active knowledge delivery in semi-structured administrative processes gregory mentzas information...
TRANSCRIPT
Active Knowledge Delivery in Semi-Structured Administrative Processes
Gregory MentzasInformation Management UnitNational Technical University of Athens [email protected]
Andreas AbeckerKnowledge Management DeptGerman Research Center for Artif. Intelligence [email protected]
Workshop on Electronic Government & Knowledge ManagementSiena, Italy, May 22-24, 2001
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 2
Overview Some facts and figures about
the DECOR project Basic assumptions and the
„overall picture“ Core technical goals and their
relevance for E-Government: Active knowledge delivery Weakly-structured processes Process-oriented structured
archives
The IKA example Summary
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 3
DECOR is a two-years European R&D project developing and testing new knowledge management methods and tools
DECOR - delivery of context-sensitive organizational knowledge
6 partners from 3 European countries
runs in the Information Society Technology Programme (IST), within the
European Union's Fifth RT Framework Programme (1998-2002)
overall funding by the European Commission 1.8 MECU - under contract
No. IST-1999-13002
Project duration: July 2000 – June 2002
Main idea: Business-Process Oriented Knowledge Management for
Administrative Processes
DECOR facts and figures
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 4
The DECOR consortium consists of technology and method providers, consultants, and end users
DECOR partner organizations
IKA - the mostimportant social security institution in Greece
PLANETERNST & YOUNG - Greek management consultancy company DHC - an innovative IT
consulting and software development firm in the SAP area
SEMA Group Benelux - one of Europe‘s leading IT consulting and systems integration companies
DFKI - the German Research Center for Artificial Intelligence
ICCS / IMU – Information Management Unit of the National Technical University of Athens
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 5
DECOR demonstrates knowledge-based document orga-nization and process-embedded, active document delivery
intelligent document archives can effectively support knowledge-intensive work
documents can be indexed and organized according to multiple criteria and different viewpoints
models for document organization may themselves be complex structures
document organization models („ontologies“) provide flexible use for information access
ontologies contain background knowledge for better search and retrieval
Basic assumptions (1/2)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 6
DECOR demonstrates knowledge-based document orga-nization and process-embedded, active document delivery
formal representations of a company‘s work routines and business
processes are useful knowledge organization model
in a given work procedure, employees often spend much time for
(sometimes unsuccessful) information search
the specific information requirements of a
given task can be analysed in advance and
automatically fulfilled during process enactment
dynamic process context and formal models of
the business process domain improve search
Basic assumptions (2/2)
How do the DECOR modules play together?
Start
purchase
Specify
good to buy
Good =
„Hardware“
Good =
„Software“
Price >= 300 €
Price < 300 €
Par. 370324
Specify detailed
HW parameters
Weakly-structured
workflows represent
knowledge-intensive
business processes
Augmented workflow models
describe information flow between
work activities and information
needs for specific tasks
For augmented work-
flow and associated
indexing ontologies,
modelling tools
and a methodology for
organisational
take-up are developed.
The pilot systems are
built on top of concep-
tually indexed archives
with domain ontologies and
business process models
as indexing structures.
The information
assistant observes
the running work-
flow and offers
active, context-
sensitive
organisational
knowledge, thus
promoting a better
exploitation
of existing know-
ledge sources.
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 8
The technical realization for active knowledge delivery extends conventional workflow concepts
Active knowledge delivery (1/2)
RetrievalAgent
SourcesInformation
Sources
WF ControlData
WorkflowEngine
WorklistHandler
Applications
WF ApplicationData
invokes
+Extensions
forknowledge-
intensiveTasks
BusinessProcessModel
interpreted by
WF Rele-vant Data
+Extensions
support
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 9
Active knowledge delivery enhances the quality of service for E-Government processes
Active knowledge delivery (2/2)
Active knowledge supply prevents that important knowledge or obligatory
information are overseen; compliance with binding (legal) regulations is
ensured
Active hints to other employees’ decisions enable equal decisions under equal
conditions and foster communication between employees
Active knowledge push ensures the dissemination of new knowledge (changed
laws etc.)
The “normal citizen” involved, e.g. via a Web interface, into an administrative
process can be part of complex processes without having in advance all
required background knowledge
Enhanced transparency even in “one-stop government” (Lenk and Traunmüller, 2000,
“innovative ways of service delivery”)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 10
The DECOR workflow concept is motivated by the characteristics of knowledge-intensive work
Weakly-structured processes (1/3)
Knowledge-intensive work is often ad-hoc to a some extent;
but not without any regularities, commonalities, and structure
Knowledge-intensive work varies along the project-process spectrum:
often no strong sequential order, but lists of required activities
often cyclic, chaotic order of steps, but logical dependencies
however, sequential parts or reusable templates for parts
at least some MS project like, or Timex like support possible
sort of top-down refinement useful while actually executing the process
Cp. (Wimmer & Traunmüller, 2000), „type 2 administrative processes“:semi-formal procedures, several people with conflicting interests involved in negotiation and decision making
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 11
DECOR’s weak-workflow support shall combine a project manager’s flexibility with traditional workflow’s complexity handling
Design goals:
1. Workflow management system as an assistant
2. Hierarchical decomposition
3. Modeling and enactment interleaved
4. Expressive process logic
5. Context-sensitive knowledge delivery
low high
reuseability
modifiability
complexity
knowledge usage
uniqueness
predictability
spontaneousmodeling decisions
structure
project management system traditional workflow
Weakly-structured processes (2/3)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 12
Several reasons ask for weakly-structured process support in Electronic Government
Some workflow support is useful anyway: many people, many documents and maybe many locations are involved; enforcing a given process structure through a workflow system may improve quality of
services and definitely increases throughput
Reasons for weak process structures: Long-living processes may change because of changing environment (regulations,
information situation, political decisions) during their enactment (see the work by Maurer et al. [MILOS System] about urban planning)
Many administrative processes are very well structured in general, but contain few central decision steps which are not further regulated, but could be planned
In some highly-complex and safety-critical administration processes, no a-priori-planning is possible, but process instances must be recorded and documented for legal reasons (DFKI currently examines such a process in the area of nuclear transports)
Weakly-structured processes (3/3)
Cp. (Lenk & Traunmüller, 2000) about E-Government processes: “They are partly … structured by legal rules which however, often demand interpretation …”
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 13
The technical basis for the DECOR intelligent archive is DHC‘s CognoVision® product
Process-based structured archives (1/4)
Knowledge objects (documents) are enriched with Metadata (author, content, quality, actuality, ...) Links to other documents Links to content-describing ontology concepts and business process
elements An ontology is an „ ... explicit representation of the shared conceptualization of a
given domain. ...“, i.e., it specifies the common vocabulary and basic assumptions (concepts, relationships, axioms) used by a number of actors to talk about a given domain
Ontologies provide the index terms for describing document content and can be used for manual navigation, information retrieval, query expansion, etc.
In DECOR, the CognoVision® tool is used as document management system: In CognoVision®, arbitrary networks (representing ontologies) can be built
from structure elements and attributed links Documents (maybe different versions) and metadata are grouped in
information units and linked to structure elements
The CognoVision® tool (DHC) allows to establish multiple views and manifold links between documents
Process-based structured archives (2/4)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 15
content of info unit 2
other info units linked to info
unit 1
links to / from information unit
1
content of information unit
1
structure elements linked
to info unit 1
structure elements liked to info unit 2
information unitslinked to info unit 2
Information units group documents, metadata, and links to structure elements / other information units in CognoVision®
Process-based structured archives (3/4)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 16
Ontologies and metadata are well-suited concepts for knowledge organization in E-Government
Process-based structured archives (4/4)
Importance of intelligent document repositories with a coherent
view on heterogeneous input material: Many different sources of knowledge: laws, comments to laws, specific
regulations, old similar cases, available case-specific documents and
information
Often, no explicit links between related knowledge items possible
Often, knowledge items prevalent at different places and in different forms and
representations, at several degrees of formality
The E-Government domain is well suited for modeling background
ontologies: Many topics (especially legal regulations) are formal by nature
People with formal education exist which can assist in knowledge acquisition
Once built, formal ontologies can be resued for different purposes: navigation
and retrieval support, explanation, reasoning, education purposes, etc.
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 17
IKA is the largest and most important organization for social security in Greece
Some facts and figures:
330 insurance branches and 360 health branches all over the country
A total personnel of 23,000 permanent employees (administrative- medical personnel and doctors)
Medical treatment to 5,6 million directly and indirectly insured people
Pensions to 1.000.000 pensioners
IKA offers a wide range of services, which fall under four categories:
Insurance
Benefits in kind and in cash
Pensions
Interstate social security
IKA case study (1/4)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 18
In DECOR we consider the process for granting full old age pension
A business process of high importance for the institution: Current beneficiaries amount to 1.000.000 persons Annual increase rate: 10% Time-consuming process, resulting in long delays in the completion of the process
A knowledge-intensive business process: A deep knowledge of the relevant legislation is required Currently, only experienced employees are involved in the process
Departments and roles involved in the business process: The pension secretariat The department of pensions or the department of payments The director of the branch
Source material involved in the business process: The source material is mainly in the form of documents It includes standard administrative forms, legal documents, official documents, etc.
IKA case study (2/4)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 19
The business process has been analysed from the knowledge perspective
Basic business process tasks:
Submission of application form and supplementary documentation (pension folder)
Check of the pension folder
Processing of the pension folder
Establishment of entitlement right to receive a pension
Calculation of the pension amount
Notification of the insured person on the final decision
The DECOR Business Knowledge Method combines elements from CommonKADS
(knowledge-oriented process analysis) and IDEF5 (ontology modeling)
IKA case study (3/4)
Ontology refinement
Ontology creation
Task analysis
Business Process Analysis
Business Process
Identification
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 20
Expected benefits in the IKA pilot
Short term benefits: Workflow introduction and electronic documents smoothen process
enactment, ease communication, and decrease processing time Active knowledge delivery facilitates access to legal documents and
comments Weak workflow structures don‘t play a significant role in this case
study Long term benefits:
Access to older, similar cases facilitates comparable decisions Assessment of „similarity“ not trivial
Presentation of similar cases fosters communication with colleagues Exploitation of tacit knowledge as most important knowledge source
Attachment of personal notes to legal resources and backup material improves personal views on knowledge base
IKA case study (4/4)
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 21
Some words on related work Process-oriented structured archives
Multiple-viewpoint organization possible in many state-of-the-art document management tools (Autonomy, ...)
CognoVision® tool provides extremely expressive structures Method-embedding and maintenance support not available on market
Weak workflow structures Several research projects and very few commercial tools with similar goals => market
study under work Coupling with knowledge delivery unique for DECOR
Active, context-sensitive knowledge delivery Active push services are state-of-the-art in many commercial KM tools, but do not take into
account the actual work context Research approaches:
„Attentive systems“ (WATSON): try to deduce local work context from open applications and document content => ignore global task context
„EULE 2“ (Reimer et al): knowledge-rich, formal model of administrative workflow, rights, obligations, regulations, etc. => excellent services, but expensive knowledge engineering
Conclusions (1/2)
Use of such approaches for Electronic Government: not known to us
DFKI, ICCS, PLANET-EY, SEMA, DHC, IKA 22
Summary Project status: Three case studies have been analysed Archives are being installed Ontologies have been built in the first
instance Workflow is in the design phase
Next steps: Fill archives with operational data Document classification support Develop workflow formalism and engine
Transfer of the concept to a real-world E-Government scenario not yet done
Conclusions (2/2)