Free your metadata: a practical approach towards metadata cleaning and vocabulary reconciliation

1. Tutorial content and its
relevance to the DH community
The early-to-mid 2000s economic downturn in the
US and Europe forced Digital Humanities projects
to adopt a more pragmatic stance towards metadata
creation and to deliver short-term results towards
grant providers. It is precisely in this context
that the concept of Linked and Open Data (LOD)
has gained momentum. In this tutorial, we want
to focus on metadata cleaning and reconciliation,
two elementary steps to bring cultural heritage
collections into the Linked Data cloud. After an initial
cleaning process, involving for example the detection
of duplicates and the unifying of encoding formats,
metadata are reconciled by mapping a domain
specific and/or local vocabulary to another (more
commonly used) vocabulary that is already a part of
the Semantic Web. We believe that the integration
of heterogeneous collections can be managed by
using subject vocabularies for cross linking between
collections, since major classifications and thesauri
(e.g. LCSH, DDC, RAMEAU, etc.) have been made
available following Linked Data Principles.
Re-using these established terms for indexing
cultural heritage resources represents a big potential
of Linked Data for Digital Humanities projects,
but there is a common belief that the application
of LOD publishing still requires expert knowledge
of Semantic Web technologies. This tutorial will
therefore demonstrate how Semantic Web novices
can start experimenting on their own with nonexpert software such as Google Refine. Participants
of the tutorial can bring an export (or a subset) of
metadata from their own projects or organizations.
All necessary operations to reconcile metadata with
Digital Humanities 2012
29
controlled vocabularies which are already a part
of the Linked Data cloud will be presented in
detail, after which participants will be given time
to perform these actions on their own metadata,
under assistance of the tutorial organizers. Previous
tutorials have mainly relied on the use of the Library
of Congres Subject Headings (LCSH), but for the
DH2012 conference we will test out beforehand
SPARQL endpoints of controlled vocabularies in
German (available for example on http://wiss-ki.
eu/authorities/gnd/ ), allowing local participants
to experiment with metadata in German.
This tutorial proposal is a part of the Free your
Metadata research project.1
The website offers a
variety of video’s, screencasts and documentation
on how to use Google Refine to clean and reconcile
metadata with controlled vocabularies already
connected to the Linked Data cloud. The website
also offers an overview of previous presentations.
Google Refine currently offers one of the best
possible solutions on the market to clean and
reconcile metadata. The open-source character of
the software makes it also an excellent choice for
training and educational purposes. Both researchers
and practitioners from the Digital Humanities
are within cultural heritage projects inevitably
confronted with issues of bad quality metadata
and the interconnecting with external metadata and
controlled vocabularies. This tutorial will therefore
provide both practical hands-on information and
an opportunity to reflect on the complex theme of
metadata quality.
2. Outline of the tutorial
During this half day tutorial, the organizers will
present each essential step of the metadata cleaning
and reconciliation process, before focusing on a
hands-on session during which each participant will
be asked to work on his or her own metadata set
(but default metadata sets will also be provided). The
overview of the different features will approximately
take 60 minutes:
- Introduction: Outline regarding the importance
of metadata quality and the concrete possibilities
offered by Linked Data for cultural heritage
collections
- Metadata cleaning: Insight into the features of
Google Refine and how to apply filters and facets
to tackle metadata quality issues.
- Metadata reconciliation: Use of the RDF extension
which can be in- stalled to extend Google Refine’s
reconciliation capabilities. Overview of SPARQL
endpoints with interesting vocabularies available
for Digital Humanists, in different languages.
After a break, the participants will have the
opportunity to work individually or in group on their
own metadata and to experiment with the different
operations showcased during the first half of the
tutorial. The tutorial organizers will guide and assist
the different groups during this process. Participants
will be given 60 minutes for their own experimenting
and during a 45 minutes wrap-up, participants
will be asked to share their the outcomes of the
experimentation process. This tutorial will also
explicitly try to bring together Digital Humanists
will similar interests in Linked Data and in this way
stimulate future collaborations between institutions
and projects.
3. Target audience
The target audience consists both of practitioners
and researchers from the Digital Humanities field
who focus on the management of cultural heritage
resources.
4. Special requests/equipment
needs
Participants should preferably bring their own laptop
and, if possible, have installed Google Refine.
Intermediate knowledge of metadata creation and
management is required.
Notes
1. See the projects website on http://freeyourmetadata.o
rg .