2. About this presentation
In this presentation I will summarise the business and
user value of supporting user created annotations in
Europeana. I will also describe different types of
annotations and suggest a relative priority from a
Europeana perspective.
This presentation is not a technical specification nor a
product development plan.
3. Table of Contents
• Why annotations?
• What are annotations?
• What is needed to support annotations?
• Risks!
4. Why annotations?
By encouraging and making it easy for our users to
annotate our content we will support a deeper and more
meaningful user experience, but we can also expect a
number of other positive outcomes:
•Meaningful connections across institutional silos
•Curated sets (galleries/collections) of objects
•Increased amounts of geodata
•Increased amounts of links to Things
•Enriched metadata to funnel back to our data partners
•Improved recall and precision in search
•Improved SEO
5. What are annotations?
• Media annotations
• Image annotations, but also text, audio and video annotations
• Geotags
• Object tags (usually just referred to as tagging)
• User created relations between objects
• User created object sets or collections
• Corrections
• Of original metadata
• Of automatic enrichments
• Translations of metadata
• Liking, and rating of objects
• Transcription of texts from images could be considered
as a form of annotation
6. Object tags
This will allow My Europeana users to tag items/records.
The system will aid the user in connecting their tags to
semantic resources and will index the multilingual labels
of those resources for search.
• It will be critical to retain a low threshold of participation (low
cognitive overhead) while allowing for user control
• Semantic resources will be autosuggested and the main input from the
user would be to disambiguate when needed
• Resources should come from SKOSified vocabularies like Dbpedia or
Freebase
•Benchmarks:
•JocondeLab
•Priority: 1
•In development in eCreative
7. Image annotations
This will allow My Europeana users to mark an area of
an image and add an annotation (comment) to that area.
Entities will be extracted from the annotation and indexed
for search.
• Creates annotations on the media items itself and thus encourages
deeper interaction with the representation of the work
• Dbpedia/Freebase to be used for the entities, possibly more specialized SKOS-compliant
vocabularies
• Dependent on direct media links in edm:IsShownBy
• Benchmarks:
• Annotorious
•Priority: 2
•In development in eCreative
8. User created collections/sets
This will allow My Europeana users to create small
collections of Europeana objects they select and then
name, describe, tag, geotag and publish them. The
collections will be made searchable and browsable in the
portal and via the API. They will be shareable,
embeddable and annotatable by other users.
•Creates amateur-curated snackable collections with personal
context
•Users should be able to collaborate in the making of
collections/sets
•Benchmarks:
• Rijksstudio, Digital NZ sets
• See Chenchen Shen's Channels report for some excellent thoughts on user
collections
•Priority: 3
•In prep for development in eCreative/eAwareness
9. User created object relations
This will allow My Europeana users to relate Europeana
objects to each other. A major outcome will be relations
created between objects from different data providers
and datasets.
• Creates meaningful connections across datasets and between
objects.
• Relationship types: To be decided but Describes/DescribedBy,
Depicts/DepictedBy, HasPart/IsPartOf, SameAs would be a good start
(these are the ones used by Kringla)
• Complements machine-created links between objects with
“communitysourced” relations
• The input of users could also be used to confirm or remove object
relations created by machine algorithms
• Benchmarks:
• Kringla where registered users can create typed relationships between
objects (and also typed relationships with Wikipedia articles)
• Priority: 4
10. Geotags
This will allow users to relate an item to a specific plaxce
on a map (or possibly multiple places). Such place tags
will be treated as resources with their own identifier and
will also be indexed for spatial search. A special type of
geotagging is georeferencing of maps.
• Creates more fine-grained geodata and improves existing geodata
• Thus improving relevance of developing geographical discovery
solutions based on our metadata
• Benchmarks:
• Historypin
• Priority: 5
11. Correction of automatic enrichments
This will allow users to correct automatically created
enrichments thus serving as a form of distributed quality
control.
•Improves the quality of semantic enrichment and thus Europeana
data overall
•And as an effect it improves discovery as precision of search is
improved by removing false hits
•Benchmarks:
•Powerhouse museum collections lets users remove machine tags
•Priority: 6
•High priority in eSounds
12. Correction of metadata
This will allow users to suggest corrections of original
metadata thus serving as a form of distributed quality
control.
• Improves the overall quality of Europeana data
• Benchmarks:
• ?
• Issues: We can't overwrite institutional data. How to feed it back to
the provider for them to officially change?
• Priority: 7
13. Translations
This will allow users to suggest translations of original
metadata thus serving as a form of distributed translation
service.
• Improves the overall quality of Europeana data and especially
improves multi-lingual retrieval
• Benchmarks:
• ?
• Priority: 8
14. Liking and rating
The most low-friction manner in which a user can
annotate an object. The resulting annotations can for
example be used to influence search result rankings or
add further sorting mechanisms.
• Liking is a good hint to ranking of search results
• Rating adds hints for ranking and a sorting mechanism for end-users
• Priority: 9
15. General requirements
• Develop API-first, features will not always be supported in the
Europeana portal
• Only logged-in My Europeana users or authorized API-keys can
annotate
• Annotations are published and made searchable in (near) real-time
• Community management of annotations, post-publication
• Users can flag annotations for removal with an annotation being removed
from display and search if flagged by 3 different users.
• Administrators can delete annotations and will have access to a
dashboard for tracking and managing annotations
• Annotations as a class must be added to EDM
• They must be time-stamped and linked to a user
• Annotations must be compliant with the Open Annotations
Community specification
• Annotations to be available for consumption via a continuous
Atom/RSS-feed
16. Specific requirement: Annotations harvest
Europeana should develop the capability to harvest and
ingest packages of annotations to existing Europeana
objects. This will be piloted in collaboration with
Historypin.
•Will allow collaboration with non-Europeana crowdsourcing platforms
like Historypin, Waisda?, Zooniverse and Crowdcrafting/Pybossa
•Priority: 1
•To be developed with Historypin as part of Europeana 3.0
17. Non-technical dependencies
• EDM must be updated to encompass user created
annotations and collections (in progress)
• The Europeana Terms of Use must be reviewed and if
necessary revised to take user created annotations into
account
• When the first annotation features are launched they
should do so as part of a prepared community outreach
campaign encouraging their use
18. System architecture
The Europeana platform must be extended to encompass
storage and (real-time) indexing of annotations and the
development of a read/write Annotations API
• A prototype Annotations service has been developed in Europeana
Creative
• Constraints:
•Must be open source
•Must be developed “API-first”
•Should ideally not add new major softwares to the European
stack
19. Risks!
• Lack of firm coordination between eCreative, eSounds
and eCloud leads to duplication of effort or development of
software modules that become abandonware post-project
• Overstretched data modelling, design and development
resources causes delay in/low quality of integration and
client implementations
• Endless “policy” discussions on the overestimated
reputational risks in user created content
• What will we do if someone annotates “Hitler’s the best. And
damned good looking as well!” Etc ad nauseaum.