2. 22
Agenda
• Key global climate and development challenges
• User needs and barriers to information use
• From taxonomies to ontologies and knowledge graphs
– How can AI technologies help?
• Example cases
– Technical uses cases
– A starting point - PLACARD Connectivity Hub
– How Knowledge Graphs support REEEP´s electrification programme
in Zambia
• Next steps for developing a Climate Action Knowledge Graph
Sukaina
3. 33
The PLACARD project
PLAtform for Climate Adaptation and Risk reDuction
https://www.placard-network.eu/
Transforming knowledge management for climate action:
A road map for accelerated discovery and learning
https://www.placard-network.eu/transforming-knowledge-management-for-climate-action/
Sukaina
5. 55
• Voluminous data - difficult to explore,
organize and analyse vast amounts of data,
particularly if there is a lack of structure or a
common format or standard.
• Fragmentation of information - Knowledge is
scattered across multiple platforms, data
portals, and websites, though objectives may
be similar.
• Disparate terminologies -
Communities each have their
own terminologies which are
often discordant.
– They often use different
terms to mean the same or
similar things.
– Results in inconsistency in
the way related content is
described or understood.
Sukaina
6. What do users want?
….too little
coordination and
planners, coordinators,
etc have too little time
to keep on top of all
initiatives. Need a
quick overview of who
is doing what
● enhanced discovery and searchability, to quickly find
related content and to filter and cluster search results
according to certain attributes;
● fewer entry points between regional, national and
international platforms so content is discoverable
from many platforms rather than searching them
individually;
● dynamic, responsive systems that help them find
relevant knowledge, e.g. through automated alerts of
new, useful content, user help desks, and expert
request services.
● clarity on language / terminology to better
understand how language is used differently and the
implications this might have.
We need clarity
on language (not
a common
language)
Sukaina
7. 77
Taxonomies as an IKM tool
Taxonomies:
● Are structured set of terms
that together describe a topic
area
● Provide an overview of the
vocabulary used in that
subject area, and how terms
related to each other
● Can include metadata, e.g.
definitions, related terms,
notes on term usage (scope
notes) and how this has
changed
Julia
8. 88
Taxonomies as an IKM tool
Term / Concept
Definition
→ supports understanding
Tagged content
Related Terms / Concepts
→ useful connections
Tagged content
Julia
15. 1515
User-agnostic and context-free data models
Do machines understand user intent? Do they have enough context?
Which Sport-utility vehicle from
France provides enough space
for my family with 3 kids?
The Kia Sportage is our
sporty SUV packed with
smart features
The Peugeot 5008
breaks new
ground as a large
SUV with many
features.
Car Loadspace Max no. of seats
KIA Sorento 550 litres 7
Peugeot 5008 823 litres 7
BMW X3 550 litres 5
1. Intent recognition
2. Entity linking
3. Background knowledge
French SUV
Martin
16. 1616
Lack of Common Knowledge
Artificial
“Intelligence” Perth
Australia
Perth is one of
the most isolated
major cities in the
world, with a
population of
2,022,044 living
in Greater Perth.
Australia is a
member of the
OECD, United
Nations, G20,
ANZUS, and the
World Trade
Organisation.
Country
City
is a
is a
is located in
Avoid illogical answers:distance between
Commonwealth
of Nations
International
Organisation
is part of
is a
Support complex Q&A:
Which cities located in the
Commonwealth of Nations
have a population of more
than 2 mio. people?
Background
knowledge
is key
Martin
17. 1717
Machine Learning per Data Silo
Machine Learning
per Data Silo
Lack of AI Strategy
Monte Carlo TS
Deep Learning
Deep Learning
Genetic
Algorithms
Neuronal
networks
Case based
reasoning
Martin
18. 1818
Desktop Data Integration
Desktop
Data Integration
Who is CEO of a Bank,
headquartered in Europe that
generates revenue per employee
higher than 400,000 Euro?
Head-
quarter
Ticker
Symbol
Revenue
Credit
Suisse
Zurich VTX: CSGN CHF 23.4b
HSBC London LON: HSBA USD 60.0b
Allianz Munich ETR: ALV EUR 122.3b
Deutsche
Bank
Frankfurt ETR: DBK EUR 33.5b
<Employees>
<VTX:CSGN>
48,200
</VTX:CSGN>
<LON:HSBA>
266,
273
</LON:HSBA>
<ETR: ALV>
147,
425
</ETR: ALV>
<ETR: DBK>
101,
104
</ETR: DBK>
</Employees>
The task awaiting Tidjane
Thiam when he takes over
from Brady Dougan as the
new chief executive at
Credit Suisse Group AG is
clear: how to pull the
Swiss bank out of a post-
financial crisis rut.
Knowledge workers
spend a large part of
their working time on
ad hoc data
integration tasks, the
so-called "research".
Which parts of it
could be automated?
Martin
19. 1919
Many Challenges in Data Management
Summary:
many challenges in
data management
Head-
quarter
Ticker
Symbol
Umsatz
Credit
Suisse
Zürich VTX: CSGN CHF 23.4b
HSBC London LON: HSBA USD 60.0b
Allianz München ETR: ALV EUR 122.3b
Deutsche
Bank
Frankfurt ETR: DBK EUR 33.5b
<Employees>
<VTX:CSGN>
48,200
</VTX:CSGN>
<LON:HSBA>
266,
273
</LON:HSBA>
<ETR: ALV>
147,
425
</ETR: ALV>
<ETR: DBK>
101,
104
</ETR: DBK>
</Employees>
The task awaiting Tidjane
Thiam when he takes over
from Brady Dougan as the
new chief executive at
Credit Suisse Group AG is
clear: how to pull the
Swiss bank out of a post-
financial crisis rut.
▸ Heterogeneous data
▸ Implicit semantics
▸ No background knowledge
▸ Data quality difficult to measure
▸ Proprietary schemas
▸ Unstructured data
▸ Ambiguity
▸ Multilingualism
▸ Various units and currencies
→ From Search to QA engines
Martin
20. 2020
FAIR Data Principles
A 4-layered Information Architecture
▸ Findability: Data and supplementary
materials have sufficiently rich
metadata and a unique and persistent
identifier.
▸ Accessibility: Metadata and data are
understandable to humans and
machines. Data is deposited in a
trusted repository
▸ Interoperability: Metadata use a
formal, accessible, shared, and broadly
applicable language for knowledge
representation.
▸ Reusability: Data and collections have
a clear usage license and provide
accurate information on provenance.
Martin
FAIR Data Principles: https://www.go-fair.org/fair-principles/
21. 2121
What is a Knowledge Graph?
From a Knowledge
Engineer’s perspective
A Knowledge Graph is a
model of a knowledge
domain created by subject-
matter experts with the help
of intelligent machine
learning algorithms.
From a Data Architect’s
perspective
Structured as an additional
virtual data layer, the KG
lies on top of existing
databases or data sets to
link all your data together
at scale – be it structured or
unstructured.
From a Data Engineer’s
perspective
It provides a structure and
common interface for all of
your data and enables the
creation of smart
multilateral relations
throughout your databases.
21
Martin
What is a Knowledge Graph: https://www.poolparty.biz/what-is-a-knowledge-graph
23. 2323
WHO uses Knowledge Graphs?
Semantic Search Index
Now I can find all the documents in
one place, and get assistance with it.
Metadata harmonization
Semantic tagging
Entity linking
Knowledge Graphs for Information
RetrievalSemantic tagging, query expansion, faceted search, classification, similarity-based recommender
Martin
24. 2424
Knowledge Graphs for KM Systems
Personalization, recommender, matchmaking, push services, smart assistants
Martin
25. 2525
Knowledge Graphs enhance KM
▸ KGs make implicit knowledge in
people’s heads/organizations
explicit.
▸ It is not the iceberg under the
water anymore, it is a hidden
treasure that can be lifted.
▸ Implicit knowledge is finally
integrated into the flow of work
and knowledge.
Martin
26. 2626
Summary: Core Principles
Summarization—core principles
▸ It’s all about things, not strings.
▸ Metadata should comply with FAIR principles.
▸ Data warehouses and data lakes are no longer
state-of-the-art paradigms of data integration,
but a data fabric will ultimately help
dismantle data silos.
▸ Use established standards and methods for
knowledge organization.
▸ Ambiguous data is often a burden on data
management. Adding more contextual
information is the key to solving this problem.
▸ Knowledge graphs are regularly confused with
a methodology for knowledge visualization.
▸ Knowledge management often strives to
design systems in which knowledge sharing on
a large scale becomes possible.
▸ Only an explainable AI creates trust.
FAIR + HITL = XAI
Martin
27. 2727
Using Knowledge Graphs to support
REEEP´s electrification programme
Content
• REEEP Background
• Beyond the Grid Fund (BGF) and
EDISON
• Climate Action Knowledge Graph to
unlock Opportunities in Data for
Development - an example from rural
electrification programme
Denise Recheis
28. 2828
REEEP - Renewable Energy and Energy
Efficiency Partnership
• International multilateral
partnership based in Vienna
• Accelerate market readiness for
renewable energy and energy
efficiency
• Founded 2002 at Johannesburg
Sustainability Summit
• Managed funds for more than
200 RE/EE projects worldwide
• Developed IT solutions and
services for knowledge
management in climate and
energy space
Denise Recheis
29. 2929
BEYOND THE GRID FUND – ZAMBIA
• BGFZ: Swedish Energy Access Programme | Initial
commitment USD 23m | Managed by REEEP
• Targets 1 million Zambians in rural and peri-urban areas
• Incentive scheme - address start-up and scale-up needs of
Energy Service Providers (ESPs)
• Market creation - public sector funding to overcome early
structural challenges in the market
• Procurement tools - incentivize innovation; comply with
customer rights and quality standards
• Builds investor confidence to mobilize downstream investment
• The BGFZ model is currently expanded to four additional
countries - BGFA
Denise Recheis
30. 3030
BGFZ | CHALLENGES
Results-based-financing Procurement – how to
reliably and practically verify the enduring
presence of 100,000s of energy services over
multiple years?
Standard RBF verification methods are
impractical / unreliable at scale:
• Physical checks of paper receipts
• Flat file exports and records
• Calling beneficiaries individually
• On-site inspection of results
Denise Recheis
31. 3131
Using Knowledge Graphs for intelligent
Analyses as part of EDISON 2
EDISON (Energy Data Intelligence System for Off-grid Networks) is a
web-based tracking tool which has been built to automatically collect,
manage and analyse programme data in a region where reliable data
is often not available and evidence-based decision making is difficult.
EDISON 2 concept
Denise Recheis
33. 3333
PROBLEM STATEMENT
Requirements of BGFA addressed with EDISON 2 include ongoing
analysis from many distributed sources, mixed formats and
difficult- to-interpret data.
• An analyst must typically spend considerable time manually exploring
various potential sources of information to uncover what exists
• Even when it is clear what information and data is available it can be
difficult to determine how data sets relate to each other, and thus time
consuming and tedious to join relevant data points.
• Unclear data e.g. deciphering often vague addresses of beneficiaries
These issues can be addressed and improved with knowledge
graph (KG) technology
Denise Recheis
34. 3434
Climate Action Knowledge Graph - an
opportunity for EDISON 2
• Geographic taxonomy including selected (Zambian) provinces, towns
and villages incl colloquial names, different spellings, local languages etc.
• External concepts from useful data sources to understand location
better
• Relationship ontology (electrification rate, census data, languages
spoken, solar radiation, postal code where available…)
• Useful keywords/tags for BGFA context for better filtering and searching
• Create a corpus of documents to improve automation of analyses
(natural language processing)
• Ongoing expansion of the KG, e.g.
– Regional info on default rate, sales, gender distribution of
customers, energy consumption through the programme per region,
village etc
– Specify currently often vague addresses with latitude / longitude,
postcodes
Denise Recheis
35. 3535
Pilot software Application for Zambia
Using a KG for a number of initial tasks related to EDISON analyses:
• Understanding more about each location, e.g. which village is in what
province, how close they are to each other and the distance to the
next larger town (remote/urban areas)
• Providing additional information such as languages spoken, rate of
energy access, national grid proximity, vicinity to important
landmarks, market saturation
• Knowing about local and regional policies such as available feed-in
tariffs
• Understanding what other organisations (NGOs, development
agencies) and companies are active in the region
• Fast analysis of resources (news, reports, project outcome
documents)
Denise Recheis
36. 3636
Knowledge Graphs for Climate Action
• Collaborating with PLACARD roadmap for IKM
transformation
• Connecting this pilot with the wider concept of the Climate
Action graph
• Complement other use cases for KM with a pilot for a KG to
support on-the-ground work
• Sharing useful ontologies for re-use in the sector as well
growing the larger Climate Action KG with specific data
• Conceptualizing on the groundwork for a linked web giving
better access to the wealth of information in the climate
sector
Denise Recheis
37. 3737
Further resources for developing a
Climate Action Knowledge Graph
Positioning paper: Creating a global Climate Action Knowledge Graph to leverage
artificial intelligence and speed progress on the climate agenda
Roadmap: Transforming knowledge management for climate action
Connectivity Hub: http://connectivity-hub.placard-network.eu
Taxonomy: PLACARD CCA and DRR Taxonomy
Short article: https://reeep.org/news/expediting-climate-change-action-through-
knowledge-graphs
Cookbook: https://www.poolparty.biz/the-knowledge-graph-cookbook
38. 3838
Thank you - any questions?
Martin Kaltenböck, Co-Founder and Managing Director at
Semantic Web Company
https://www.linkedin.com/in/martinkaltenboeck/
Semantic Web Company, https://www.semantic-web.com
PoolParty Semantic Suite, https://www.poolparty.biz
Denise Recheis, Knowledge and taxonomy manager at REEEP
REEEP www.reeep.org
Climate Tagger www.climatetagger.net
Sukaina Bharwani, Senior researcher and weADAPT Coordinator,
SEI, Oxford, UK
https://www.sei.org/people/sukaina-bharwani/
Julia Barrott, Researcher and weADAPT Knowledge Manager,
SEI, Oxford, UK
https://www.sei.org/people/julia-barrott/