Transcript Slides - Academia Sinica

Spatiotemporal Infrastructure for
Semantic Network in Digital
Archives
Eric Yen
Computing Centre, Academia Sinica
Dec, 2002
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Outline
 Introduction
 NDAP Approaches – Space-Time-Language Coordinates
 Archiving and processing of millions of geospatial materials
in AS
 Characteristics
 How to delve into the knowledge level
 Experiences & Lessons we learned
 Extend to more general solution
 Geolibrary
 The Trends
 Conclusions
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Introduction to Digital Archive
 Digital Archive is a collection of digital objects.
 A digital object is defined as something (e.g., an image, an audio
recording, a text document, a movie, a map) that has been
digitally encoded and integrated with metadata to support
discovery, use, and storage of those objects.
 Goals for Digital Archive (functional point of view)








Protection of the original
Duplication for safety
Search and Retrieval
Easy Access
Resource Sharing
Lower cost of maintenance and dissemination
Max. flexibility for integration of heterogeneous/homogeneous
information resources
Providing abundant resources for knowledge discovery and knowledge
construction
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Knowledge Discovery and Construction
 Knowledge construction means the active process of
manipulating data to arrive at abstract models of relationships
among phenomena in the world that facilitate our
understanding of those phenomena and, ultimately, of the
world. [1]
 Knowledge discovery is a nontrivial process of identifying
valid, novel, useful, and understandable pattern in data. [2]
 Persistent cataloging, classification, and segmentation of
digital objects is the ground for finding patterns, models, and
trends of large volume data.
Reference:
1. MacEachren, A. et al, Constructing knowledge from multivariate spatiotemporal Data: integrating
geographic visualization with knowledge discovery in database methods
2. Fayyad, U., Piatetsky-Shapiro, G. and Smyth, P., 1996, From data mining to knowledge discovery:
An overview. In advances in Knowledge Discovery and Data Mining, pp.1-34.
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Types of Elementary
Knowledge Organization Systems












Classification Systems
Ontologies
Taxonomies
Index Languages
Thesauri and other controlled lists of keywords
Glossary
Dictionaries
Clustering Approaches
Lexical Databases
Concept Maps/Spaces
Semantic Road Maps
…
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Why Knowledge-based
Approach for Digital Library ?1
 Providing “Conceptual Infrastructure”
 Mapping out the conceptual structure and providing a common language for a field
 Providing classification/typology and concept definitions. Clarifying concepts by putting
them into context. Thus providing orientation and serving as a reference tool for individual
researchers and practitioners and thereby
 Assisting with the exploration of the conceptual context of a research problem and in
structuring the problem, thereby providing the conceptual basis for the design of good
research, for the consistent definition of variables, and thus the cumulation of research
results.
 Providing the conceptual basis for the exploration of the various aspects of a program in
program planning, in the identification of approaches and strategies, and in the
development of evaluation criteria
 Assisting users in understanding context
 Assisting information providers with conceptualizing a topic and with finding
the proper term
 Discovery of high quality resources
 Providing frameworks for information exchange and resource interoperability
Dagobert Soergel, Evaluation of Knowledge
Organization
(KOS)and Digital Libraries
2002APEC
WorkshopSystems
on e-Learning
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Why Knowledge-based
Approach for Digital Library ?2
 Information Storage & Retrieval
 Information system(s) in which the vocabulary is to be used
 Use of the vocabulary
 Vocabulary control in indexing and searching (controlled vocabulary)
 Vocabulary control only for searching. Assist with clarifying a search topic and
assembling all applicable concepts and terms, whether searching with a controlled
vocabulary of free-text.
 ISAR technique(s) (such as: printed index, computer search system). Support of
inclusive (hierarchically expanded) searching
 Automated vs. manual indexing or query formulation. Approach to indexing to be
supported: Request-oriented vs. entity-oriented
 Techniques for eliciting user needs (e.g., menu based on search tree; questions based
on facet structure)
 Summary evaluation of the vocabulary's adequacy for the stated purpose on the
more detailed analysis as outlined below.
 Translation
 Language learning
2002APEC
WorkshopSystems
on e-Learning
Dagobert Soergel, Evaluation of Knowledge
Organization
(KOS)and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Digital library requirements for
knowledge organization schemas
 The need for knowledge organization in subject gateways
and discovery services, issues of application and use
 Web-based directory structures as knowledge
organization systems
 Knowledge organization as support for web-based
information retrieval, query expansion, cross-language
searching
 Semantic portals
ECDL2000, Special Workshop on Networked Knowledge Organization Systems,
http://nkos.slis.kent.edu/ECDL-NKOS-final.htm
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Digital library requirements for
knowledge based data processing
 Knowledge organization for filtering, information
extraction, summary
 Knowledge organization support for multilingual systems,
natural language processing or machine translation
 Structured result display, clustering
 End-user interactions with knowledge organization
systems, evaluation and studies of use, knowledge bases
for supportive user interfaces, visualization
ECDL2000, Special Workshop on Networked Knowledge Organization Systems,
http://nkos.slis.kent.edu/ECDL-NKOS-final.htm
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Digital library requirements for
knowledge structuring and management
 Suitable vocabulary structures, conceptual relationships
 Comparison between established library classification
systems and home-grown browsing structures
 Methodologies, tools and formats for the construction and
maintenance of vocabularies and for mapping between
terms, classes and systems
 Frameworks for the analysis of assumptions and
viewpoints underlying the construction and application of
terminology systems
 Methods for the combination and adaptation of different
vocabularies
ECDL2000, Special Workshop on Networked Knowledge Organization Systems,
http://nkos.slis.kent.edu/ECDL-NKOS-final.htm
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Digital library requirements for access to
knowledge structures
 Data exchange and description formats for knowledge
organization systems, the potential and limitations of
XML and RDF schemas
 Handling of subject information in metadata formats
 Standards and repositories for machine-readable
description of networked knowledge organization
schemas (as collections/systems)
 Interoperability, cross-browsing and cross-searching
between distributed services based on knowledge
organization systems
 Distributed access to knowledge organization systems:
standard solutions and protocols for query and response,
taxonomy servers
ECDL2000, Special Workshop on Networked Knowledge Organization Systems,
http://nkos.slis.kent.edu/ECDL-NKOS-final.htm
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Discover Knowledge from
Digital Archive
 Geospatial information means those geo-materials that are
georeferenced and having well-documented metadata
 Ref. Components of a digital object in digital archive
 Geospatial Content Based
 Extracting knowledge by space-time-language
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Knowledge about Space
 Temporal Characteristics is embedded and could not be neglected
 Acquisition
 Direct Experience
 Locomotion thru environment(crawling, walking, running, bicycling, driving,
flying, etc.)
 Stationary viewing
 Secondary Environmental Experience
 Static medium: maps, diagrams, paintings, photos, etc.
 Dynamic medium: animate static visual figures to show changes over time
 Other ways to conceive those that can not be viewed
 Characteristics
 Multimodal: proprioceptive, kinesthetic, auditory, visual, etc.
 Language is often used to convey spatial information
 Multi-perspective and scales
 充分瞭解人類獲取、整合與利用空間資訊模式，將可促進此類
資訊的更有效利用，以及建立更符合實際需求的應用機制(e.g.,
aid for decision making)
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Spatial Representation in GIS
 Data Model
 Vector: explicit
 Basic elements: point, line and polygon
 Raster: implicit
 Geographic space is organized into partitions (layers)
 Space-dominant representations focus on the spatial arrangement
of entities based on the geometric and thematic properties of
these entities.
 Space is a neutral container
 Entities only exist when associated to a layer or theme
 Applied primarily in traditional mapping
 Layer-based raster and vector models
 Each layer is associated to a period or point in time
 Change- or update-based scenario
 Analysis based on similarity or dissimilarity between aggregations
(layers) at different points of time
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Why Thinking in Spatio-Tempoal
ways?
Because the earth is running: It’s
incomplete to describe an events/object in
spatial domain only.
Learn from the past, and plan for (predict)
the future.
Characteristics of Space & Time
Importance
To organize space over time
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Discover Knowledge from
Geospatial Information
 Geospatial information means those geo-materials that are
georeferenced and having well-documented metadata
 Ref. Components of a digital object in digital archive
 Geospatial Content Based
 Feature Identification
 Feature comparison: enhance the likelihood of relationships among
features
 Feature interpretation: merge the identified features and their
relationships with real world entity, by domain knowledge
 Linking to other resources that are related to this feature, this place
and the time  parsing the collected information from metadata or
lexical analysis
 Demands
 Link spatiotemporal data analysis techniques to GIS
Feature interpretation tools must provide connections between abstract representations
of data, metadata that describe those data, an analyst’s
knowledge,
andonknowledge
2002APEC
Workshop
e-Learning and Digital Libraries
sources external to the data set being explored (e.g., thru digitalAcademia
library) Sinica, Taipei, Taiwan. Dec. 16-20
Discover Knowledge from Geospatial Information
Feature Identification
 Def: Finding instances of identifiable features in spatiotemporal data
 Emphasis is on examining the distribution of data in all of its dimensions in an effort to
notice any distinct object, regularity, anomaly, hot spot, etc.
Example:
Distribution of Tombs in
Han Dynasty
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Integrated Support for
Research
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
WebGIS-based System Architecture
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Challenges of
Geospatial Information Processing
 High threshold for general users
 Hard to find required geospatial content/service
 New retrieval technology for geospatial
information
 Persistent metadata and archive
 Mechanism for effective management of huge
volume of data set
 Efficient ways for digitization/vectorization of
geospatial materials
 Integration with other information resources
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Discover Knowledge by
Space-Time-Language Coordinates
 Constructing the linkage among diversified archives thru
language (vocabulary)
 Lingual coordinate has both spatial and temporal extents
 Lingual-Temporal Plane: evolution of language thru time
 Lingual-Spatial Plan: spatial distribution in dialect
 Multi-lingual support for digital archive
 Establishment of domain-specific controlled vocabulary sets,
and serve as basis of ontology
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Discover Knowledge by
Space-Time-Language Coordinates
Time
Space
Language
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Space, Time and Language Coordinates for Digital Archives
Time
Historical
GIS
Space
Digital
Archives
Language
in Time
Language Changes
Language
in Text, in
Speech...
Language
Language
in Space
Language variations
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Lingual Coordinate in NDAP
 A lexis/vocabulary in context is analogy to the basic unit of a concept in knowledge
 Lexis is the basic unit for any kind of language process, such as recognition, parsing,
wordformation, semantics, conversation and analysis
 Thru lexical analysis, collection of all the lexical types(詞類), lexical patterns(grammar文法),
and instances could pave the base as lingual coordinate.
 Collection of enough description(context incl. metadata) for a specific domain(could be a set of
digital objects), ontology(collection of concepts for the domain) of that field is constructed. 
How do we know if that is enough?  Need the self-learning capability in the mechanism
 Atomic attributes of a place name
 Name
 Glyph & stroke: original writing, all the historical and contemporary writing, and Romanization(pinyin)
 Pronunciation: indigenous and evolutions afterward
 meaning (if we could restore to original fonts & sound)
 Footprint
 Could be ambiguous: M N
 Time: (start, end), could be vague for historical names
 Type: (geographic type, also could know the administrative level if it represents an administrative
area)
 Atomic attributes of a datum
 People, event, time, place, object
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Constructing Space-Time-Language Coordinates for NDAP
 Geographic searching is a powerful and important tool
 More than 80% information resources pertain to specific geographic areas and are either
explicitly or implicitly geo-referenced.
 To utilize benefits of geographic search, we have to geo-reference information contents first.
 the cost of creating geographic footprints for each record (the Alexandria Digital Library Project
spent $4m over four years) is very high. The automatic extraction of geo-referenced information
is also possible but there is a need for sophisticated tools that go further than geographic name
extraction.
 Moving from information management toward knowledge management
 (Demands) New ways of information search & retrieval




Traditional full-text search
Keyword-based or query by example search
Query by information content (image, audio, video, and multimedia contents)
Incorporation of geographic & temporal search
 Versatile ways for presenting information & knowledge
 2D, 3D, or 4D
 Multimedia, virtual reality
 Map-on-demand, thru the parser of geographic names from context, or directly by the coordinates
 Separation of content representation & presentation
 The core is the metadata-based content analysis
 CA(Information Content)Metadata Schemes for management of contents
 Identify the best way of information representation and become persistent archive
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
中國歷史文化地圖之整合應用
清代地方誌檢索
漢籍全文檢索
圖書聯合目錄
查詢
人物資料庫查
詢
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Roles of Visualization in
Knowledge Discovery
 Role




Useful in finding holes or errors in data sets
Useful for noticing abstract features and patterns
Predigest complex relations of data sets into visual form
Facilitate access to multiple perspectives on information, thru
interactivity
 Facilitate decisions on appropriate model representation during
analysis stage.
 Process tracking: uncover key aspects of a process
 Parameter control to get corresponding outcome on the fly
 Functionality
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Geolibrary
 Objective: Lower the barriers for applying GIScience
technologies
 Approaches
 Collecting and providing basic georeferenced spatial data/knowledge
persistently
 Building up application environment and tools for utilization of
spatiotemporal knowledge and technologies
 Development of spatiotemporal-based technologies for multi-disciplinary
contents integration, aggregation, knowledge discovery in map-metaphor
 Focus & Approach
 Construction of the System Infrastructure for Spatial and Temporal
Information Technology
 Development of Core Technology
 Establishment of Effective Service Model for Research Support
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Clearinghouse
 An instance of implementation of interoperability
 Functionality
 Locating the required resources/services
 Maintaining a persistent catalog of resources/services for
sharing
 Exchange of information content
 Format transformation
Clearinghouse (catalog)
Metadata
Framework GEOdata
Standards
Partnerships
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Effective Management System for Huge Volume
of Data
 Remote sensing data: 2TB/day;And will accumulate to 5 Peta
Byte in 2005。
 According to the statistics of EU Space Center
 Raw data from satellite : 100GB/day, 500GB/day (after Feb. 2002)
 800 TB data had been archived
 Big Challenge of IT for cataloging, searching, retrieval,
management, identification, knowledge discovery, and integration、
 Trading off between decentralization and consolidation on cost,
 Convergent to multi-centers of information resources in Internet
 Think about how to facilitate the collaboration among those centers –
Community and virtual organization
 Demands for complete architecture and services Data Grid
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
What’s the Solution
 Support sharing and coordinated use of diverse resources in
dynamic “virtual organizations” – Grid !
 Good technical solutions for key problems, such as




Security enhancement like authentication and authorization
Resource discovery and monitoring
Reliable remote service invocation
High-performance remote data access
 -- Grid !
 Good quality reference implementation, multi-lingual support,
interfaces to many systems, large user base, industrial support,
etc. – Grid !
 Persistent Web Services – Grid !
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Measuring Success






High degree of component autonomy
Low cost of infrastructure
Ease of contributing components
Ease of using components
Breadth of task complexity supported by the approach
Scalability in the number of components
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20
Conclusions and Future Work
 Building the right infrastructure will be crucial
 Intersection of spatiotemporal coordinates and lingual
coordinate constitutes a good framework both for knowledge
extraction and interoperability
 Consensus gathering and technology development still the
major challenges for interoperability
 Open System, Open Standard, and Open Source
2002APEC Workshop on e-Learning and Digital Libraries
Academia Sinica, Taipei, Taiwan. Dec. 16-20