management of distributed knowledge sources for complex application domains
DESCRIPTION
This presentation focuses on knowledge management for complex application domains using Collaborative Multi-Expert-Systems. We explain how different knowledge sources can be described and organized in order to be used in collaborative knowledge-based systems. We present the docQuery system and the application domain travel medicine to exemplify the knowledge modularization and how the distributed knowledge sources can be dynamically accessed. Further on we present a set of properties for the classification of knowledge sources and in which way these properties can be assessed.TRANSCRIPT
Management of Distributed Knowledge Sources for Complex Application Domains
Meike Reichle, Kerstin Bach, Alexander Reichle-Schmehl and Klaus-Dieter Althoff
University of Hildesheim
{lastname}@iis.uni-hildesheim.de
2 of 27FGWM @ LWA‘2009 | 2009-09-23
Outline
• Motivation• Knowledge Modularization• Knowledge Map
• Classification of Knowledge Sources– Knowledge Source Properties
• Conclusion and Outlook
3 of 27FGWM @ LWA‘2009 | 2009-09-23
Motivation
• Knowledge-based systems deal with increasingly complex application domains
• Distributed, knowledge-based systems – Distributed knowledge processing– Distributed knowledge acquisition
• Realisation of distributed, knowledge-based systems using well-known AI techniques
4 of 27FGWM @ LWA‘2009 | 2009-09-23
The docQuery Project
• Travel Medicine– Prevention, management and research of travel
related medical aspects– Interdisciplinary: Requires expertise in other areas like
geography, activities, etc.
• Our main goal within the docQuery project– Provision of individualized and reliable information– On-demand query processing– Up-to-date information
5 of 27FGWM @ LWA‘2009 | 2009-09-23
SEASALT
• Sharing Experiences using an Agent-based System Architecture LayouT
• Instantiation of the CoMES (Collaborative Multi-Expert-Systems) approach
• Features– Application-independent architecture
– Knowledge acquisition from a web-community
– Knowledge modularisation,
– Agent-based knowledge maintenance
6 of 27FGWM @ LWA‘2009 | 2009-09-23
SE
AS
ALT
7 of 27FGWM @ LWA‘2009 | 2009-09-23
SE
AS
ALT
8 of 27FGWM @ LWA‘2009 | 2009-09-23
SE
AS
ALT
9 of 27FGWM @ LWA‘2009 | 2009-09-23
SE
AS
ALT
10 of 27FGWM @ LWA‘2009 | 2009-09-23
SE
AS
ALT
11 of 27FGWM @ LWA‘2009 | 2009-09-23
SE
AS
ALT
12 of 27FGWM @ LWA‘2009 | 2009-09-23
Knowledge Modularisation
• Knowledge Line (KL) within SEASALT– KL consists of complex knowledge in smaller,
reusable units (knowledge sources)
• Distribution of knowledge – Reflects structure of complex (interdisciplinary)
domains– Facilitates knowledge acquisition – Facilitates knowledge maintenance
FGWM @ LWA‘2009 | 2009-09-23
Knowledge Line
14 of 27FGWM @ LWA‘2009 | 2009-09-23
Knowledge Sources
• Topic Agents + external sources• Contain different kinds of information
– Multiple knowledge sources for the same purpose
• Knowledge sources are accessed dynamically– according to their properties
• Retrieval results (can) serve as input for a subsequent query
15 of 27FGWM @ LWA‘2009 | 2009-09-23
Knowledge Map: Motivation
• Term originates in Davenport’s and Prusak’s work on Working Knowledge1
• Organises all available knowledge sources– Who is the expert on a certain topic?
• Coordination Agent (Broker, Mediator)– Access to knowledge sources– Combines retrieved information– Uses Knowledge Map
1 Thomas H. Davenport and Laurence Prusak. Working Knowledge: How Organizations Manage What they Know. Harvard Business School Press, May 2000.
16 of 27FGWM @ LWA‘2009 | 2009-09-23
Knowledge Map: Definition I
• Knowledge Map KM consists of a number of Knowledge Sources KS:
• A Knowledge Source KS consists of a knowledge base KB and an interface I:
KM= {KS1 , KS2 , KS 3 , .. . KS n }
KS= {KB, I }
17 of 27FGWM @ LWA‘2009 | 2009-09-23
Knowledge Map: Definition II
• Dependencies between the Knowledge Sources– Input/Output dependencies enabling a subsequent retrieval
• Constraints on the Retrieval– Constraints over all Knowledge Sources
→ Availability, Costs, etc.
• Individual Retrieval Graph– Representing requested knowledge sources for an individual
query
18 of 27FGWM @ LWA‘2009 | 2009-09-23
Knowledge Map: Example
19 of 27FGWM @ LWA‘2009 | 2009-09-23
Computing Retrieval Graphs
• Computed based on – The information a user gives in an individual query– Pre-defined constraints– Knowledge Source dependencies
• A-priori computation of the retrieval path• Modified Dijkstra2 algorithm to determine an
optimal route over the graph2 Edsger W. Dijkstra. A note on two problems in connexion with graphs. NumerischeMathematik, 1:269–271, 1959.
20 of 27FGWM @ LWA‘2009 | 2009-09-23
Classification of Knowledge Sources
• Different properties referring to– Meta-information – Content
• Complex Knowledge Source properties– Compound properties
21 of 27FGWM @ LWA‘2009 | 2009-09-23
Meta-Properties• Access Limits:
– Number of requests per time unit, e.g. Projekt Deutscher Wortschatz3
• Format:– XML, HTML, data base tables, pure text, ...
• Syntax:– HTTP, SQL, agent, web service, ...
• Trust / Provenance: – Trustworthiness and reliability knowledge sources
3 http://wortschatz.uni-leipzig.de/
22 of 27FGWM @ LWA‘2009 | 2009-09-23
Content Properties
• Content: – Semantic description: What knowledge is provided?
• Coverage: – How good is the knowledge source’s topic covered?
• Completeness: – How complete is the information offered?
• Up-to-dateness• Expiry
23 of 27FGWM @ LWA‘2009 | 2009-09-23
Complex Knowledge Source Properties
• Complex properties– Compound properties as a (weighted) sum of
the presented simple properties
• Example: Quality– Comprises different aspectsQuality= 2× Coverage 2×Up−to−Dateness 2 × Answer Speed
24 of 27FGWM @ LWA‘2009 | 2009-09-23
Assessment of Knowledge Source Properties
• Automatically assessable properties– Speed, language and structure
• Manually maintained properties– Knowledge engineer assigns property values
• Relations between properties– Syntax, format, structure and cardinality are partially
related basic sanity checks of their assigned values
• Similarity-based reasoning
25 of 27FGWM @ LWA‘2009 | 2009-09-23
Values of Knowledge Source Properties
26 of 27FGWM @ LWA‘2009 | 2009-09-23
Conclusion
• Knowledge modularisation:
Knowledge Line approach in SEASALT• Focus on distributed knowledge acquisition
Dynamic access and assessment of distributed knowledge sources
• Retrieval over distributed knowledge sources
• Management of distributed knowledge sources
27 of 27FGWM @ LWA‘2009 | 2009-09-23
Outlook
• Retrieval Path computation– More flexible computation– Algorithm extension towards a more flexible and
subsequent result dependent routing– Automated integration of feedback about knowledge
sources
• Application and evaluation in docQuery
Thank you for your attention!
Questions | Suggestions | Comments