Slides for the presentation given at the MTSR 2016 conference in Gottingen, Germany for the paper "Exploring Audiovisual Archives through Aligned Thesauri" by Victor de Boer, Matthias Priem, Michiel Hildebrand, Nico Verplancke, Arjen de Vries, and Johan Oomen.
In this paper, we present a case study where partial
collections of two audiovisual archives (Netherlands Institute for Sound and Vision and VIAA) are connected by aligning their thesauri. We report on the conversion of one of the thesauri to SKOS and on the subsequent application of an interactive alignment tool CultuurLINK. Finally, we introduce an cross-collection browser which uses the produced alignment to allow users to explore connections between the two collections.
Exploring Audiovisual Archives through Aligned Thesauri
1. Exploring Audiovisual Archives
through Aligned Thesauri
Victor de Boer, Matthias Priem, Michiel Hildebrand,
Nico Verplancke, Arjen de Vries and Johan Oomen
6. Case: Flemish Institute for Archiving (VIAA)
and the Netherlands Institute for Sound and Vision
(NISV)
• pipeline of a real-world, international use
case that illustrates the end-user benefit of
aligned SKOS thesauri
• method and tools for converting XML
thesauri to SKOS;
• CultuurLINK, an interactive tool for
thesaurus alignment;
• application that enables cross-collection
search and browsing using the aligned
thesauri.
7. Sound and Vision VIAA
Dutch AV heritage
> 1.000.000 hrs of Tv (public
broadcasters)
radio, music, docu, film, commercials, etc
Flemish archive,
including Flemish broadcaster (VRT)
8. Gemeenschappelijke Thesaurus
Audiovisuele Archieven (GTAA)
184,484 terms (concepts, persons, geo,…)
19,695 terms in hierarchy
9 conceptSchemes
90,708 scopeNotes
33,542 relations
Published as SKOS Linked Open Data
http://gtaa.beeldengeluid.nl/
11. VRT Thesaurus
102,172 terms
97,744 in hierarchy
4,429 top concepts
212 scopeNotes
6,828 relations
Conversion code available at https://github.com/viaacode/skoscreator
Triples available at http://semanticweb.cs.vu.nl/test
12. Collections
VIAA
• Part of the VRT AV collection
• +/- 35,000 items
(out of ~1Million)
• Annotated with VRT
thesaurus
• Not publicly available
NISV
• Openimages.eu
• +/- 3,000 items out of 800K hrs
• Mostly news broadcasts
• Annotated with GTAA
• Publicly available (CC-by-SA)
15. ‘Happy alignments are all alike;
every unhappy alignment is unhappy
in its own way’
Jacco van Ossenbruggen, (with apologies to Tolstoy)
16. http://cultuurlink.beeldengeluid.nl/
Semi-automatic SKOS vocabulary alignment service
Successor of EuropeanaConnect’s Amalgame
Users can upload vocabularies and match with
existing vocs.
Users can design, experiment, improve their
alignment strategy
Matching, selecting, excluding, sampling, evaluating
22. Demonstrator: Information Retrieval tool
using Spinque search-by-strategy paradigm
No programming needed,
just modelling the IR strategy
Keyword, vocabulary term or
Related-Object search
Search on titles, description,
vocabulary labels
Weight on collection (user-
input)
24. Input for keyword
search or
thesaurus
concepts
Search results
Collection
indicator
Thesaurus terms
associated with
video. Terms
may appear in
one thesaurus or
in both thesauriThesaurus terms
associated with
retrieval results
(grouped by type)
Slider used to
indicate
collection
preference/weigh
t
Per results, the
thumbnail, title,
description,
identifier and
thesaurus terms
are shown
25. The selected
video appears in
the search field.
Thesaurus terms
associated with
search results
and selection.
Play screen
In this case, the
user positioned
the slider all the
way to the right,
indicating that
he/she is
interested in
Open Images
videos related to
this VRT item.
List of
OpenImages
videos related to
this VRT video.
Matching terms
are highlighted.
26. Conclusions
Conversion of structured vocabularies to SKOS
opens possibilities for connecting collections
Interactive alignment produces many useful links
Demonstrator shows possibilities of aligned
collections
Demonstrator will be extended whenmore
collections are available
> Complete NISV collection metadata (?)
> Compete VIAA collection metadata(?)