Today's enterprises heavily rely upon accurate, consistent, and timely access to data. However, company data is typically scattered across multiple databases and file shares in a multitude of forms and versions. Moreover, an increasing amount of valuable background information is available outside the companies' influence and control. This situation is typical for many enterprise information integration scenarios, also in Austria's largest broadcasting media archive. Our demonstration argues for an information integration approach that uses semantic web principles to interlink archival media content of the Austrian Broadcasting Corporation (ORF) with the web of data and with internal knowledge resources to facilitate semantic search and to increase the user experience of browsing and discovering media content in the daily production workflow.
The Linked Open Data (LOD) community project was initiated in 2007 by the W3C
[
With this article we propose the integration of large datasets available on the web
by following the Linked Data principles as outlined in [
The Austrian Public Broadcaster’s (ORF – Österreichischer Rundfunk) arch ive is the central repository for all video and audio material created by the ORF in the last 60 years and contains a vast amount of media content in different formats. Th e primary objective of the archive is to preserve audio/video content for potential future use and make it accessible to editors. When archiving new content, several archiv ing tools restricted to expert users are used; FESAD4 as an example is used to man age video based content. However, for journalists, editors and program planners the archiving division uses a web based tool for federated search and investigation. For now, the work of describing the clips (e.g. annotating the content) is actually carried out solely by members of the archiving division. The users of the search tool currently cannot modify/annotate content in order to improve data quality or search confidence.
The main objective for the ORF is therefore to (i) provide additional information to the end users like editors and journalists, (ii) to allow simple annotation means which are not restricted to the archiving division and (iii) provide/integrate seman tic search facilities for improved search results. As an integrated solution we integrated the LMF as Linked Media Server in the Archival Toolset of the ORF. In addition to the existing tools, the LMF provides extended semantic search facilities and also allows for interlinking of archival content with publicly available linked data sources. As shown in Fig. 1, the LMF extends the search tool mARCo by adding itself as an additional data source and by providing means for annotating mARCO search results. The annotations are then subject of future searches in mARCo. 2.1
When browsing search results, editors or journalists are enabled to annotate the
content. With the help of a special annotation plugin, the formerly “read-only” search
result page becomes editable by injecting the annotation features into the w eb page.
Parts of the page such as the content description are analyzed by Apache Stanbol5. As
a result, eligible resource annotations are provided to the user as shown in Fig. 2.
By selecting a suggestion, the journalist can review the proposal and finally annotate
the content. The Linked Media Framework stores the annotation by means of
SPARQL Update [
The search experience can be improved by facilitating the semantic relatio ns of the archived data. By using the semantic concepts of the data which are either p roduction related (e. g. moderator, editor, program etc.) or content related (e.g. persons named or in video, content description, location of the clips content), it is possible to provide a faceted search as shown in Fig. 3, for example to narrow down the search, the user may select one or more facet properties shown in the search interface. 5 http://incubator.apache.org/stanbol 3
The Linked Media Framework (LMF) serves as the backend whereas the both clients
for search and annotation are lightweight JavaScript implementations using RESTful
webservices for the communication with the backend service. The LMF is a service
oriented framework which uses semi-structured data representation (RDF) and HTTP
URLs as uniform resource identifier to store and identify resources, as recommended
for Linked Data [
The LMF integrates/connects the linked data cloud as possible sources for background information and finally enables annotation by storing selected concepts in the (local) Linked Data server by means of SPARQL Update statements. In particular this annotation functionality will be subject of the demonstration given at I-Semantics to first show the where we will preload the LMF with a selection of news articles out of the Austrian Broadcasters Archive. The demonstration will also cover how the news articles are presented to journalists for annotation. Finally, the demonstration of the search interface is also available online at the NewMediaLabs demonstration site9. 4
The potential of Linked Data in general and the Linked Media Framework as a platform for supporting semantic search has been proven in several projects. With this demonstration we aimed to outline its potential for the use in an Enterprise Information Integration scenario where Linked Data technology is used to support users in their daily work and to improve the amount and quality of content annotation. The latter directly leads to an improved search result with respect to precision which is a fundamental requirement in the news domain. Because of the smooth integration in existing processes, the functionality is offered as an optional add-on to the users. The improved search results as well as the provided background information are the inducement for the users to use the offered functionality. In contrast to the increasing number of semantic web case studies10, the demonstrated scenario Linked Media Framework allows the publication of structured information as Linked Data and also 6 http://szabyg.github.com/vie-annotation-bookmarklet/ 7 http://dbpedia.org 8 http://www.w3.org/2004/02/skos/ 9 http://labs.newmedialab.at/ORF/orf/search/index.html 10 http://www.w3.org/2001/sw/sweo/public/UseCases/ enables the full read-write management of the published data and in particular enables the full roundtrip of annotations for further usage during search and retrieval. 5
The media content enhancement and the semantic search described in this paper were planned and developed in the Austrian research centre "Salzburg NewMediaLab - The Next Generation" (SNML-TNG). The centre is funded by the Austrian Federal Ministry of Economy, Family and Youth (BMWFJ), the Austrian Federal Ministry for Transport, Innovation and Technology (BMVIT) and the Province of Salzburg. The demo content is taken from the ORF archive by courtesy of the Austrian Broacasting Corporation. The development of the LMF has been inspired by the needs & requests of our industrial partners. As a result, the Linked Media Framework currently serves several real-world scenarios. 6