Ontology-centric knowledge discovery in a Contact Centre for - - PowerPoint PPT Presentation

Ontology-centric knowledge discovery in a Contact Centre for Technical Product Contact Centre for Technical Product Support

Christopher JO Baker bakerc@unb.ca p @ Bradley Shoebottom bradley.shoebottom@innovatia.net Alex Kouznetzov alexk@unb.ca

Knowledge Discovery: Contact Centre

• Business Challenges:

Lengthy diagnosis phase / insufficient time for

– Lengthy diagnosis phase / insufficient time for

troubleshooting

• Technical support teams spend 25 to 50%
• f time

hi f searching for answers

• Unlinked information in knowledge-base silos and

heterogeneous formats

• Case escalation due to poor information find-ability

– Cases languish when Tier 2 (second level) not

available available

– OEMs outsourcing to low cost solution providers

drives Contact Centres to be more productive.

2

Semantic Knowledge Discovery: Contact Centre

• OWL-DL
• Custom Telecom Gazetteers
• GATE
• OWL API
• Pellet Reasoner
• TopBraid Composer

3

• TopBraid Live/ Ensemble

Semantic Solution: Visual Query over Telecom KB

Visual Query: Network Routing Server has a Configuring and Enabling Procedure TopBraid Ensemble

4

Telecommunications Ontology

• Describes phone routing software

Based on OWL DL (OWL 2)

High Level Ontology

• Based on OWL-DL (OWL-2)

– Classes: 506 – Instances: 12,000+ – Data Properties: 47 – Obj ect Properties: 167 – Class Equivalencies: 37 – Class Equivalencies: 37 – Class Disj unctions: 34 – S

ubclass Axioms: 37 I Obj 50

– Inverse Obj ects: 50 – Description logic: ALCHI(D) – Depth: 8 classes

5

Technical Support Contact Center FAQs

• What are the product software error codes?

E g ADM0234

– E.g. ADM0234

• What are the problem symptoms?

– E g Unable to call 911 – E.g. Unable to call 911

• What are the possible causes for a problem

symptom? symptom?

– E.g. Mis-configured system settings

• What is the solution for a possible problem?

p p

– E.g. Reset Emergency S

ervices settings

• Where is the procedure for a solution?

6

p

Case Resolution Process

1 1 T f t Tier 1 Monitor CRM Queue Tier 1 Resolve Using KB Transfer to Tier 2 Product Specialist Annotate CRM Case If No Solution p Tier 2 Attempts to p Resolve CRM Annotate Create Share Point Annotate Case Create Solution Document

P d t B D t CR

Federated Access

KB=Knowledge Base DB=Data Base CRM=Customer Relationship Management DB OEM=Original Equipment

Handoff to OEM Technical

Product Bug, Document CR is OEM Responsibility

Point If No Solution

OEM=Original Equipment Manufacturer

7

Publications Manuals Wiki KB CR DB

Contact Center Environment

• Tier 1:

Information gathering/ validation

– Information gathering/ validation – Initial problem solving – Requires highly precise information

Requires highly precise information

– Needs simple-to-use user interface

• Tier 2:

– Problem escalation or information not

found

– Requires high information recall – Requires advanced search capabilities

8

Pilot Study: Query Types and Source Content

• S

earches involve up to 4 terms, links to granular literature metadata and data in diverse (un)- t t d f t structured formats.

Type of Query Existing Form query only CRM DB Bulletins Technical Publications Existing Knowledge Base Form query only S emantic Form query on all Existing Knowledge Base HTML PDF only* PDF only* S e a t c S

• lution
• que y o all
• ntology entities

Pre-configured visual query (F AQ) S emantic S

• lution

HTML Word* 2 kinds of XML, FrameMaker* Ad hoc visual query

*Unstructured

9

Pilot Test Design

• Phase 1: Ontology and Interface Usability

Test (completed)

f d

– Can users find answers to 5 common queries

• Tier 1: Form search, pre-built general visual query, pre-

built specific visual query

• Tier 2: Form S

earch, Create pre-built general visual query, pre-built specific visual query

• Phase 2: S

cenario Testing (ongoing)

• Phase 2: S

cenario Testing (ongoing)

– Role-play of interaction with customer to test

• Troubleshooting decision tree

g

• Did information retrieved address symptoms, provide

procedures for solutions (recall and precision)

• At what point did escalation occur and why

10

At what point did escalation occur and why

Testing Challenges

• S

election of enough Tier 1 and 2 with a limited supply of people (especially Tier 2) limited supply of people (especially Tier 2)

• Time need for training and testing
• Complexity of testing to ensure all search
• Complexity of testing to ensure all search

paradigms covered

• Extra time needed to gather baseline
• Extra time needed to gather baseline

metrics in old toolset since they did not exist

• Adoption of new software toolset had

Adoption of new software toolset had glitches

11

Pilot Study: Results

d Search ared to Old ge Compa % Chang Toolset

• Tier 1 found the right information with less need for escalation

–

Now able to find documents 90%

• f the time (old toolset 75%

)

• Tier 2 has more tasks and toolset features to learn – longer

12

Tier 2 has more tasks and toolset features to learn longer learning curve

Contact Center Performance Metrics

Metric Impact of Semantic Solution Tier Impact Utilization (Productive versus Less time in training/ mentoring 2 non-productive time) and more time solving cases First Call Resolution Information found the first time, 2 , less time spent in research Case Closed Timeframe (Total Decreased case duration due to 1 and 2 Case Closed e a e ( otal elapsed time) ec eased case du at o due to less time spent in research a d Filtration Rate (Escalation) [Linked to First Call Resolution] Less cases escalated to Tier 2 or to Manufacturer 2 [Linked to First Call Resolution] Manufacturer Revenue Model Move from a per person headcount/ per client to a per case handled and multi-client support 1 and 2

13

handled and multi client support model

Business Outlook in Contact Centre Vertical

• Contact Centres employ 18,000 in New

Brunswick Canada and provides over CDN$1 Brunswick, Canada and provides over CDN$1 billion to provincial economy

• S

emantic S

• lution:

S emantic S

• lution:

– Proj ected saving for Tier 1 is 26%

• f overall case

resolution cost

– Re-usable methodology applicable across multiple

telecommunications products Business driver in cost reduction platform

– Business driver in cost reduction, platform

customizations, professional services

14

Innovatia Research

• Funded by a CDN $4 million grant from the

Atlantic Innovation Fund of the Atlantic Atlantic Innovation Fund of the Atlantic Canada Opportunities Agency

• Research Focus

– S

ingle source content development and re-use S ingle source content development and re use

– S

emantic knowledge management

15

Acknowledgements

• Proj ect Team

Christopher JO Baker1: Primary Investigator

– Christopher JO Baker1: Primary Investigator – Bradley S

hoebottom2: Knowledge Engineer

– Alex Kouznetzov1: Text Mining Engineer

Alex Kouznetzov : Text Mining Engineer

– Michael Doyle2: Network Infrastructure

• Testing Team

g

– Karen Lewis2: Information Architect – Innovatia Technical support team

• Dearran Townes, Amanda Chase, Darrell Flynn, Gregg

Knight, Corey Harris, Andrew Madsen

1 UNBSJ 2 Innovatia

16

1 UNBSJ, 2 Innovatia