smart data webinar: a semantic solution for financial regulatory compliance
TRANSCRIPT
A Semantic Solution for Financial Regulatory Compliance
Dataversity Webinar 12 November 2015
Mike BennettHypercube Ltd. + EDM Council
Hypercube
Agenda• Regulatory requirements in finance
– Overview of risk data aggregation /BCBS239• Non disruptive use of reference ontology
– semantic querying of existing systems of record – RDA-compliant reporting (BCBS239) – Risk / compliance dashboards
• Use of R2RML compliant wrappers – create SQL queries from SPARQL queries
• How to extend conceptual FIBO: – Modeling guidelines for the required reference ontology
• Data strategy considerations – ETL versus Pass-through querying– When to stand up a separate triple store for data and when not to
A Question
• What is the sound of one hand, clapping?
Systemic Risk: What Happened?
Network of Financial Exposures
5
Financial exposure to counterparty
Network of Financial Exposures
Some of these are lookinga little shaky…
6
Network of Financial Exposures
POP!
Where does that leave the survivors?
7
Systemic Risk: What Happened?
• Firms took a long time to establish their exposures to endangered banks
• Data wasn’t the problem
• Knowledge was
BCBS239 RDA Principles1. Governance2. Data architecture and IT infrastructure3. Accuracy and Integrity4. Completeness5. Timeliness6. Adaptability7. Accuracy8. Comprehensiveness9. Clarity and usefulness10. Frequency11. Distribution12. Review13. Remedial actions and supervisory measures 14. Home/host cooperation
BCBS239 Themes
I Overarching governance and infrastructure
II Risk data aggregation capabilities
III Risk reporting practices
IV Supervisory review, tools and cooperation
Objectives of the RDA Principles• Enhance the infrastructure for reporting key information, particularly that used by
the board and senior management to identify, monitor and manage risks;
• Improve the decision-making process throughout the banking organisation;
• Enhance the management of information across legal entities, while facilitating a comprehensive assessment of risk exposures at the global consolidated level;
• Reduce the probability and severity of losses resulting from risk management weaknesses;
• Improve the speed at which information is available and hence decisions can be made; and
• Improve the organisation’s quality of strategic planning and the ability to manage the risk of new products and services.
EDM Council Observations
1. Affects not just G-SIBs2. RDA concepts understood the same by
everyone3. Common financial language4. Cultural not compliance objective5. States of book: internal v what is reported;
harmonizing these
Source: Mike Atkin, John Bottega, EDM Council Industry Webinar Oct 2014
Towards a Culture of Compliance
• Requires established governance: – Operation and controls.
• RDA defines the goal of what you mean by control environment.– The firm needs to take control of knowledge and
therefore of concepts– Data invited into the conversation along with
business and IT
Bringing Data Into the Conversation
• Control environment for applications and process is well known
– Data has been for too long the neglected sibling of applications
– We are seeing a move to a more data-centric world
– Financial markets are made of data
• Time for Data to step up!
From BCBS239 Principle 2
• “A bank should establish integrated data taxonomies and architecture across the banking group, which includes information on the characteristics of the data (metadata), as well as use of single identifiers and/or unified naming conventions for data including legal entities, counterparties, customers and accounts”
– Not a requirement for a single data model but a requirement for robust reconciliation among multiple models
BCBS239: A Trojan Horse for Effective Data Management?
The Zachman framework
18 Copyright © 2010 EDM Council Inc.
Model Positioning
Conceptual Model
Logical Model (PIM)
Physical Model (PSM)
Realise
Implement
19 Copyright © 2010 EDM Council Inc.
Model Positioning
Conceptual Model
Logical Model (PIM)
Physical Model (PSM)
Realise
Implement
The Language Interface
Business
Technology
Conceptual Model
• Model Formalism:– Any technology independent formalism– First Order / Higher Order Logic is good
• Model Theory:– Everything in the model should represent something in the
business domain
• Model application:– Provides technology-neutral view of some aspect of the
problem domain– Point of reference for solutions implementers
20 Copyright © 2010 EDM Council Inc.
FIBO History• Multi-year project sponsored by the Enterprise Data
Management Council• Initial draft material subjected to business subject matter
expert reviews• Reference data for principal instrument classes is in “Beta”
– This means it is stable enough to refer to but we expect changes as we come up against real data and real projects
– Proof of Concept projects and early adopters ongoing• Teamed up with OMG to submit a series of proposals for
formal FIBO standards• FIBO Foundations has been accepted by the OMG• Business Entities, Financial Instrument common terms, Indices
to follow
So…
• We have defined FIBO as a kind of conceptual ontology– What does a conceptual ontology “Do”?– Typically nothing – it’s a management tool
• Here we will show a way of using a conceptual ontology as a reference ontology in a particular technical application
• It still has to be conceptual
FIBO Proof of concept
• A US Globally Systemically Important Bank• Cambridge Semantics
• Focus on Derivatives / Swaps• Automated classification of IR Swaps etc.
Proof of Concept• Ontology editor:
– Loaded existing ontology (FIBO) – View, modify and extend FIBO.
• Tools for mapping and loading data from varied sources against the ontology into a graph store.
• For the PoC:– Loaded the FIBO swap model– Mapped data extracted from G-SIB’s system (in a
spreadsheet) onto the model – Loaded the data into Anzo– Ran classification rules on the swaps – Built dashboards to visualize the data.
Challenges
• Triple store and Data Management– Timeliness (how often to update)– Provenance– System of record– Data lineage
• What if we could use the ontology to report on the data in situ?
Reference Ontology
Solution Architecture with Triple Store
www.capsenta.com 26
Source 1 Ontology
Source 2 Ontology
Source N Ontology
Source DB 1 Source DB 2 Source DB N
…
Graph Triplestore
Reporting
Query ResponseET
L
ETL ET
L
Solution Architecture with R2RMLReporting
R2RML based Ontology to Legacy Database Adapters
Semantic Queries
Risk, Compliance etc.
Reference ontology
Legacy Data Sources and Systems
Source DB Q
Combined Architecture
www.capsenta.com 28
Source 1 Ontology
Reference Ontology
Source 2 Ontology
Source N Ontology
Source DB 1
Source DB 2
Source DB N
…
Reporting
Query Response
Graph Triplestore
ETL
Source P Ontology
Source DB P
QueryResponse
Reasoning Apps
Source Q Ontology
Data Lineage – Metadata management
Financial Institution with Golden Copy
29
Reference Ontology
Source DB 1
Golden Copy DB Other DB N
…
Source DB N Market Data Feed 1..n
Market Data Feed
…
Reporting
ResponseQuery
Observations
• Reference ontology is the key to semantics based reporting and querying
• Needs to capture the concepts in the data• Needs to disambiguate concepts across that data• Use classification – deep hierarchies, faceted
classification• Recognizing the meaning means thinking about
meaning – Not words. Not data. Not technology. Just meaning
Some issue around meaning of concepts in finance
• Faceted classification• Different contexts e.g. price• Recognizing the meanings• Principles for creating single, coherent set of
unambiguous concept– Unambiguous shared meaning
• Scope and coverage– What FIBO covers– What else you need for BCBS239 and other regulatory
History: Financial Standards
• Messaging: MDDL– XML schema for market data
• ISO 20022 FIBIM (ISO TC68/SC4)– Logical Data Model Design via UML profile
• FpML (ISDA)– Derivatives message models
• What the industry really needed
Financial Data Standards / MDDL
• Equity as Classes, Debt as faceted classification
• The Meaning of Price– Price in the market– Price of a transaction
• Meanings and Regulations
Investment Roadmap
• Based on…– “Investment Roadmap”
• September 2010
– As maintained by• FIX Protocol• FpML• ISITC• SIA/FISD• SWIFT• XBRL
Investment Roadmap – FIX, ISO, FpML, XBRL syntax (HIGH LEVEL)
(1) Represents ISO 20022 , ISO 15022 and MT messages(2) See OTC Derivatives breakout for details:
- Syndicated Loans, Privately Negotiated FX, and OTC Equity, Interest Rate, Credit, and Commodity Derivatives- FpML payload may be used in combination with FIX business processes in dealer to buy side communication
Function Cash Equities & Fixed Income Forex(2) Listed
DerivativesOTC
Derivatives(2) Funds
Issuer Pre-investment decision N/A N/A
Front OfficePre-Trade
Trade
Middle Office
Post-Trade
Clearing / Pre-Settlement
Back Office
Asset Servicing N/A
Collateral Management N/A N/A
Settlement
Pricing / Risk / Reporting
Investor Supervision Regulatory Reporting
Issuer Supervision Regulatory Reporting N/A N/A
FIX ISO (1)
FpML XBRL
ISO 20022 Common Business Model (HIGH LEVEL)
(1) Represents ISO 20022, ISO 15022 and MT messages(2) See OTC Derivatives breakout for details:
- Syndicated Loans, Privately Negotiated FX, and OTC Equity, Interest Rate, Credit, and Commodity Derivatives- FpML payload may be used in combination with FIX business processes in dealer to buy side communication
Function Cash Equities & Fixed Income Forex(2) Listed
DerivativesOTC
Derivatives(2) Funds
Issuer Pre-investment decision N/A N/A
Front OfficePre-Trade
Trade
Middle Office
Post-Trade
Clearing / Pre-Settlement
Back Office
Asset Servicing N/A
Collateral Management N/A N/A
Settlement
Pricing / Risk / Reporting
Investor Supervision Regulatory Reporting
Issuer Supervision Regulatory Reporting N/A N/A
FIX ISO (1)
FpML XBRL
FIBO Business Semantics (HIGH LEVEL)
(1) Represents ISO 20022, ISO 15022 and MT messages(2) See OTC Derivatives breakout for details:
- Syndicated Loans, Privately Negotiated FX, and OTC Equity, Interest Rate, Credit, and Commodity Derivatives- FpML payload may be used in combination with FIX business processes in dealer to buy side communication
Function Cash Equities & Fixed Income Forex(2) Listed
DerivativesOTC
Derivatives(2) Funds
Issuer Pre-investment decision N/A N/A
Front OfficePre-Trade
Trade
Middle Office
Post-Trade
Clearing / Pre-Settlement
Back Office
Asset Servicing N/A
Collateral Management N/A N/A
Settlement
Pricing / Risk / Reporting
Investor Supervision Regulatory Reporting
Issuer Supervision Regulatory Reporting N/A N/A
FIX ISO (1)
FpML XBRL
Meaning
Messaging
Common Data
Industry Conclusions• Good design is weak semantics• Business knowledge gained during reviews is either
– Lost– Buried in meeting minutes– Kept in uncontrolled spreadsheets in a variety of structures
• Data Dictionaries try to link business definitions to data elements – but data elements are reused across business meanings and usage
contexts (good design again)
• Industry conclusion– “We need a semantics standard”
Concepts• First we must recognize concepts.• Conceptualization is abstracting away from the sensory
stuff that makes up our world, into discrete and useful meaningful pieces.
• A concept has an ‘intension’ (a set of logical statements about what it means to be that kind of thing), and an ‘extension’ (the set of individual things in the world which match those statements). Optionally it has a name or label, which can be used to refer to it.
• Some situations in the world can be conceptualized in more than one way. But the concepts are what they are.
Not DesigningSome Stuff
DETECTION:What kind of Thing?
What distinguishes it?AbstractionsClassificationPartitioning
Ontology Representation:
How to model conceptsPatterns
Validation Reference Ontology
Classification and Abstraction
Thing
Classification and Abstraction
Thing
Red Thing Blue Thing
Differentiae: what distinguishes the sub types of the Thing
Classification and Abstraction
Thing
Round Thing
Square Thing
Differentiae: what distinguishes the sub types of the Thing
Faceted Classification
Thing
Round Thing
Square Thing
Red Thing Blue Thing
Round Red Thing
Round Blue Thing
etc.
Use of Partitions
• 1: Independent Relative and Mediating Things• 2: Continuant and Occurrent• 3: Concrete and Abstract
46
Recognizing the Philosophical Requirements
• Consider this dog:
What is a Pet?
• A dog is a thing in itself• A pet is defined in relation to some interaction
between the animal and some person - it is somebody’s pet
• Pet ownership is a kind of implied contract between some human(s), and some animal
Definitions• Independent Thing:
– Something that exists in itself, its essential meaning doe not depend in any way on being in a relationship with anything else . That is to say, defined by a set of immutable characteristics.
– e.g. rock, person– There is a property P(x)
• Relative Thing – Something whose essential meaning is determined by one or more
relationships it is in with at least one other thing – e.g. buyer, broker– There is a property R(x,y)
• Mediating thing – Something whose essential meaning derives from the fact that it brings two or
more things together in some way.– E.g. Trade, Agreement, Reified relationships– There is a property M(x,y,z)
What is a Pet?• A Dog has intrinsic properties that are not dependent
on context– There is a property P(x) where x is a dog
• A Pet has at least one property which relates to the interaction between some independent thing and something else– There is a property R(x,y) where x is a dog and y is a person
• Pet ownership has some property which relates to two or more things being brought together into some interaction– There is a property M(x,y,z) where x is pet ownership, y is a
dog and z is a person
Ontology Partitioning 1
51
Thing
Independent Thing
Relative Thing
Mediating Thing
“Thing in Itself”
• e.g. some Person
Thing in some context
• e.g. that person as an employee, as a customer, as a pilot…
Context in which the relative things are defined
• e.g. employment, sales, aviation
• Everything which may be defined falls into one of three categories:
Ontology Partitioning 2
Thing
Continuant Occurrent
• Continuant: where it exists, it exists in all its parts– Even if these
change over time
• Occurrent: the concept is only meaningful with reference to time
© Hypercube 2015 52
Ontology Partitioning 2Thing
Continuant
Person Contract Pilot
Occurrent
Event State Etc.
• Things which are independent or relative are also either continuant or occurrent
Ontology Partitioning 3
Thing
Concrete Abstract
• Concrete: A physical thing– Or a virtual thing in
some reality
• Abstract: the concept is only meaningful as an abstraction from reality
Example Concept Patterns
Transactions (REA Ontology)
56
Transaction Event
57
58
Transaction Event Undertakings
59
Example Detailed Instrument:Credit Default Swap
Scope for RDAR /BCBS239
Risk
• Basic Risk Formula:
RISK = PROBABILITY x IMPACT
Risk
• Basic Risk Formula:
RISK = PROBABILITY x IMPACT
• Probability of what? = EVENT
• Impact to what? = GOAL
Three Levels of Risk
• Institutional risk– Including credit risk, operational risk etc.
• Sub-system Risk– E.g. risk in a given market
• Systemic Risk– Risks to the entire financial system
Three Levels of Risk
• Institutional risk– Including credit risk, operational risk etc.
• Sub-system Risk– E.g. risk in a given market
• Systemic Risk– Risks to the entire financial system
Three Levels of Risk
• Institutional risk– Including credit risk, operational risk etc.
• Sub-system Risk– E.g. risk in a given market
• Systemic Risk– Risks to the entire financial system
Classification: Types of “Thing”
• Static terms– Reference data– Commitments in the
Contract– Embedded options– Business Entities / LEI
Classification: Types of “Thing”
• Static terms– Reference data– Commitments in the
Contract– Embedded options– Business Entities / LEI
• Temporal terms– Time to Maturity– Credit Ratings– Analytics
Classification: Types of “Thing”
• Static terms– Reference data– Commitments in the
Contract– Embedded options– Business Entities / LEI
• Temporal terms– Time to Maturity– Credit Ratings– Analytics
• Real-time terms– Pricing– Market rates– Valuation
Classification: Types of “Thing”
• Static terms– Reference data– Commitments in the
Contract– Embedded options– Business Entities / LEI
• Temporal terms– Time to Maturity– Credit Ratings– Analytics
• Real-time terms– Pricing– Market rates– Valuation
• Environment Factors– Behavior– Predictions
System and Sub-system Risk Factors
– Individual sub-system risk factors: volatility, speed, liquidity
• Factors on these: volume, access to price information– Price: timeliness, accuracy, access
• Sub-system specific factors e.g. property market, loans (recourse v non recourse, secured v unsecured; collateral valuation movements)
– Emergent systems: the things you need to measure are the links among the systems from which it emerges.
Implications for RDAR• Analysis of the risk requires that you introduce different
things into the ontologies.– For example the behaviors themselves, the organizational
groupings and so on.– Derived risk factors– Temporally sensitive concepts
• Ontology: the meanings of the variables– Regardless of their origin e.g. whether computed or from data
feed– Ontology of input factors to risk apps (and other apps)– Ontology of output factors form those apps, from data feeds
etc. (assertions)– Risk model assumptions
FIBO: Scope and ContentUpper Ontology
FIBO Foundations: High level abstractions
FIBO Contract Ontologies
FIBO Pricing and Analytics (time-sensitive concepts)Pricing, Yields, Analytics per instrument class
Future FIBO: Portfolios, Positions etc.Concepts relating to individual institutions, reporting requirements etc.
FIBO ProcessCorporate Actions, Securities Issuance and Securitization
Derivatives Loans, Mortgage Loans
Funds Rights and Warrants
FIBO Indices and Indicators
Securities (Common, Equities) Securities (Debt)
FIBO Business Entities FIBO Financial Business and Commerce
FIBO: StatusUpper Ontology
FIBO Foundations: High level abstractions
FIBO Contract Ontologies
FIBO Pricing and Analytics (time-sensitive concepts)Pricing, Yields, Analytics per instrument class
Future FIBO: Portfolios, Positions etc.Concepts relating to individual institutions, reporting requirements etc.
FIBO ProcessCorporate Actions, Securities Issuance and Securitization
Derivatives Loans, Mortgage Loans
Funds Rights and Warrants
Securities (Common, Equities) Securities (Debt)
Key OMG in process
OMG in preparation OMG Complete
Draft in Semantics Rep
FIBO Indices and IndicatorsFIBO Business Entities FIBO Financial Business
and Commerce
FIBO Where is What!• 29 FIBO Business Conceptual Ontologies have been built since 2008
• http://www.edmcouncil.org/semanticsrepository/index.html• Contains much detailed downloadable information including models, spreadsheets and XLS files
for 29 FIBOs• Github Working Wiki page”
• https://github.com/edmcouncil/fibo/wiki• For those who want to get serious soon – Links to UML and RDF/OWL downloadable files for all
29 FIBOs and much much more of Pink and Yellow and Green FIBOs• Browseable and searchable repository with workspaces for all ontologies
• http://us.adaptive.com/FIBO/a3/
• http://www.omg.org/spec/EDMC-FIBO/FND/Current• Contains FIBO-FND in final OMG documentation form including UML and RDF/OWL models for FIBO
Foundations• Github wiki is at:
• https://github.com/edmcouncil/fibo/wiki/FIBO-Foundations• http://www.omg.org/spec/EDMC-FIBO/BE/Current
• Contains FIBO-BE (Business Entities) In OMG documentation form. • Github wiki is at
• https://github.com/edmcouncil/fibo/wiki/FIBO-Business-Entities• A working version in testing (“David’s Branch”) is at
• https://github.com/dsnewman/fibo/tree/pink/be• http://www.omg.org/spec/EDMC-FIBO/IND/Current
• Contains FIBO-IND (Indices and Indicators) In OMG documentation form• Github wiki is at
• https://github.com/edmcouncil/fibo/wiki/FIBO-Indices-and-Indicators .• Pointer to Loans FIBO Github Wiki page
• https://github.com/edmcouncil/fibo/wiki/FIBO-Loans• Pointer to Securities and Equities FIBO Github wiki page
• https://github.com/edmcouncil/fibo/wiki/FIBO-Securities-and-Equities
• General Information - http://www.edmcouncil.org/financialbusiness• Historical perspective and status
The art of Not Designing• Conceptual Ontology is an exercise in detection not
creation• We do not ask “What does this word mean?” but,
“What would be a good word for this concept?” • There are no choices about what things mean• Requirements have no effect on meaning• Concepts are unaffected by whether or not you choose
to include them in an application• The choices are about which meanings to stand up in
an ontology• You start by must recognizing concepts
What is the Sound of One Hand Clapping?
• It is exactly the same as the sound of the other hand clapping
Take-aways
• A good reference ontology starts with the aptitude and motivation to consider these kinds of questions– Technology considerations come later– This may be a recruiting challenge– Also require that you ask the right questions of
domain experts• This gives you a conceptual model which can be
used with R2RML wrappers to directly query legacy data sources semantically
Thank You!
• Mike Bennett– [email protected]– [email protected]
– www.edmcouncil.org– http://www.edmcouncil.org/semanticsrepository/
index.html