Adoption-Centric Knowledge Engineering

12
Computer Human Interaction & Software Engineering Lab Department of Computer Science, University of Victoria Adoption-Centric Knowledge Engineering AC KE Neil A. Ernst [email protected]

description

Presented at the Workshop on Adoption-Centric Software Engineering, ICSE 2003, Portland.

Transcript of Adoption-Centric Knowledge Engineering

Page 1: Adoption-Centric Knowledge Engineering

Computer Human Interaction & Software Engineering Lab

Department of Computer Science, University of Victoria

Adoption-Centric Knowledge Engineering

AC KE

Neil A. Ernst [email protected]

Page 2: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 2

Overview

•  Background

•  What is ACKE?

•  Our experiences: Jambalaya

•  Suggestions for creating user-centered knowledge tools

May 2003 CHISEL Research Group, University of Victoria 2

Overview

Background

ACKE

Jambalaya

Suggestions

Page 3: Adoption-Centric Knowledge Engineering

Neil A. Ernst, University of Victoria 3

•  KE refers to the creation of knowledge-based systems

•  Typical methodology: design, acquisition, entry, refinement

Background: knowledge engineering

Overview

Background

ACKE

Jambalaya

Suggestions

May 2003

Page 4: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 4

•  Most design occurs at what Allan Newell called the ‘Knowledge Level’

–  What exactly is being captured?

•  Multiple domain experts are sometimes necessary to explain the often complex subject areas

•  Some tools exist to simplify these steps

–  Help with modelling, acquisition,, and/or maintenance

•  Two chief user types:

–  End user (query, add, update)

–  Knowledge engineer (maintain, upgrade, model)

•  Similar to software engineer who maintains a legacy program

•  Difference between KE and SE: KE is maintaining an ontological commitment, not a tool

Background: knowledge engineering

Overview

Background

ACKE

Jambalaya

Suggestions

Page 5: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 5

Adoption-Centric Knowledge Engineering

•  Knowledge engineering (KE) has not had a strong end-user focus –  FOL oriented, mathematical syntax, research focus

–  Nevertheless, an increasing use of KE tools to develop applications

–  Semantic web initiatives increase this –  How can we make Semantic Web tools as simple as

early HTML tools were? –  Doing more complex things, so the feedback cycle is

slower, and the barrier to entry is higher

•  Move to leveraging existing cognitive support users have

•  Develop tools and processes with a human-centered focus

•  E.g. Rich Site Syndication (RSS) standard

Overview

Background

ACKE

Jambalaya

Suggestions

Page 6: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 6

Jambalaya

•  project: implementing information visualization in Protégé

–  Protégé is a popular knowledge-based system used to create and manage ontologies (specifications of concepts in a domain)

–  Jambalaya provides alternate views and tools to explore, understand, and

interact with these complex datasets

•  goals

–  know there is a problem with current tools (such as navigation and editing

problems)

–  our theory: visualization is an essential cognitive aid for conceptualizing a domain model and communicating that model to others

–  examine issues in user adoption of cognitive aids

•  how can an adoption-centric knowledge engineering focus help us?

–  conduct user studies for theory verification and generation

Overview

Background

ACKE

Jambalaya

Suggestions

Page 7: Adoption-Centric Knowledge Engineering

Neil A. Ernst, University of Victoria 7

Jambalaya (2)

•  demonstration: a research knowledge base,

Shrimpbib

•  current work

–  Initial approach: a graph visualization in Protégé would be useful!

–  Problem: convince real-world users of this

–  Refinement: ethnographic studies of this ‘real-world’

•  Surveys – large numbers of domains and scopes

•  Interviews - Do they need our tool?

–  How can we get people to adopt the tool?

Overview

Background

ACKE

Jambalaya

Suggestions

May 2003

Page 8: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 8

Page 9: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 9

Page 10: Adoption-Centric Knowledge Engineering

Neil A. Ernst, University of Victoria 10

ACKE: suggested approaches

•  Leverage existing tools such as Protégé

–  Reasonably large userbase, 8000+ registered

–  Extensible, open-source, cross-platform

–  What about different representation formalisms?

(FOL, frames, Description Logic)

•  Recall Shaw: “90% of code goes to UI, 10% to function”

•  What practices are currently used? How can WE adapt to them? (not, “here's a neat tool”)

•  Work on tool interoperability as well e.g. common exchange mechanisms (KIF, RDF, OWL)

Overview

Background

ACKE

Jambalaya

Suggestions

May 2003

Page 11: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 11

ACKE: suggested approaches

•  Support common tools

–  What are these tools?

•  Obvious ones: Office, Email

–  Eg. SemTalk (semtalk.com)

•  Web-centric tools

–  SVG or Flash

–  XML data interchange (GXL, GraphXML)

•  Custom applications: learn through qualitative

analysis on case by case basis (no one solution)

•  Aim to support the most with the least?

Overview

Background

ACKE

Jambalaya

Suggestions

Page 12: Adoption-Centric Knowledge Engineering

May 2003 Neil A. Ernst, University of Victoria 12

Questions?