When Semantics support Multilingual Access to Digital Cultural Heritage - the Europeana case

Presenters: Valentine Charles / Juliane Stiller
(Europeana Foundation, Netherlands, The / Humboldt-Universität zu Berlin, Germany)

Abstract:
For Europeana, the platform for Europe’s digital cultural heritage from libraries, museums and archives, multilingual access is one priority. Breaking down language barriers is an ongoing challenge with many facets as Europeana provides content coming from 36 different countries serving users across the world. For Europeana, multilingual access does not only mean the translation of the interface, but also comprises retrieving, browsing and understanding documents in languages the users do not speak. This talk will present the solutions implemented at Europeana enabling multilingual retrieval and browsing. Europeana leverages the semantic data layer by linking multilingual and open controlled vocabularies to objects. The Europeana Data Model (EDM) allows for semantic and multilingual metadata descriptions and gives support for contextual resources including concepts from “value vocabularies” either coming from Europeana’s network of providers or third-party data sources. To enable retrieval across languages and enhance data semantically, Europeana performs automatic metadata enrichment with external value vocabularies and datasets such as GEMET, GeoNames and DBpedia. Providers are also encouraged to send links from open vocabularies such as AAT, GND, Iconclass and VIAF or from their domain vocabularies following the EDM recommendations for contextual resources, especially when these vocabularies contain labels in different languages. By re-using these vocabularies, Europeana does not only pursue efforts in demonstrating the potential of Linked Open vocabularies by exploiting the semantic relations and translations but also aims at making Europeana truly multilingual.

Transcript

1. When Semantics support Multilingual Access to Cultural Heritage The Europeana Case Valentine Charles and Juliane Stiller SWIB 2014, Bonn, 2.12.2014
2. Our outline 1. Europeana 2. Multilinguality in digital libraries - challenges 3. Europeana Data Model – a framework for multilingual data 4. Semantic and multilingual enrichment
3. Europeana, the platform for Europe’s digital cultural heritage
4. Europeana Aggregates metadata from the cultural heritage sector in Europe • Libraries, museums, archives and audio-visual archives • Metadata in 33 languages  Provides a portal for users to access data and objects • http://www.europeana.eu/ in 31 languages • Metadata under Creative Commons Zero - public domain • Previews and links to source Data distributed via • API http://labs.europeana.eu/api/ • Linked Data (currently being updated) http://data.europeana.eu/
5. 5 33M objects from 2,200 galleries, museums, archives and libraries CC Europeana.eu, Europe’s cultural heritage portal
6. Challenges Multilinguality issues • Provide access to multilingual resources • Allow the search for items in various languages • Make sure users can understand the descriptions of these items
7. Multilinguality in digital libraries - challenges
8. Dimensions of multilinguality Interface and portal display Search • Translation of query • Translation of documents Representation and refinement of search results • User needs to be able to determine relevance of documents Browsing
9. Portal display • Which language will be displayed to the (first) user? • Will a cookie be set? • What will be translated? • Which language dimensions does the drop-down menu impact?
10. Cross-lingual search Mona Lisa AND La Joconde Search External Dataset External Dataset
11. Cross-lingual search Determine source language Determine target language Pick translation Translation of result list Translation of object • Queries are short • 39% of queries can belong to more than one language • 60% of queries are named entities
12. Europeana Data Model- a framework for multilingual data
13. Create new data framework Europeana Data Model (EDM) • Re-uses several existing Semantic Web-based models: Dublin Core, OAI-ORE, SKOS, CIDOC-CRM…  More granular metadata • Links e.g. between objects and context entities (persons, places) • Multilingual & semantic linked data for contextual resources (e.g. Concepts)
14. Rely on knowledge organisation systems  Create a “semantic layer” on top of cultural heritage objects • Include multilingual “value vocabularies” • From Europeana’s providers or from third-party data sources
15. Encourage providers to contribute their own vocabularies Benefit from data links made at data providers’ level Ingestion of vocabularies is made possible if the vocabularies used the data structures EDM expects • For instance SKOS for concepts
16. An example the integration of AAT URIs in EDM hourglasses@en uurglazen@nl reloj de las horas@es http://vocab.getty.edu/aat/300206197 edm:ProvidedCHO Hourglass urn:imss:instrument:401058 skos:Concept http://vocab.getty.edu/ aat/300198626 skos:prefLabel skos:prefLabel skos:prefLabel skos:broader dc:type
19. Automatic enrichments
20. Object Object Enrichments in information retrieval Mona Lisa AND La Joconde Search Goal: reaching higher visibility of documents within the document space
21. Enrichments in the linked data space Goal: contextualization which goes beyond the scope of a particular platform Object Object External Dataset and Vocabulary External Dataset and Vocabulary
22. Automatic enrichment process in Europeana
23. Enrichment types and vocabularies Enrichment Type Target vocabulary Source metadata fields Number of enriched objects Places GeoNames dcterms:spatial, dc:coverage 7 mio Concepts GEMET, DBpedia, dc:subject, dc:type 9,2 mio Agents DBpedia dc:creator, dc:contributor 144,000 Time Semium Time dc:date, dc:coverage, dcterms:temporal, edm:year 10,2 mio
24. Europeana Enrichment - Example
25. Quality of enrichments Olensky et al. (2012) analyzed 200 enrichments of Europeana -> found enrichment flaws and problems Incorrect enrichments lead to • Devaluation of curated metadata • Loss of trust from providers • Propagation of errors to different languages • Irrelevant search results • Bad user experiences Better understanding of impact of enrichments needed 
26. To conclude Continue tofocus on cross-domain multilingual vocabulary alignment and publish the results as Linked Data • More pivot vocabularies such as AGROVOC, STW Thesaurus for Economics integrated in Europeana More domain-specific and targeted vocabularies for enrichment Multilingual interactions Better understanding of impact of multilingual strategies on Search and Browse and User Interactions
27. Thank you Valentine Charles & Juliane Stiller valentine.charles@europeana.eu, juliane.stiller@ibi.hu-berlin.de
28. Toolbox Replace text and adjust size Replace text and adjust size Replace text and adjust size Replace text and adjust size Replace text and adjust size Replace text and adjust size