Old Maps, Annotations, and Open
Data Networks
Bernhard Haslhofer (Cornell University)
...with input from Rainer Simon (Austrian Institute of Technology)
Harvard University, January 2013
My plan for today...
• Background and Context
• Annotations on Historical Maps
• Maphub Demo + Technical Details
• yuma.min.js / Annotorious
• Next Steps, Ideas, ...
• Related Projects
Harvard, January 2013 2
About me...
• Postdoc at Cornell
• Research in the area of global data networks
• how can open data support scholarship?
• how can scholarly knowledge flow back into
open data networks?
Harvard, January 2013 3
The big picture...
W3C Open Annotations
maphub
Semantic Tagging (Zoomable) Historic Maps
JavaScript, Ruby on Rails
LEMO YUMA yuma.min.js / Annotorious
Linked Data Annotation Model OAC Model Audio, (Zoomable) Images
Audio,Video, Images Audio,Video, Images JavaScript
Server + Flash Clients Server + GWT Clients
yuma4j
Storage Server (Java)
Open Annotation OAC Model Open Knowledge
Collaboration (OAC) Foundation Annotator
2007 2013
Harvard, January 2013 4
Possible Benefits
• Engage people; let them tell their stories
• Collect knowledge; enhance metadata
records
• Connect digitized maps with other Web
resources
Harvard, January 2013 9
Maphub Demo + Technical Details
Harvard, January 2013 10
Geo References
Affine Transformation
Model
y x = a*lat + b*lng + c
y = d*lat + e*lng + f
lat = a*x + b*y + c
lng = d*x + e*y + f
x
Harvard, January 2013 14
Map Overlays
• Generate overlay tilesets using GDAL and
gdal2tiles.py
/raw
/thumbnails
/ts_zoomify
/ts_google
/scripts/convert.rb
new control points?
...
Harvard, January 2013 15
Semantic Tagging
annotation text
extracted named entities
Wikipedia (Wikipedia Links)
Miner
Harvard, January 2013 20
Semantic Tagging
Label-based Tagging Model Semantic Tagging Model
Resources Tags Users Resources Semantic Users
Tags
t1, t2, t3
t1, t4, t5
t1
t6, t7
t1, t6, t7
Knowledge Context
based on Marlow et al. (2006)
Harvard, January 2013 21
Multilingual Search
(a
bs
fe ct +
tr
tc l
a
h ab
en el
ric tr
hm an
en slat
ts ion
s)
index map metadata + annotations + enrichments
search & retrieval
Harvard, January 2013 22
Maphub API
• Central Maphub information entities:
maps, annotations (comments), control points
• Each instance of these entities becomes a first-
class Web resource, identified by a URI
• http://maphub.herokuapp.com/maps/355
• http://maphub.herokuapp.com/annotations/1
• http://maphub.herokuapp.com/control_points/1
Harvard, January 2013 23
Open Annotation API
"bernhard.haslhofer@ oa: "2012-07-18T02:
cornell.edu" Annotation 01:07Z"
"behas"
"2012-07-18T02:
foaf:mbox rdf:type 01:07Z"
oa:annotated
foaf:name
oa:generated
http://
http://maphubdev.
uuid:1235 maphubdev.mminf.univie.ac.at/
mminf.univie.ac.at
annotations/1
oa:annotator oa:generator
oax:hasSemanticTag oax:hasSemanticTag
oa:hasTarget oa:Specific ct:ContentAs
Resource rdf:type Text
dbpedia: rdf:type
dbpedia:
Pillars_of_Hercules Strait_of_Gibraltar uuid:4567 dcterms:
oa:hasSelector
format
image/svg
cnt:Content oa:hasBody uuid:2752
oa:hasSource
AsText oa:hasSelector
rdf:type
uuid:2751
ct:chars <svg…..>
"text/plain" http://
samos.mminf.univie.ac. rdf:type
cnt:chars uuid:2753
dc:format at/maps/raw/
g3200.ct000725C.jp2
ct:ContentAs
dcterms:
Text
In antiquity, the Strait of Gibraltar (which format
dc:format
connects the Atlantic Ocean with the rdf:type
Mediterranean Sea) was also known by the
name "The Pillars of Hercules". This is the
reason for this inscription! image/jp2 application/wkt
ct:chars
POLYGON((5315
dcterms:StillImage
5639,5444….)
Harvard, January 2013 26
Semantic Tagging Study
• Effects of semantic tagging on
• comments, tag usage / types
• user satisfaction
• in-lab user experiment
• currently in-progress @ Cornell
Harvard, January 2013 30
Maphub Construction Areas
• Infrastructure stability
• GIS functionality + overlay generation
• Learning from Semantic Tagging (accept/
reject)
• Annotation moderation
• ...
Harvard, January 2013 31
Possible Extensions
• Semi-automatic annotation (toponym
detection) (http://rsimon.github.com/
toponym_identification/)
Josephinische Landesaufnahme: Gebiet von Mooskirchen bis Grazer Feld, Steiermark, Österreich, 1764-1787. Wikimedia Commons.
Harvard, January 2013 32
Possible Extensions
• Semantic Tagging with placenames from
different Gazetteers
• http://pleiades.stoa.org
• Pleiades dataset (http://thedatahub.org/dataset/pleiades)
• 34,000+ ancient places
• cool URIs for these resources
• http://pelagios-project.blogspot.com
Harvard, January 2013 33
Annotorious
• Light-weight image annotation library; no big client/
server framework
• Adds annotation features to existing Web pages
<html>
<head>
<link rel="stylesheet" type="text/css" href="css/annotorious.css" />
<script type="text/javascript" src="annotorious.min.js"></script>
</head>
<body>
<img src="example.jpg" class="annotatable" />
</body>
</html>
Harvard, January 2013 35
Annotorious
• Plugin-Framework
• Different storage backends and GUI components
(Tags, Replies)
• Additional selectors
• Interoperable with OKFN Annotator
https://github.com/annotorious/annotorious
Harvard, January 2013 36
Lucene-SKOS
• Plugin for Apache Lucene / Solr
• Works with SKOS-based knowledge contexts
• Uses semantic tags (links to SKOS concepts)
for query expansion
• query time
• indexing time
https://github.com/behas/lucene-SKOS
Harvard, January 2013 37
References
• Haslhofer et al. (2007): The LEMO annotation
framework: weaving annotations with the Web (IJDL)
• Haslhofer et al. (2010): Augmenting Europeana Content
with Linked Data Resources (I-Semantics)
• Simon et al. (2011): Semantically augmented annotations
in digitized map collections (JCDL)
• Haslhofer et al. (2012): Open Annotations on Multimedia
Web Resources (MTAP)
• Haslhofer et al. (2013): Old Maps and Open Data
Networks (iConference Demo / Poster)
Harvard, January 2013 38
Further Information
http://maphub.github.com
Thank you!
Harvard, January 2013 39
Acknowledgements (Maphub)
• Rainer Simon
• Werner Robitza
• Keith Newman
• Amanda Stefanik
• Carl Lagoze
• Francois Guimbretiere
Harvard, January 2013 40