzk6wu/ OvertureDemo-v0.
          <source>mp4?dl=0</source>
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>