Tatiana Baquero Cakici, Senior KM Consultant, and Jennifer Doughty, Senior Solution Consultant from Enterprise Knowledge’s Data and Information Management (DIME) Division presented at the Taxonomy Boot Camp (KMWorld 2022) on November 17, 2022. KMWorld is the world’s leading knowledge management event that takes place every year in Washington, DC.
Their presentation “Climbing the Ontology Mountain to Achieve a Successful Knowledge Graph” focused on how ontologies have gained momentum as a strong foundation for resolving business challenges through semantic search solutions, recommendation engines, and AI strategies. Cakici and Doughty explained that taxonomists are now faced with the challenge of gaining knowledge and experience in designing and documenting complex solutions that involve the integration of taxonomies, ontologies, and knowledge graphs. They also emphasized that taxonomists are well poised to learn how to design user-centric ontologies, analyze and map data from various systems, and understand the technological architecture of knowledge graph solutions. After describing the key roles and responsibilities needed for a team to successfully implement Knowledge Graph projects, Cakici and Doughty shared practical ontology design considerations and best practices based on their own experience. Lastly, Cakici and Doughty reviewed the most common use cases for knowledge graphs and presented real world applications through a case study that illustrated ontology design and the value of knowledge graphs.
Climbing the Ontology Mountain to Achieve a Successful Knowledge Graph
1. Climbing Ontology Mountain to Achieve
a Successful Knowledge Graph
Taxonomy Boot Camp 2022
November 7, 2022
2. Agenda
Federal
The Value of Knowledge
Graphs
1
2
Key Roles for Knowledge
Graph Projects
3 Ontology Design Approach
4
Knowledge Graph Case
Studies
3. ENTERPRISE KNOWLEDGE
10 AREAS OF EXPERTISE
KM STRATEGY & DESIGN
TAXONOMY & ONTOLOGY DESIGN
AGILE, DESIGN THINKING & FACILITATION
CONTENT & DATA STRATEGY
KNOWLEDGE GRAPHS, DATA MODELING, & AI
ENTERPRISE SEARCH
INTEGRATED CHANGE MANAGEMENT
ENTERPRISE LEARNING
CONTENT AND DATA MANAGEMENT
ENTERPRISE AI
Clients in 25+ Countries Across Multiple Industries
Meet Enterprise Knowledge
HEADQUARTERED IN
ARLINGTON, VIRGINIA,
USA
GLOBAL OFFICE IN BRUSSELS,
BELGIUM
Top Implementer of Leading Knowledge
and Data Management Tools
400+ Thought Leadership
Pieces Published
Jenni Doughty
Senior Consultant, EK
Tatiana Cakici
Senior Consultant, EK
5. FOLKSONOMY CONTROLLED
LIST
TAXONOMY ONTOLOGY KNOWLEDGE
GRAPH
ARTIFICIAL
INTELLIGENCE
Free-text tags. List of predefined
terms. Improves
consistency.
Predefined terms &
synonyms.
Hierarchical
relationships.
Improves
consistency. Allows
for parent/child
content
relationships.
Predefined classes
& properties.
Expanded
relationships types.
Increased
expressiveness.
Semantics.
Inference.
Capture related
data. Integration of
structured and
unstructured
information. Linked
data store.
Architecture and
data models to
enable machine
learning and other
AI capabilities.
Drive efficient and
intelligent data and
information
management
solutions.
@EKCONSULTING
6. Taxonomy Ontology
● What content covers
certain concepts?
● What is a more
specific/general version
of the concept?
● What are related pieces
of content based on
shared concepts?
● What are other names
for the same concept?
Types of questions we
can answer:
● Who wrote book A?
● Which books were published by Publisher X?
● Which books were published after 1995 by
authors from the UK?
● Which author worked with the most
publishers?
Types of questions we
can answer:
@EKCONSULTING
8. Business Questions Knowledge Graphs Answer
DATA FINDABILITY FOUNDATIONS FOR AI
Can users find the right
information at the right
time?
Does your organization
need to unify data silos to
capitalize on the
relationships between
organizational data
resources?
Is your data organized to
support the cutting-edge AI
and cognitive computing
solutions that will maintain
your organization’s
competitive edge?
DATA GOVERNANCE
Do data resources make it
clear to users what
information they contain?
Do current data procedures
support your organization’s
business success?
DATA AGILITY AND
SCALABILITY
Does your organization need
more flexibility from its data
architecture to rapidly iterate
and grow new products and
services for its users?
Do new use cases, legacy data
models, and the scale of the
data ecosystem cause delays
and challenges?
@EKCONSULTING
9. ENTERPRISE KNOWLEDGE
Semantic Capabilities
Personalization &
Insights
NLP Applications
Identification of Risks &
Opportunities
Recommendation Logic
Data/Content Aggregation
Reasoning
Disambiguation
Reporting & Decision-Making
Entity Recognition
Inferencing
Auto-tagging
Querying
Query Expansion (Stemming & Synonyms)
Discovery, Standardization &
Quality Control
Search within Results
Spell Checker
Type Ahead
Browsing and
Navigation
Sort Results
Facet/Filter Selection
Hierarchy Display
Taxonomy
Knowledge
Graph
Taxonomy
Ontology
Modeling
Solution
Functionality Use Case Business Value
Semantic
Formalization
&
Expressivity
Informs
Development
&
Maintenance
@EKCONSULTING
10. Knowledge Graph Applications
Recommender Systems
Data Management &
Quality
Auto-tagging
Taxonomy & Ontology
Development
Standardization and
Dereferencing
Natural Language and
Semantic Search
Data Visualization and
Reporting Dashboard
Data Governance
@EKCONSULTING
12. Key Roles for Knowledge Graph Projects
Core
Technical
Team
Business
Team
Ontologist
Designs the ontology,
taking use cases and
inferencing needs into
account
Information Analyst
Maps the ontology to
existing data sources,
determining which fields
in a source “match” to
which properties, classes
in the ontology
Semantic
Developer
Transforms data in
various source systems
to generate a semantic
knowledge graph
System Admin/IT
Professional
Installs and maintains
software resources (e.g.
ontology management
tool, graph database)
Subject Matter
Expert
Understands the
domain being modeled
and can validate
ontology design and
knowledge graph data
Business
Stakeholder
Defines the goals of a
knowledge graph
project, prioritizes
knowledge graph use
cases
Product Manager
Defines the knowledge
graph as a product and
ensures it is well-scoped
and managed
@EKCONSULTING
13. ● Ability to design simple
and complex ontology
solutions that may involve
integration of taxonomies,
ontologies, and knowledge
graphs
● Good understanding of key
semantic web standards
like RDF, OWL, and SKOS
● Model and document
ontologies for priority use
cases using various types of
semantic tools for ontology
management
Ontologists
● Good understanding of
foundational principles and
common applications of
taxonomies, ontologies,
and semantics
● Ability to analyze content
and data sources to
discover core components
and relationships
● Make sense of large
quantities of data and help
uncover unexpected data
connections
● Identify and document
ontology and knowledge
graphs use cases and
requirements
Information
Analysts
● Lead and support the
technical implementation
of semantic solutions
● Leverage common
taxonomy/ontology
management tools and
graph databases.
● Create and work with RDF
graph data, including
semantic inference,
structured and
unstructured data, auto-
tagging, SPARQL, SHACL
validation, and graph
machine learning
techniques
Semantic
Developers
Skills Required from Core Technical Team Roles
@EKCONSULTING
15. ONTOLOGY DESIGN
Not Agile Approach
Wait until the ontology is almost complete to share it with the user.
Agile Approach
Involve the users from the initial use case definition and gather feedback throughout the design process.
@EKCONSULTING
16. Involve the users from the beginning and gather feedback throughout the process.
VISION and
PLANNING
ANALYSIS DESIGN VALIDATION IMPLEMENTATION
Ontology Projects Approach
@EKCONSULTING
17. Vision and Planning
1. Define Use
Cases
2. Identify
Business Value
3. Develop User
Personas
SALES CUSTOMER
ACCOUNT
MANAGER
INTERNAL
SUPPORT
Semantic
Search
Chatbots Content
Recommendation
Entity
Resolution
@EKCONSULTING
19. Design
Sketch it out
Get a mental picture of how things are connected
Potential Tools:
● A whiteboard
● LucidChart
● Microsoft Visio
● PowerPoint
● gra.fo
Formalize in RDF
Assign official labels, URIs, properties, cardinalities, etc.
Potential Tools:
● gra.fo
● PoolParty
● Protégé
● Semaphore (Smartlogic)
● Synaptica
● TopBraid EDG
@EKCONSULTING
20. Let’s walk through design, Imagine that…
…we’re building an ontology for a large,
multinational retailer.
This retailer sells products, which are ordered by
customers and delivered by shippers.
How do we go about conceptualizing this ontology?
@EKCONSULTING
21. What are we trying to answer?
Who worked on project X?
Who can help me with topic
Y?
Who worked on project X?
What orders include Category X?
Product recommendations based
on Category Z?
Is there a Shipper trend for any
Product?
Step 1: Determine the questions we want to be able to answer
@EKCONSULTING
22. What are we trying to answer?
Step 2: Determine which classes are necessary to answer each question
Who worked on project X?
Who can help me with topic Y?
Product
Category
Shipper
Order
Who worked on project X?
What orders include Category X?
Product recommendations based
on Category Z?
Is there a Shipper trend for any
Product?
@EKCONSULTING
23. What are we trying to answer?
Who worked on
project X?
Who can help me with
topic Y?
Product
Category
Shipper
Order
Who worked on
project X?
What orders include
Category X?
Product
recommendations
based on Category Z?
Is there a Shipper trend
for any Product?
Supplier
Shipper
Product
Category
Customer
belongsToCategory
includedInOrder
Territory
managesTerritory
shippedByShipper
suppliesProduct
Employee
processedByEmployee
submitsOrder
Order
Step 3: Determine which relationships between
the classes are necessary to answer the questions
@EKCONSULTING
24. Validation
Perform a mix of techniques to validate your
model
● Sanity Check
● Sensitivity Check
● Data Fit Check
● Technical Check
● Best Practices Check
Potential Tools
Ontology Pitfall Scanner (OOPS) or similar open-
source tools can be used to check for:
● Missing type declarations
● Missing labels
● Missing domain/range
● Multiple domains/ranges
● Cyclical hierarchies
● Incorrect inverse properties
@EKCONSULTING
25. Implementation
Position the ontology so that it can
fulfill the use case(s).
Often, implementation of an ontology
involves the creation of a knowledge
graph.
Tooling Considerations:
● Ontology Management/Editors
● Governance Workflows and Controls
● Documentation
● Integrations or Consuming
Applications
@EKCONSULTING
26. Ontology Best Practices
Ontology Design Best Practices Ontology Implementation Best Practices
Identify a clear
use case
Specify expected
data-types for
attributes
Reuse
standards and
existing
vocabularies
Prioritize
relationships
Leverage
consistent
naming
conventions
Use singular nouns
for classes
Start small and
grow iteratively
Define &
document your
purpose
Plan for the long-
term
Focus on the
end user
Leverage
governance
Use simplest
language
possible
Look to usability
best practices
These best practices will help enhance the usability of the ontology.
However, these rules are slightly flexible – use your best judgement and keep business need centered. @EKCONSULTING
27. Design and Implementation Challenges
Complexity: Domains may be
complex, and thus developing an
ontology to describe them require
intensive research and validation.
Data & Technology: The data
contained in the legacy technology
may lack a clear organization scheme
or require additional transformations..
Understanding: Internal experts often
have conflicting ideas on the process
and about data intent or usage.
Scaling: Beyond a prototype.
Challenges
Linked Open Data Analysis: Analyze
existing ontologies available as linked
open data that may provide clarity and
understanding to a complex process.
Top-Down Analysis: To overcome the
lack of a clear organization scheme,
combine bottom-up analysis approach
with focus groups and validation
sessions.
Federation and Virtualization:
Present the ontology in numerous
ways to help communicate the
ontology design effectively, show it can
be used on real data, and build
consensus among subject matter
experts.
How we addressed them
@EKCONSULTING
29. .
THE CHALLENGE
THE SOLUTION
THE RESULTS
● We developed a cloud-hosted semantic course
recommendation service powered by a redesigned taxonomy
that was applied to a healthcare-oriented knowledge graph.
● EK extracted key terms and topics from the content in
order to rapidly build relationships between content
components.
● The recommendation engine was integrated with the
organization’s learning platform, successfully delivering
courses relevant to each user’s exam performance.
Personalized Course Recommendations
A healthcare workforce solutions provider:
● Had failed to consistently deliver relevant tailored course
content to healthcare professionals.
● Wanted to increase engagement and learning outcomes
across their learning platform.
● Wanted to deliver personalized content offerings to
connect users with the exact courses that would help them
master key competencies.
The recommendation service is
beating accuracy benchmarks
and replacing manual
processes, supporting higher-
quality, more advanced, and
targeted recommendations
that provide clear reasons why
the course was recommended
to the user.
@EKCONSULTING
30. Solutioning Challenge
Questions Courses
What is the
Question about?
What is the Course
about?
How are Courses related to Questions?
How are the Concepts
relevant to each other?
Healthcare
Professional
(Assessment)
@EKCONSULTING
33. Process of Generating Semantic Networks
Data Integration
Connecting existing data models
& concepts
Data Enrichment
Organizing & enhancing data via
extraction, tagging, &
classification
Data Creation
Adding new data concepts via
taxonomy development, data
entry, etc.
● Taxonomy and Ontology
● Questions
● Courses
● Competency Concepts
● Evaluation Methods
● Proficiency Level
● Extracting Topics from
Assessments for Taxonomy
Enrichment
● Tagging Questions
● Classifying Competency
Concepts
@EKCONSULTING
34. ENTERPRISE KNOWLEDGE
● Start with a small scope
● Involve SMEs each
knowledge domain
● Leverage ontology design
best practices
● Identify “gold standards” to
adjust the model along the
way
● Explore how the knowledge
graph can help with other
solutions in the future
Key Takeaways
@EKCONSULTING