Upload
trannga
View
219
Download
4
Embed Size (px)
Citation preview
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 1
Mastering Enterprise Metadata with Seman2c Modeling
Enterprise Metadata: The descrip4on of the organiza4onal context – processes, roles, policies, products and offerings, etc. – that are implicitly part of the enterprise informa4on ecosystem. Seman2c Modeling: A data model linked to the real world through a conceptual model
If we combine enterprise metadata with an enterprise seman4c web ini4a4ve, we can create a knowledge fabric that can completely change the way we think of enterprise soPware
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 2
Cust. ID Name Address Drivers License
The seman4c model is not just part of the metadata, it is the metadata
SID: ORCL
Seman2c Model: A data model linked to the real world through a conceptual model
Ted
A person in the real world, who just happens to be playing the role of the customer at this
point in 4me
We need a legal name for this person: A name registered with the government body within whose legal jurisdic4on we are
engaging with this person
An address is not just a couple of lines. Is it a residence? Does the customer have “in the town / in the country” homes? What sort of neighborhood is the house in?
The state that issued the license tells us the primary state of domicile of the person. The date of first issue tells us how long the person has been in the state
Data in & about a record also gives us technical context: which table in which database on
which server. When was this record created & by whom
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 3
Modern Enterprise IT Environment: Ecosystem of meta driven systems
Smart Devices Modern Modular UI’s
Business Process Management
Customer Rela4onship Management
Service Oriented Architecture
Business Rules Engine
Accoun4ng Pla_orm
Master & Reference Data Management
Message and Data Integra4on
Data marts and Warehouse
Analy4cs Ecosystems
Enterprise Resource Planning
Enterprise Content Management
Security & En4tlement
Local & Wide Area Networks
On demand Data Centers & Virtual
Desktops
Modern development pla_orms are also heavily meta data driven: JEE, Spring, .Net, Cococa & iOS, Android
One primary role of designers and developers today is as translators between the “dialects” that these systems speak
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 4
Case specific Process
Evolu2on of Metadata Driven Systems
R2RML
Parameter Driven Applica2on
Logic Parameters
Custom Extensions
Orchestra4on Rules
Parameters
Logic Orchestra4on Rules
Custom Extensions
Context
Metadata Driven Applica2on
Parameters Rule Fragments
Seman4c Model
Process Model
Enterprise Context
Logic Fragments
Dependency Tree
Parameters Orchestrator
Context Rule Fragments Logic Fragments
Current Generation Next Generation
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 5
Scenario Report
En2tlement
Swim Lane
Business Rule
Test Case
Involves
Realized as
Segmented into
In a Performed by
Belongs to
Requires
History analyzed through
Metrics Generates
Realized as
Requires / generates
Behavior driven by
Validated by
The business (problem) space has metadata too…
Data
Func2on
Use Case Process
Actor
Task
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 6
John Smith is a customer of the bank with a large investment porQolio. He is a cau4ous investor with a preference for energy stocks and commodity futures. PB banker is James M
Real world facts
<a human >
“John Smith”
Has name
<a por_olio account>
Operates
<a banker> [James M]
Is advised by
<a security holding> includes
<a security>
For Security
<a company> [Venture Solar]
<a corporate ac4on>
ini4ates Is no4fied
of
Issued By
<a banker> [Frank K]
Is advised by
plans
Public Informa4on
Works with
“Energy”
Ac4vity Area
plans
<a country> (Germany)
Tax Jurisdic4on
<a company> Benthik
Petroleum
Will become subsidiary of
<a country> (Syria)
Operates in
Has posi4ons in Headquartered in
<a state> [California]
<LOB> [Private Bank]
Works in
Works in
<LOB> [Corporate Advisory
Services]
John Smith holds posi2ons in Venture Solar, an energy company, currently headquartered in California, USA
John Smith currently works in Germany and is subject to German tax laws
Venture Solar is corporate customer. Frank K from corporate advisory services is currently helping them put together a reverse merger with Benthik Petroleum
Benthik Petroleum is authorized to operate in Syria
<a planned corporate ac4on>
Non public Informa4on
Taking this one step further… express actual data as graphs as well
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 7
Mobile Device
Process
“John Smith”
Has name
<a por_olio account>
Operates
Is advised by
includes
<a security>
For security
<a corporate ac4on>
ini4ates Is no4fied
of
Issued By
Is advised by
<a planned corporate ac4on>
plans Non public informa4on
Public Informa4on
Works with
“Energy”
Ac4vity Area
plans
<a country> (Germany)
Tax Jurisdic4on
Will become subsidiary of
<a country> (Syria)
Operates in
Has posi4ons in Headquartered in
<a state> [California]
<LOB> [Private Bank]
Works in
Works in
<LOB> [Corporate Advisory
Services]
Por_olio Review Advise ac4on on a posi4on
Task Part of
Personal Banker
Customer Service Team
Performed by
iPad 3
On device
Mobile Applica4on On app
Reference Pla_orm
Table
Db2 on Mainframe
Data Pla_orm
CIF Aqribute
Legal Name EU Prospect Policy
enforces
Table Column
Applies to
Prospect informa4on
cannot be held for more than 3
months
Customer Master
CIF
<a company> [Venture Solar]
<a banker> [Frank K]
<a banker> [James M]
<a human >
Customer Name
Advisor Desktop
<a security holding>
<a company> Benthik
Petroleum
If we put this all together …
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 8
We now know…
Which processes update data in table “X”?
• SELECT ?process WHERE { ?table rdf:type db:Table . ?app app:canUpdate ?table . ?process proc:hasTask ?task . ?task proc:performedOn ?app }
Which development teams query the customer master table?
• SELECT ?team WHERE { ?app app:canUpdate <CustomerMaster> . ?app app:maintainedBy ?team }
Which func4ons are impacted by an outage of the DB2 mainframe?
SELECT ?process ?func4on WHERE { { ?process prod:supports ?func4on . ?process prod:runsOn prod:DB2_Mainframe } UNION ?{ process prod:supports ?func4on . ?func4on prod:dependsOn prod:DB2_Mainframe } }
All the possible provenance alterna4ves for customer phone numbers
Which systems are impacted by this new EU policy?
Who are the possible downstream consumers of the field I am capturing on this screen?
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 9
And change the way we write apps...
Pick the right connec4on pool paqern based on table proper4es
Invoke the right valida4ons based on the region of the customer and region of the user capturing the customer’s data
Enforce addi4onal valida4ons because of the needs of downstream processes
Enforce addi4onal process steps because of the poor controls upstream
Look across customers’ por_olios and iden4fy synergies across customers and foster collabora4ons
Based on how oPen folks have to override an upstream automated decision, add criteria upstream at the decision point
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS
Overview of the Program GeWng There …
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 11
Enterprise Metadata Model … an Ini2al Inventory
Processes
Actors
R2RML
Tasks Interac4ons
Transac2onal Systems
CRM Core Systems ERP/corporate
Opera2onal Meta Data
KPI’s SLA’s Taxonomy
Rules and Policies
Compliance Risk Regula4on
Process Orchestra2on
Front office Back office Ops
Roles
Systems
Dependencies
Inputs & outputs
Interfaces
Messages
Organiza2onal Context
Departments Products & services
Legal Structure
Charter
KPI’s
Dependencies
Inputs & outputs
Legal Jurisdic4on
Legal En4ty
Informa2on + IT Asset Inventory Func4onal Inventory
Structured Data
Unstructured Data
Use Cases
Reusable modules
Opera4onal
Analy4cal
Documents
Knowledge Systems
Real World Overlap
Geography Business domain knowledge
Opera4onal Regula4ons
Common sense knowledge News & events
Business Regula4ons
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 12
Program Overview – Possible func2onal End State
My Task Con
text Reference Data
Customer Transac2onal Interac2on
Eternal Data Sources
R2RML
My seman4c m
odel
Policy Overla
y
Governance / control / oversight
My seman4c m
odel
Process C
ontext + Current state
Access Layer (m
essage / service / fi
le /
Subject a
rea specific on
tologies
Analy2cs
Discovery
External
Ontologies
Product Org details
Transac2onal Systems
Account Trades GL
Opera2onal Meta Data
Process Role Taxonomy
Rules and Policies
Compliance Risk Regula4on
Process Orchestra2on
Front office Back office Ops
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 13
Build a context model
Approach Overview
Build a context model
Unified View of the Enterprise
Encode the process models
Applica2ons access the seman2c knowledge
Build a sta2c business ontology
Process Models
Organiza4on Structure
Services & Products
External context
Current state: Ver 1.0 Future state: Ver 2.0
Concepts
Taxonomy
Rela4onships
Inferences
Reveal data through the ontology + process context
BPM Opera4onal Reports
Business rules
Applica2ons access the seman2c knowledge
Internal Data Ontology
Encoded process Encoded rules
Encoded Architecture External Data
Actors & roles
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 14
SPAR
QL qu
ery fron
t end
Seman2c Repository
Business Process Models
Database Schema Metadata
Service Signatures + Data Types
• Build repository & import meta data • Boqom up seman4c modeling • Import Business Process Models • Import Enterprise Architecture models • Import Interface specifica4ons • Wrap data sources with SPARQL
Enterprise Architecture
Tasks
Basic Ontologies for each Domain
Interrogate and discover business & IT knowledge across the en2re ecosystem
BPMN
UML
XMI
XMI
WSDL / XML Schema / XML / JSON
XSLT
DB En4ty Model
3rd Party tools
Map to
Ontology
Database Tables
Document Stores
JDBC
JDBC Document Metadata
Actual Data
Phase 1: Integra2on
R2RML
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 15
Notes on Process Modeling
Driv
er Arrive at lot
Hand over keys
Leave for meetings /
appointments
Keys
Driv
er
Par
k P
lus
Take key
UC 1: Capture
registration
UC2: Print tags
Attach copy to
keys
Return copy to
customer
Park car in designated
location
Hang up keys in booth
Return copy to
customer
TagLocationTime Serial numberRegistration
TagLocationtime Serial number
Volume:Max: 10 per hourNormal: 1 / 2 per hour
SLA:Prints in < 10 seconds
ScreenRegistration
ScreenConformation of tags
Volume informa4on for the process shown in diagrams
Data tags shown in each element
Where SLA’s are cri4cal, show them
All interac4ons must show data between users and systems
Use cases are color coded and number labeled.
Actor for the use case
Atte
ndan
t
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 16
Meta data’s meta data
Building Domain Models
Inventory of terms to be modeled
Process context: Actor
Task
Interac4on
Importance
Role
Interfaces
Targeted process
Driv
er Arrive at lot
Hand over keys
Leave for meetings /
appointments
Keys
Driv
er
Par
k P
lus
Take key
UC 1: Capture
registration
UC2: Print tags
Attach copy to
keys
Return copy to
customer
Park car in designated
location
Hang up keys in booth
Return copy to
customer
TagLocationTime Serial numberRegistration
TagLocationtime Serial number
Volume:Max: 10 per hourNormal: 1 / 2 per hour
SLA:Prints in < 10 seconds
ScreenRegistration
ScreenConformation of tags
Indian Ci4zen
Suresh
Indian Passport
No: Z12345
Foreign ci4zen authorized to work
in the US
Suresh
US Social security number
No: Z12345
Holds passport
Is identified by
Person
Suresh
US Address
460 Park Ave S, NYC 11016
Works at
Statement Request
_ x _
Acme Bank Account
No: 12345 Has statement request
Prime holder on
Send to
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 17
Avoid deep ontology modeling
Deep, complete seman4c models are very difficult to manage in a project context, and do not add significant value in the integra4on phase
Lessons Learned – Phase 1
Avoid “shortcuts” – model the real world
Tradi4onal database designs use a variety of short cuts to make real world complexity manageable. Le}ng these propagate into the seman4c model results in an ontology that is specific to the project context, and therefore does not survive well into later phases
Avoid abstract classes
Without a context to anchor them on, discussions on abstract classes tend to go into free fall, and added liqle to know value
Evolve classes & taxonomy from defining characteris2cs
Explicitly declared taxonomies proved difficult to reverse engineer in later phases of the project
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 18
Phase 2: Ontology based Inferencing
• Enrich OWL with inference rules • Add inferencing capability to repository • Migrate business rules to RIF • Encode IT paqerns into RDF + RIF
Tasks
SPAR
QL qu
ery fron
t end
Seman4c Repository
Business Process Models
Database schema meta data
Service signatures + data types
Enterprise Architecture
Enriched ontologies
BPMN
UML
XMI
XMI
WSDL / XML Schema / XML / JSON
XSLT
DB En4ty model 3rd Party tools
Map to
ontology
Database Tables
R2RML
Document Stores
JDBC
JDBC Document meta data
Actual data
Ontology Ba
sed Inferencing
Interrogate and discover business & IT knowledge across the en4re ecosystem
Code generators (Java using Jena API + SPARQL)
Meta data for code generators (RDF) + design paqerns (RDF + RIF)
Applica4on logic driven by seman4c models
Applica4
ons
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 19
Add backward chaining rules engine Add rule cura4on to govern “learned” rules
Tasks
Phase 3: Computer Intelligible Models (just star2ng)
SPAR
QL qu
ery fron
t end
Seman4c Repository
Business Process Models
Database schema meta data
Service signatures + data types
Enterprise Architecture
Enriched ontologies
BPMN
UML
XMI
XMI
WSDL / XML Schema / XML / JSON
XSLT
DB En4ty model 3rd Party tools
Map to
ontology
Database Tables
R2RML
Document Stores
JDBC
JDBC Document meta data
Actual data
Ontology Ba
sed Inferencing
Interrogate and discover business & IT knowledge across the en4re ecosystem
Meta data for code generators + design
paqerns
Reason
ing
Event history
Business Rules Rule Cura4on Process
Code generators (Java using Jena API + SPARQL)
Applica4on logic driven by seman4c models
Applica4
ons
Public Ontolgies OWL
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 20
If a rule is not easy to define, check the model first
• Majority of cases where a “reasonable” rule was proving difficult to implement, the root cause was a poor model of the real world. Fixing the model made rule defini4on easier
Retain graph models through all the layers
• When consuming seman4c models in Java or other non-‐seman4c languages, we learned to retain the graph models through all applica4on 4ers
• Majority of developers do not use Java objects as pure “logical models” of the real world. Instead they use families of classes to enable implementa4on of logic and ensure maintainability
• Retaining the seman4c model through all 4ers was the only way to retain the Ontology overlays and enable use of business logic in all 4ers
Gaps in governance of ontology, rule and data cura4on will kill projects
• Errors in ontology and rule models are rela4vely painless to fix, they can require people to walk back mul4ple threads of thought, which can be painful
Lessons Learned – Phases 2 & 3
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS 21
Data graphs • Quads, not tuples (nothing in the default graph) • Reified statements where provenance is important
Ontology Encoding • OWL 2.0 • Tight version control • All ontologies must be uploaded to the repository
Elements in the Solu2on
Repositories • Single scalable repository for meta data • En4tlement enforced at named graph level • Federated front end only to merge meta data with actual data
Inference levels • Inferred rela4onships • Inferred taxonomy • Seman4c Web Rule Language = declara4ve rules. Rules produce new facts based on exis4ng facts in the model • Rule Interchange Format = forward & backward chaining + other business rules variants: given an outcome, can figure out the star4ng configura4on that would lead to this result
Interfaces • Interac4ve SPARQL endpoint for power users • Custom HTML 5.0 screens + canned queries + model naviga4on for occasional users • Import + export through RDF / TTL / TRIG format. Convert to RDF / TTL / TRIG using custom Java code
Unlocking th
e Pow
er of Sem
an4c Kno
wledge
1 November 2013 | Proprietary and confiden4al informa4on. © 2013 MphasiS
THANK YOU About MphasiS MphasiS an HP Company is a USD 1 billion global service provider, delivering technology-‐based solu4ons across industries, including Banking & Capital Markets, Insurance, Manufacturing, Communica4ons, Media & Entertainment, Healthcare & Life Sciences, Transporta4on & Logis4cs, Retail & Consumer Goods, Energy & U4li4es and Governments around the world. MphasiS’ integrated service offerings in Applica4ons, Infrastructure Services and Business Process Outsourcing help organiza4ons adapt to changing market condi4ons and derive maximum value from IT investments. For more informa4on about MphasiS, log on to www.mphasis.com
Presenta2on by Suresh Nair Vice President & Chief Architect, Banking & Cap Markets