Alexandria Digital Library Project ALEXANDRIA DIGITAL LIBRARY PROJECT Larry Carver James Frew Greg Janée Mike Goodchild Linda Hill Terry Smith

Alexandria Digital Library Project

ALEXANDRIA DIGITAL LIBRARY PROJECT

Larry Carver James Frew Greg JanéeMike Goodchild Linda Hill Terry Smith

www.alexandria.ucsb.edu


2Smith et al • NSF • April 3, 2003

Outline

Alexandria Digital Library Project (ADLP) History Goals, activities, partners

Distributed DL supporting georeferenced access Research and development issues Operational collections and services

Knowledge organization systems (KOS) Gazetteers and related KOS

ADEPT learning environment Concept-based learning spaces Collections and services



ADLP History

Pre-1994: UCSB geo-information and map library 1994-98 DLI-1: georeferenced collections/access 1998-99: Operational ADL (UCSB Library/CDL) 1999-2004 DLI-2: distributed DL

Extension of architecture and access services Knowledge organization services Integration of learning services Geo/GIS-based interfaces Basic CS research

2004-2008: Large-scale DLs and beyond NSDL Core Infrastructure and services Cyber Infrastructure



ADLP Goals

Current goals: Distributed DLs and applications Operational distributed digital library

– services for construction/use of georeferenced collections– DL federation and interoperation– scalability over many heterogeneous collections

Development/integration of KOS services Integration of concept-based learning spaces

– services for creating/using learning environments Development of geo-based interfaces Evaluation of services Basic computational science research

Emerging goals: Large-scale DLs and beyond Extending NSDL Core Infrastructure and services Cyber Infrastructure



ADLP Major Collaborative Activities

1994-98 4 DLI-1 partners: CMU, Illinois, Stanford, UCB SDSC, U.Arizona, US Navy, NIMA, LoC, MSFT, ESRI,…

1999-2004 UCSB Library, CDL DLI-2 partners: UCLA, GT, SDSC/NPACI, Stanford, UCB DLESE NSDL CI partners: Cornell, Columbia, U.Mass NSDL Services partners: IIT Chicago, UCSD JISC partners: Penn State, Southampton, Leeds …



• georeferenced DL tutorials• distributable software packages• operational libraries: UCSB library, ...• outreach; federated nodes

OPERATIONAL APPLICATIONS

• gazetteers: research and community• gazetteer content standard• web service protocols for gazetteers,

thesauri, and other KOS• ADL gazetteer• thesauri for feature and object types• duplicate detection for gazetteers• textual-geospatial integration services

KNOWLEDGE ORGANIZATION

• distributed georeferenced DL services• NSDL core infrastructure• data environment (e.g., GIS) integration• hardware acceleration for spatial data• collaborative tools• Z39.50 support• ingest and workflow systems

GEOREFERENCED DIGITAL LIBRARIES

• knowledgebase and lecture composing, visualization, and presentation tools

• physical geography concept space and learning object collections

• applications to undergraduate education• educational evaluation• learning services and DL integration• digital classrooms• metadata content standards

• learning objects• computational models

EDUCATIONAL APPLICATIONS

• reusable user interface components• contextual maps, footprint creation• KOS navigation• lightweight GIS functionality

• Digital Earth visualization• image processing

• query-by-content, classification• spatial extent determination

USER INTERFACES

ADLP Activities



Outline







Goals

Digital library architecture for geospatial/georeferenced information heterogeneous rich services scalable

– many providers– collections, large and small

DL infrastructure, not artifact standard components and interfaces distributed participants



Issue: discovery

Naïve approach I want a map of Boulder “Downtown street map of Boulder, Colorado”

But... remote-sensing imagery is nameless AVHRR NOAA-13 2002-06-03 14:33 UTC

But... direct placename search is unreliable I want a map of the Flatirons in the Rocky Mountains just

behind Boulder, Colorado USGS topographic map “Eldorado Springs” generally: many names for any given place



ADL approach

Coordinate-based representation and discovery lat/lon coordinates rich geometry

– polygons, polylines spatial operators

– overlaps, contains

Gazetteer content standard defines

representation service maps placenames

coordinates

client

gazetteer

library

coordinates

placenames



Issue: multiple data types

Geospatial discovery is not amenable to text treatment constitutes new data type

Adding notion of different data types has many implications: input validation internal structures, external representations query language and processing ranking user interface components



ADL approach

Discovery: “bucket framework” extensible data type system for metadata

– XML representations, search operations native metadata is explicitly mapped to buckets software supports bucket views over arbitrary RDBMSs 9 Dublin Core-like standard buckets

User interface components background maps, item footprint identification/creation

Spatial ranking by spatial similarity to query region



Bucket mapping

Originator

FGDCCitation/Originator

USGS DOQProducer

U.S. Geological Survey

Photo Science, Inc.

bucket-levelsearching

field-levelsearching

collectionstatistics



Collection statistics

Object Typecartographic works mapsimages photographs aerial photographs • • •

Count324,876324,876

2,014,799484,083484,083

Temporal

Spatial



ADL approach

Discovery: “bucket framework” extensible data type system for metadata

– XML representations, search operations native metadata is explicitly mapped to buckets software supports bucket views over arbitrary RDBMSs 9 Dublin Core-like standard buckets

User interface components background maps, item footprint identification/creation

Spatial ranking by spatial similarity to query region



ADL in context

ADL

Web

affo

rdan

ces

generality • structure

DLsGreenstone

ODL

OAI

GIS



Issue: scalability

Size easy to accumulate lots of data

– satellites image continuously geospatial discovery scales... not so well

– indexing unwieldy at 106 items efficiently joining spatial, other constraint types is difficult

Burden & management collection building is labor-intensive providers have differing content, services, IP concerns,

policies, lifetimes providers already exist

– MS Terraserver: 3 TB, 750 million items



ADL approach

Distributed library of peer nodes library nodes host collections other nodes host gazetteers, thesauri, other KOS other components, e.g., map servers

Federated item-level search over buckets over individual metadata fields mapped to buckets

Centralized collection-level search/ranking over collection statistics derived from bucket mappings

– space, time, type, format any library node can act as collection registry

Collection aggregation



Issue: context & use of library items

Context is critical in geospatial DLs formulating queries evaluating result sets and individual results

Use of geospatial data need access descriptions

– “item content single URL” is insufficient– multiple formats– multiple access methods– multiple components

need integration with common data environments– ARC/INFO, etc.



Geospatial context

Does this answer your question?

Flatirons #1-5

Flagstaff Rd.

Green Mountain



ADL approach

All library functionality is accessible via... web service APIs Java RMI

Content access model characterizes methods of access multiple “access points”

– download, service, web interface, offline– hierarchies of alternatives, decompositions

Context background maps library-supplied lightweight GIS functionality



Incorporation into NSDL/CI

Geospatial/georeferenced data is an instance of science data complex, well-defined structure rich metadata large size poorly served by traditional information retrieval methods

Science data belongs in NSDL

For NSDL: comparable infrastructure enabling... distributed, content-specific search services association of DL items and content-specific helper tools



Operational status

ADL co-developed with UCSB Library production-quality software foundation of operational library since 2000 complete system in 2003

UCSB Library: Map & Imagery Laboratory (MIL) self-supporting, 5 full-time employees 2.6 million items, 6.5 TB, growing 1.5 TB/year 4.5 million item gazetteer

Remote sites ESSW, CNR, DLESE, SIO, NTNU, AUT



Outline







KOS activities & contributions

KOS as primary components of DL architecture Heretofore not acknowledged as a major component ADL/ADEPT thesaurus and gazetteer service protocols

Gazetteer components of DLs Growth of a research and development community,

adopting/adapting/sharing our ADL Gazetteer components Gazetteer research issues NSDL Textual Geospatial Integration Project

KOS integration into learning environments Terry Smith will address this in detail



Digital Library ComponentsCATALOG

OF

METADATA

SERVICES

ACCESSING

ANALYZING

ARCHIVING

CATALOGING

DIGITIZING

RETRIEVING

SEARCHING

VISUALIZING

KNOWLEDGE ORGANIZATION SYSTEMS

AUTHORITY FILESCLASSIFICATION SYSTEMSCONCEPT SPACESDICTIONARIESGAZETTEERSGLOSSARIESONTOLOGIESSUBJECT HEADING SETSTHESAURI

DATA STORE

OF

OBJECTS

LibrariesCollections



KOS Generalization

Relationships

Label

TypeDefinition

Meaning

Navigation TranslationSense-making



Digital Gazetteer Essentials

(controlled vocabulary)

•None of these elements are unique identifiers of a particular place



Building gazetteer research community

1994-1996: ADL built the first multi-million-entry international gazetteer and integrated it into the ADL system

1996-1999: ADL created... Gazetteer Content Standard Feature Type Thesaurus (210 preferred terms; 1046 non-

preferred) rebuilt the ADL Gazetteer (over 4 million entries) provided web interfaces for searching the ADL Gazetteer



Building a research community 1999-present

Digital Gazetteer Information Exchange (DGIE) Workshop, funded by NSF (66 participants), 1999

JCDL 2002 workshop on Digital Gazetteers – Integration in Digital Library Services (38 participants; sponsored by NKOS)

NAACL 2003 workshop on Analysis of Geographic References

ADL-hosted discussion list for gazetteer issues; archived by NSF DLI2 (146 subscribers)

Set of 5.9 million geographic names available for download – useful for placename recognition in text

Gazetteer Service Protocol and protocol server code An “external identifier” for ADL Gazetteer records New gazetteer client that is based on the gazetteer protocol



Our network of gazetteer interactions

Electronic Cultural Atlas Initiative (ECAI) gazetteer project Academia Sinica’s Taiwan Gazetteer UK Historical Boundaries project

UK Geo-crosswalk project Digital Library for Earth System (DLESE) Education Biodiversity research, such as the “Specify” system –

University of Kansas

State projects, such as NY Agricultural History project (in proposal stage) and Florida statewide gazetteer project

University of Redlands internship proposal (mini-GIS) Bulgarian Antarctic Place-Names Commission

SRI’s Artificial Intelligence Center (spatial reasoning) Navy’s SPAWAR Systems Center (natural language process.) THREDDS project at UCAR (event gazetteers) Illinois Institute of Technology (geoparsing research)



Advancing and extending gazetteers

Named Time PeriodsWorld War I ___|_____|___

1914 1918

Named Spatiotemporal EventsSuch as Hurricane Hugo




What happens when we extend the digital gazetteer model to anatomy: named structures in the brain, for example?

http://www.ohiou.edu/~linguist/l550ex/brainpic.htm

Credit & Copyright: Sherry Buttnor http://antwrp.gsfc.nasa.gov/apod/ap011120.html

Anticline Famennian sandstone, Hastièrehttp://www.nitg.tno.nl/eng/iccp_tripj.shtml

Or to celestial space and 3-d features?




• Recognizing patterns

• Identifying features from gazetteers

• Deriving the extent of the features from feature analysis

• Adding bounding box footprints to gazetteer entriesSanta Barbara Municipal AirportSanta Barbara Municipal Airport

Obtaining extents from image analysis




Lake Bigler, thru 1920s Lake Bonpland (also Bondland),

thru 1890s Da-ow-a-ga, thru 1850s

The duplicate detection problem.

Given variant names and variant footprints, how do we determine that two pieces of information are about the same place?




From Michael Freeston, New Generic Indexing Technology

Effective and efficient database indexingtechniques for large spatial + text data collections

Test database of 2-d shapes in a geographic area to test the “sufficiency” of spatial generalizations (e.g., bounding boxes) for information retrieval based on spatial similarity (e.g., degree of overlap or containment)



Gazetteer ITR ProposalAdvancing and Extending Georeferencing

Interoperability and Services (AEGIS) Medium ITR proposal for 2003 Michael Goodchild, UCSB, PI Lewis Lancaster, Berkeley/ECAI, co-PI

Formalization and extension Performance and scalability Cross-cultural issues Cognitive and behavior issues Extents: representation of a feature’s geometry Integration of locator services



NSDL Textual Geospatial Integration

GoalsExtend NSDL infrastructure by

enabling geographic queries

across heterogeneous, text and non-text resources

spatial georeferencing of arbitrary texts without

explicit geographic cataloging

2001 - 2003

Participants

University of California, Santa Barbara

James Frew, PI Terence Smith Michael Bueno Linda Hill

Information Retrieval Lab, Illinois Institute of Technology

Ophir Frieder David Grossman Eric Jensen Steve Beitzel

The American Geological Institute (AGI) has permitted us to use a set of their GeoRef records for system training.


Example text -> Estimated footprintStructure and petrography of the schist of Skookum Gulch, Callahan-Yreka area, eastern Klamath Mountains, Northern California

<key>blueschist | California | Callahan California | foliation | Klamath Mountains | melange | metamorphic rocks | Ordovician | Paleozoic | petrology | schists | Silurian | Siskiyou County California | Skookum Gulch | United States | Yreka California</key>

<ab>The schist of Skookum Gulch (SSG) is an informal name applied to a fault-bounded melange composed mainly of schistose metamorphic rocks and less abundant sedimentary and igneous rocks located in the eastern Klamath Mountains of Northern California. The SSG features outcrops of lawsonite+sodic amphibole blueschist and epidote+sodic amphibole rocks transitional to the greenschist facies. Isotopic dating indicates that the schist was metamorphosed during the Ordovician. The SSG is the oldest known Paleozoic blueschist-bearing melange in California and one of the oldest preserved blueschist terranes in North America. Tonalitic rocks associated with the schist have Early Cambrian ages and are among the oldest rocks yet dated within the Klamath Mountains. Field relations indicate that the schist of Skookum Gulch is a complex tectonic melange composed of metavolcanic, ...</ab>

<coord>N410000N420000W1220000W1230000</coord>

• Derived footprint - small

• Blue: derived footprint – large

• Red: GeoRef footprint



KOS activities & contributions

KOS as primary components of DL architecture Heretofore not acknowledged as a major component ADL/ADEPT thesaurus and gazetteer service protocols

Gazetteer components of DLs Growth of a research and development community,

adopting/adapting/sharing our ADL Gazetteer components Research issues NSDL Textual Geospatial Integration Project

KOS integration into learning environments Terry Smith will address this in detail



Outline







Applications services based on DLs

Integrate applications with DL infrastructure Web portals lack library organization “packages” not integrated with DLs

Important applications include Services/collections supporting learning environments Services/collection supporting research

Apply domain-specific KOS principles for organizing collections/services for given application Geospatial applications: use georeference Science learning environments: use concept spaces



Science learning spaces: Concept KOS

Concepts of science as basic knowledge granules Sets of concepts form bases for scientific representation DL and KOS technology can support organization of science

learning materials in terms of concepts– Collections of models of science concepts (knowledge base)– Collections of learning objects (LO) cataloged with concepts– Collections of instructional materials organized by concepts

Organize learning materials as “trajectory through concept space” Lecture, lab, self-paced materials Services for creating/editing/displaying such materials



Learning environment components/services



Application to learning environments

Application Introductory physical geography (F2002, S2003)

Collections created Knowledge base (KB) of strongly structured concepts Structured lectures and labs Learning objects cataloged by ADN metadata (+ concepts)

Services created For concepts

– Web-based concept input tool– Graphic and text-based display tools

For instructional materials– Web-based “lecture composer”– “Conceptualization” graphing tool

For learning objects– Metadata input tool



Learning environment display (lecture mode)

The lecture is presented on three projection screens, showing the Concept window (left) Lecture window (center) Object window (right)



Model of science concepts

Representing a concept involves more than terms Objective, information-rich, scientific representations

– e.g., for concepts of heat diffusion, DNA, drainage basin, … Associated semantics

– e.g., relating to measurement, recognition,… Many interrelationships

– e.g., hierarchical, causative, property,…

Models of science concepts Already exist for chemistry (ASA), materials (NIST),… Generalize such models for this application

Structure items in concept KB using model



Model of science concepts

ID TYPE and FACET CONTEXT (KNOWLEDGE DOMAIN) TERM(S) (P/NP) DESCRIPTION(S) HISTORICAL ORIGIN(S) EXAMPLE(S) HIERARCHICAL RELATIONS DEFINING OPERATIONS SCIENTIFIC REPRESENTATION(S)

– Scientific classifications– Data/Graphical/Mathematical/Computational reps

PROPERTIES CAUSAL RELATIONS CO-RELATIONS APPLICATION(S)



Item in concept knowledge base



Concept input tool



Collections of learning materials

Lecture/lab composer Creates learning materials with

– Tailorable structure– Underlying organization as “forest of trees” of concepts

Small reusable granules for– Easy creation/edit/access/re-use

Can link in – Concepts from concept KB– Items from learning object collections– Items from lecture collection



Current instructional material window The left-hand

frame displays the structure of the lecture

The right-hand frame displays the content of the lecture

ADL icons (globe image) attached to a concept link to a display of concept properties in the concept window

Other icons attached to a concept link to a display of concept examples in the illustration window



View of learning material by concepts



Lecture/lab/… composer tool



Learning object collections

Cataloged with tool for metadata creation ADN metadata content standard with concept fields

Use of ADL/ADEPT middleware search services E.g., in creation of lecture/lab presentation materials

Display of collection items in collection window Photos, images, maps, text, videos,…

Support in display window for ADL browser Allows dynamic search of collection holdings



The illustrations window



Evaluation of concept-based approach

Evaluation of efficacy for student learning Do students attain “deeper levels” of understanding? Comparison approach to evaluation

Evaluation of value to instructors/TAs UCLA evaluation team

Evaluation issues Instrumenting students’ use of course materials

Time to assess pedagogic value of approach



Example of lessons learned

Importance of “conceptualizations” of concept e.g., characterize concept of Fluvial Landscape with

concepts of {River, Watershed} Embed conceptualizations in lecture/labs (not in KB) Idea of learning materials as trees in concept space

Construct labs using analogous “lab” composer Tailored for lab presentations/work Supports of logic of using concepts as framework Can import material from lecture/other collections



Summary

DL infrastructure as basis for Learning Environments Collections

– Concept KBs, Lectures, DL objects Services

– Creation/Search/Display

Evaluation of efficacy of approach Community-based development of KBs, Learning

Materials, Collections



• georeferenced DL tutorials• distributable software packages• operational libraries: UCSB library, ...• outreach; federated nodes

OPERATIONAL APPLICATIONS

• gazetteers: research and community• gazetteer content standard• web service protocols for gazetteers,

thesauri, and other KOS• ADL gazetteer• thesauri for feature and object types• duplicate detection for gazetteers• textual-geospatial integration services

KNOWLEDGE ORGANIZATION

• distributed georeferenced DL services• NSDL core infrastructure• data environment (e.g., GIS) integration• hardware acceleration for spatial data• collaborative tools• Z39.50 support• ingest and workflow systems

GEOREFERENCED DIGITAL LIBRARIES

• knowledgebase and lecture composing, visualization, and presentation tools

• physical geography concept space and learning object collections

• applications to undergraduate education• educational evaluation• learning services and DL integration• digital classrooms• metadata content standards

• learning objects• computational models

EDUCATIONAL APPLICATIONS

• reusable user interface components• contextual maps, footprint creation• KOS navigation• lightweight GIS functionality

• Digital Earth visualization• image processing

• query-by-content, classification• spatial extent determination

USER INTERFACES

ADLP Activities

Documents

Alexandria Digital Library Project ALEXANDRIA DIGITAL LIBRARY PROJECT Larry Carver James Frew Greg Janée Mike Goodchild Linda Hill Terry Smith