Knowledge Technologies 2001 Siemens Automation and Drive Help Desk:

1

Knowledge Technologies 2001 Siemens Automation and Drive Help Desk:A Knowledge Work-Place with Self-Service

Norman Zimmerempolis NA, Inc.Burlington, MA

2

SIEMENS Automation & Drives

• Process Control Systems

• Machinery

3

Distributed Organization

4

Expert

Call-Center

Customers

Same sort of ProblemsSame expert

Same sort of Problemsmany different experts andagents

Lots of differentproblems and customers

1st -Level

2nd -Level

3rd -Level

Call-Center Pyramid

5

What is CBR ?

Experience is documented as a case.

A new problem is solved by adapting the solution of a stored case to the new situation.

Case-Based Reasoning (CBR) is a problem solving approach, that applies known solutions of past problems to solve new ones.

6

Examples

• A doctor remembers past patient records.• An advocat argues by precedence.• An architect reuses designs of existing buildings.• A sales agent explains a new product by referring

to satisfied customers.• A service technician remembers a similar defect from

another machinery.

7

? !Knowledge-Server Idea

QuestionsAnswers

KnowledgeServer

ContentBase

Knowledge is key to transform Data into InformationKnowledge is key to transform Data into Information

8

Motivation

Reuse experience to solve new problemsKnown examples utilize structured data in databasesbut in most cases there is a lot of existing unstructured information in free text form

Is it possible to apply the CBR paradigm to such text information?

9

In many areas knowledge is stored as weakly structured text:

Frequently Asked QuestionsDocumentationsManualsNotes and CommentsCustomer queriesProposalsand many more ...

Knowledge in Text

10

Documents contain a lot corporate knowledgeDocuments have specific characteristics:

restricted topicmostly free textpartly structured (chapters, section, ...)many documents address the same topic

Knowledge in Text

11

Example: FAQ

FAQ document

Hardware:PC & HP DeskJet 870

Software: Windows 95

Question:My new printer crops graphic print outs.

Answer:load and install new printer driver

12

Example: Dictionary

Computer

Down

PC

Machine

Sun

Crash

Storage InputWin3.1

13

Example: Ontology

Computer

Down

PC

Machine

Sun

Crash

Storage InputWin3.1

14

Example: Synonyms

Computer

Down

PC

Machine

Sun

Crash

Storage InputWin3.1

15

Example: Antonyms

Computer

Down

PC

Machine

Sun

Crash

Storage InputWin3.1

16

Example: Query

Q: On my PC the input of a long street name causes a crash. The error message is “Memoryfault”.

17

Example: Query

Q: On my PC the input of a long street name causes a crash. The error message is “Memoryfault”.

18

Example: Query and Results

Q: On my PC the input of a long street name causes a crash. The error message is “Memoryfault”.

F1: On Windows 3.1 there is not enough memory allocated for the name of the street. This may cause the system to go down.

F2: The PC-Version stores the street name incorrectly.F3: Typing German characters causes a Sun to crash.

19

Example: Query and Results

Q: On my PC the input of a long street name causes a crash. The error message is “Memoryfault”.

F1: On Windows 3.1 there is not enough memory allocated for the name of the street. This may cause the system to go down.

F2: The PC-Version stores the street name incorrectly.F3: Typing German characters causes a Sun to crash.

20

Analyzing Text

Create a dictionary of relevant termsCreate relations and similaritiesUtilize layers of knowledge:

Keywords: relevant common termsPhrases: application specific termsFeature Values: structured informationThesaurus: relations among keywordsGlossary: relations among phrasesDomain Structure: e.g. products Information Extraction: feature values from text

21

Prerequisites

Availability of appropriate documentsthe more the better (initially)extensible

Semi-automatic construction of dictionariesdatabases, other documents

Semi-automatic construction of the knowledge modeldatabases, existing glossaries

22

Ideal

Many documents electronically availableHTML, TXT, DOC, PDF, ...

Clearly distinguished topicsspecific application area

Documents correspond to cases1 Case = 1 Document1 Case = 1 Section in a document

Many userscustomers and technicians via WWWin-house teams via Intranet

23

Example: Document

Clear topic

sub-structure by productsspecific vocabulary

24

Maintain systemand publish

Knowledgeextraction

2nd levelproblem solving

1st levelproblem solving

Customer

Editor

ExpertExpert

Expert

Hotline

Administrator

SKM

solveproblem

problem

request

solveproblem

discuss

discuss

discuss

report report

document

FAQ

maintain

publish

Knowledge Capture Process

25

Text

26

“Ontology”

27

SIMATIC Knowledge manager

www.ad.siemens.de

orenge:Server

Structure

Informa-tion about SIMATICProduct structure

Products

Order no.

Product name

DictionaryInform-ationunits

Simil-arities

Knowledge model

Documents within the customerssupport information

system

SearchResults

Document view

28

Analysis of Queries

29

SIMATIC Knowledge Manager

Search in 20.000 FAQs

CD-Rom & Internetseamless integration

Online since 1998

German & English

FAQ Support

30

0

50

100

150

200

250

300

350

Ju

l-9

9

Au

g-9

9

Se

p-9

9

Oc

t-9

9

No

v-9

9

De

c-9

9

Ja

n-0

0

Fe

b-0

0

Ma

r-0

0

Ap

r-0

0

Ma

y-0

0

Ju

n-0

0

Call Avoidance = Savings

2.5 Million Dollar Savings in

2.5 Million Dollar Savings in

12 Months

12 MonthsSavin

gs in

Th

ou

san

ds

31

Maintain systemand publish

Knowledgeextraction

2nd levelproblem solving

1st levelproblem solving

Customer

Editor

ExpertExpert

Expert

Hotline

Administrator

SKM

solveproblem

problem

request

solveproblem

discuss

discuss

discuss

report report

document

FAQ

maintain

publish

Measurement

• Number of Calls• Time to Solve Problems• Amount of Knowledge• Coverage• User Satisfaction• Cost of Evolution

32

empolisBERTELSMANN MOHN MEDIA GROUP

Transforming Information into Value