Transnational Digital Government Research: Building Regional Partnerships

1
Transnational Digital Government Research
Building Regional Partnerships
José Fortes (on behalf of the TDG team) Dept. of
Electrical and Computer Engineering and Dept. of
Computer and Info. Science and Engineering Univers
ity of Florida
2
The TDG team

• University researchers and staff members from OAS
  and agencies in the US, Belize and Dominican
  Republic
• North Carolina State University (A. Anton)
• Carnegie Mellon University (J. Carbonell, V.
  Cavalli-Sforza)
• University of Belize (C. McSweeney)
• University of Florida (J. Fortes, S. Su, I.
  Krsul)
• University of Colorado (R. Cole, W. Ward)
• University of Massachusetts (D. Towsley)
• U. Pontificia Católica Madre y Maestra (L. de
  Brens, J. Ventura, P. Taveras)
• OAS CICAD (R. Connolly, C. Ortega)
• National Drug Abuse Control Council (Belize) (O.
  Brooks)
• Consejo National de Drogas (Dominican Republic)
  (M. Herrera)
• Agencies in Belize and Dominican Republic

3
Outline

• The TDG team and project background
• Transnational digital government
• Challenges
• Research
• Organization
• Coordination
• Infrastructure and funding
• Political
• Conclusions

4
The long path behind TDG

• Int. Collaboration in Computer Science, WA97
• benefits/barriers
• to advance ST solve global problems
• Western Hemisphere Collaboration, FL98,
  Mexico99
• scientists from Latin America, Canada, US, OAS
• OAS/CICAD meeting, Montevideo99
• approves drug information network in the Americas
• US-Chile/Argentina Collaboration,Chile/Argentina0
  0
• collaboration and development plan
• Transnational Digital Government (TDG), Belize
  01
• TDG Kickoff(s), Dominican Rep. 02

5
Transnational Government

• Agreed actions among agencies of distinct
  countries to address global problems
• Transnational Digital Government
• Uses information technology to enable or improve
  government processes.

6
Research goals of the project

• To advance the state-of-the-art of spoken
  dialogue systems, machine translation,
  collaborative information management, Internet
  portals and services, network performance
  optimization, and software requirements in the
  context of a specific TDG process the MEM.

7
Research challenges

• Heterogeneity everywhere!
• Languages
• Agency and staff skills
• Infrastructures
• Networks and other communication means
• Computer hardware and software
• Regulations
• Government structures
• Culture and politics
• Time!

8
Building a partnership

• Representative transnational scenario
• Requirements
• Two small countries with distinct languages
• High-level commitment to TDG activities
• Local universities teamed with agencies
• Choices
• Belize
• Dominican Republic
• US (OAS headquarters)

9
Organization of the American States

• oldest regional international organization
• North, Central and South America and the
  Caribbean (the Western Hemisphere)
• forum for multilateral dialogue and action
• English, Spanish, French and Portuguese
• Inter-American Drug Abuse Control Commission
  (CICAD)
• creating a multilateral evaluation tool to
  measure progress of countries in meeting
  anti-drug goals

10
Context the OAS CICAD MEM

• states mutually evaluate progress in all aspects
  of the drug problem collect, share and analyze
  information per agreed-upon indicators
• goal reduce production, traffic consumption
• countries to improve counter-drug performance by
  own means, or with technical assistance by OAS
• challenges objective, up-to-date, comparable and
  exchangeable data
• transnational surveys, traffic data/documents
• currently, much data is not readily accessible
  because it is processed manually

11
A glimpse at the MEM questionnaire
12
Belize

• 1981 - present
• size of Massachusetts
• 270,000 people
• Languages
• English (official), Spanish, Mayan, Garifuna,
  Creole
• University of Belize
• School of Engineering IT

Source The World Fact book 2002
13
Dominican Republic

• 1844 - present
• 2 X size of N. Hampshire
• 9 Million people
• Languages
• Spanish
• Pontificia Univ. Catolica Madre y Maestra
• Computer systems

Source The World Fact book 2002
14
Transnational Issues 

• Disputes - international
• the "Line of Adjacency" established in 2000 as an
  agreed limit to check squatters settling in
  Belize, remains in place while the OAS assists
  states to resolve Guatemalan territorial claims
  in Belize and Guatemalan maritime access to the
  Caribbean Sea Honduras claims the Sapodilla Cays
  off the coast of Belize
• Illicit drugs
• major transshipment point for drugs small-scale
  illicit producer of cannabis for the
  international drug trade some money-laundering
  activity related to offshore sector

Source The World Fact book 2002
15
Conversational interface systems

• User does not need to write or read in order to
  receive/provide information
• Conversation between user and computer as
  questions and answers
• Examples of usage
• animated characters who can talk with children to
  educate them on drug-related matters
• assistance on data collection
• detect and respond to suspicious activities at
  countries' borders

16
Why are conversational interfaces hard?

• variability in languages, dialects, environmental
  noise and communication channels
• annotated speech from native speakers required to
  model variability and achieve good performance
• expensive studies - human operator controls the
  computer response - to model language use
• approaches
• statistical modeling of speech patterns to
  generate word sequence hypotheses
• robust semantic parsing to assign words and
  phrases to conceptual categories
• sophisticated modeling of dialogue interaction to
  achieve task goals

17
Machine translation (MT)

• For documents, dialogues
• Example of use
• Documents shared among collaborating countries
• Records of conversations between officials and
  travelers
• Descriptions of travelers at border crossings
• Text Information in fixed-format databases
• Problem very hard unless domain-restricted
• Words must have unambiguous meaning, e.g. drug
• Language differences meaning, structure, idioms

18
Comparing Approaches to MT

• Interlingua and syntactic transfer MT
• labor intensive/expensive, no learning from
  experience
• Statistical MT
• huge amount of parallel text (billions of words)
  to work well
• not easily available, especially in our domain
• Example-Based MT is a good compromise
• less parallel in-domain texts than statistical MT
• like statistical MT, learns from additional data
  w/ less labor

I met the Pope ? Conocí al Papa The tallest man
is my father ? El hombre mas alto es mi padre I
met the tallest man ? Conocí el hombre mas alto I
met the tallest man ? Conocí al hombre mas alto
Given examples
New sentence
Slide by J. Carbonell and V. Cavalli-Sforza
19
Data/knowledge management

• Enables secure, timely, authorized access to
  shared data and distributed information
• Enforces rules and policies of interactions among
  agencies to ensure security, privacy and
  coordination of collaborations
• Example of use
• checks of timely complete drug-use reports while
  protecting identities and data integrity

20
End goal Virtual Collaboration Grids
1. Grid 1
2. Grid 2
Country X
3. Grid 3
US
Belize
Country Y
Internet
Country Z
Dominican Republic
Country W
Slide provided by S. SU
21
Purposes/Challenges of Collaboration Grids

• Share heterogeneous data, knowledge and
  application system resources and allow human
  programmatic accesses to these resources
• Coordinate inter-government and
  inter-organizational activities by
• querying for distributed data,
• automatic event notification,
• information delivery,
• application system invocation, and
• knowledge rule processing.

Slide provided by S. SU
22
How does our approach work
Detailed presentation and discussion of
this prototype system will take place in session
2B this afternoon at 130 pm Presentation
title Enabling Transnational Collection,
Notification, and Sharing of Information
23
Internet portals and services

• Web sites where collaborating agencies can find
  and use data, documents and software to process
  information
• Web site where agencies can place data, ask
  questions or assistance, interact with similar
  agencies
• Example of use
• MEM survey templates and statistical tools to
  summarize survey results can be obtained and used
  on the OAS web site

24
A possible new OAS portal
25
What is behind portals
26
Why virtual resources and grids?

• Hiding of heterogeneity
• resource differences, lack of resources
• security and privacy
• sustainability and scalability
• time independence
• natural for collaborative efforts
• A basis for
• flexible and portable test beds
• cyberinfrastructure for digital government
  research and deployment

27
Network support for collaboration grids

• services to compensate for variations in
  performance of the Internet
• example of use
• keeping authorized copies of data on local
  computers in anticipation of needs of
  users/agencies (dissemination)
• application-level overlay networks (fast,
  transparent)

28
Network support for collaboration grids

• Goals
• improve performance
• mask local performance impairments
• support for event advertisements/subscriptions
• policy driven security
• authorization/access in heterogeneous environment
• anonymity

29
Network support for collaboration grids

• Approach Pub/Sub network
• domain-specific portals
• brokers responsible for
  distribution, inter-domain
  interactions
• Benefits
• domain-level security
• performance enhancements
• application-level multicast
• routing around problems

30
Legal and cultural heterogeneity

• Raised by reviewers and agency representatives
• IT solutions must be mindful of needs and
  expectations of users and institutions
• A. Anton joins TDG team - expert on
• software requirements analysis
• aligning privacy and security

31
Software requirements

• The inquiry cycle

Slide provided by A. Anton
32
Manageable first steps

• TDG Kickoff(s), Dominican Rep. 02
• identified MEM IT systems/processes requiring
  transnational collaboration
• MEM has 83 indicators
• many agencies interact for each one
• need country/agency buy-in
• Selected indicator 83 Displacement
• new trends in the global phenomenon of the
  mobility of the different manifestations of the
  drug problem - e.g. trafficking/routes

33
Remote border control

• Very specific domain
• timely
• support at highest-levels
• transnational in many ways
• builds on existing IT infrastructures

34
Participant roles
Dominican Republic
Belize
USA
Pontificia Universidad Católica Madre y Maestra
University of Belize
CMU, NCSU U. Colorado U. Florida U. Mass.
Universities
National Drug Council
Ministry of Health
OAS headquarters
Agencies
35
The reality
36
Transnational Dig. Gov. research is

• transnational!
• Heterogeneity rules!
• Language barriers
• Security
• Buy-in from agencies
• Infrastructure differences
• Bureaucracy and different institutional cultures
• Thinly divided funds

37
Solutions to language barriers

• Translators for political meetings
• Mostly bilingual team (12/17)
• Selective translation of documents
• by OAS
• by other team members
• Translation services are expensive!
• can automatic translation help?

38
Coordination challenge

• 17 participants
• plus students
• Web-based forum
• Review meetings
• 3 in 2003
• Exchange visits
• Calls, email
• Teleconferencing

39
Institutional bureaucracy/cultures

• academia vs. OAS vs. agencies
• similar institutions in different countries
• one and half year from intention to fund to
  arrival of funds
• diplomatic concerns
• distinct contract languages
• first-time scenario
• deep subcontracting
• solution re-budgeting and other sources

40
Thinly divided funds

• 1 student 1 faculty-month per university
• very limited funding for infrastructure
• supplements for
• undergrad research
• work on security-privacy alignment
• leveraging ongoing work at PIs laboratories
• seeking complementary funding sources

41
Infrastructure differences

• Everywhere
• computers, networks, software
• Three-phase approach
• U.Florida-centered system
• system distributed across universities
• system distributed across countries
• Virtualization technology planned
• for test bed
• for deployment

42
Summary

• an exciting unique project (3 years, just started
  )
• strict goal to advance research on IT that
  facilitates transnational collaborations
• heterogeneity is major barrier
• virtualization-supportive infrastructure needed
• broad potential to lead to
• solutions adoptable by many countries
• IT can help TDG but will it?
• need systems-oriented research, international
  collaboration and flexible infrastructure

inspired by Gio Wiederhold
43
Acknowledgements

• NSF
• The Digital Government program
• Larry Brandt and Valerie Gregg
• Program Directors
• Rita Rodriguez
• Gary Strong
• OAS CICAD
• Ministry of Health, Belize
• National Drug Council, Dom. Republic
• The TDG team

44
The real reality
45
Sustainability and scalability