cognitive computing: the art of the possible,calit2.uci.edu/uploads/media/text/ibm watson art...
TRANSCRIPT
Cognitive Computing:
The Art of the Possible,Innovation and Inspiration through IBM Watson
Ali Arsanjani, PhD
IBM Distinguished Engineer,
Chief Technology Officer,
IBM Analytics
@AliArsanjani
© 2015 IBM Corporation 2
Please Note
IBM’s statements regarding its plans, directions, and intent are subject to change or withdrawal without notice at IBM’s sole discretion.
Information regarding potential future products is intended to outline our general product direction and it should not be relied on in making a purchasing decision.
The information mentioned regarding potential future products is not a commitment, promise, or legal obligation to deliver any material, code or functionality. Information about potential future products may not be incorporated into any contract. The development, release, and timing of any future features or functionality described for our products remains at our sole discretion.
Performance is based on measurements and projections using standard IBM benchmarks in a controlled environment. The actual throughput or performance that any user will experience will vary depending upon many factors, including considerations such as the amount of multiprogramming in the user’s job stream, the I/O configuration, the storage configuration, and the workload processed. Therefore, no assurance can be given that an individual user will achieve results similar to those stated here.
© 2015 IBM Corporation 3
Topics
• Preface
• Cognitive Computing
• IBM Watson
• Adoption Patterns
– Engagement
– Watson Platform
– Exploration
– Discovery
– Decision
• Few Demos
© 2015 IBM Corporation 4
Let’s get started
Preface
© 2015 IBM Corporation 5
© 2015 IBM Corporation 6
Cognitive is the new
competitive advantage.
© 2015 IBM Corporation 7
Chapter 1
Cognitive Computing?
© 2015 IBM Corporation 8
Not …
Yet !
Cognitive Systems
© 2015 IBM Corporation 9
Cognitive systems amplify human
cognition.
© 2015 IBM Corporation 10
Cognitive systems don’t do your
thinking for you. They augment &
aid your research so that you have
more time to innovate.
© 2015 IBM Corporation 11
To do that, cognitive systems attempt
to reason about a problem space in a
way that is
similar to human thinking
so that the conclusions are intuitively obvious, transparent, and immediately
useful.
© 2015 IBM Corporation 12
Provides answers not links and webpages
Answers with evidence not guesses
Not restricted to a predefined question-answer set
Learns from every interaction
© 2015 IBM Corporation 13
In 20 years, cognitive systems will be to computing what
transaction processing is today
• Amplify human creativity
– Inspiring us to new alternatives to decision options
– Bringing the breadth of all human knowledge to the tip of our
tongue
• Learn their behavior through formal and informal training processes
• Interact with humans on our terms – in the language of humans
• Demonstrate their expertise through trust and depth of
character
• Evolve strategies of success – adapting to ever changing
knowledge and understanding
© 2015 IBM Corporation 14
What is Cognitive Computing?
• Cognitive systems are able to learn their behavior through education
• That support forms of expression that are more natural for human interaction
• Whose primary value is their expertise; and
• That continue to evolve as they experience new information, new scenarios, and new responses
… and does so at enormous scale.
© 2015 IBM Corporation 15
Principles of Cognitive Computing
• Cognitive Systems –– Applying human-like characteristics
– to conveying and manipulating ideas,
– that when combined with the inherent strengths of digital computing
– can help users to address problems with • higher accuracy, • more resilience, and • on a massive scale • over very large bodies of information.
• Watson is a Cognitive System– Teases apart the human language to identify inferences between text
passages
– Can demonstrate human-like accuracy
– Performs at speeds far faster, and at a scale far bigger, then a human.
© 2015 IBM Corporation 16
How does it work?
• Cognitive systems interpret the natural world by listening to the signals emitted in that world, and evaluating patterns in strong and weak signals that convey meaning
– Algorithmic annotators detect and score the strength of the features (signals) relevant to the domain
– Machine Learning algorithms then use the resulting vector of feature scores to learn which patterns represent meaning
– These Machine Learning algorithms are taught to recognize meaningful patterns through training data representing ground-truth in the domain of natural experience
• Cognitive systems are able to boost their comprehension through metaphorical comparison to models of the real world that are constructed from shared contextual history
• Confidence is reinforced through redundancy and supporting evidence
• Different combinations of features and Machine Learning represent different reasoning strategies that can be useful for different situations
© 2015 IBM Corporation 17
A Key driver for Cognitive Computing… we want to interpret
overwhelmingly large amounts of increasing data
2000
Vo
lum
e in
Exa
byte
s
9000
8000
7000
6000
5000
4000
3000
2015
Percentage of uncertain data
Sensors & Devices
VoIP
Enterprise Data
Social Media
Pe
rce
nt o
f un
ce
rtain
da
ta
100
80
60
40
20
0
© 2014 International Business Machines Corporation
© 2015 IBM Corporation 18
Watson began from DeepQA invoking the need for multiple reasoning
strategies for answering questions
Num
be
r o
f tim
es a
sim
ilar
qu
estio
n is
aske
d
Different questions that might be asked (including similar variations)
The “long tail” of
questions
Passage QA
Pipeline
Prediction
QA Pipeline
Knowledge QA
Pipeline
Factoid QA
Pipeline
FAQ QA
Pipeline
Query QA
Pipeline
Yes/No QA
Pipeline
Directed QA
Pipeline
Action QA
Pipeline
• Some questions are asked frequently
• We call this the ‘long tail’ of the question space
• Different types of questions demand different types of answers
• Note: Jeopardy questions only ever demanded Factoid answer types
• WEA engagements most often demand FAQ and Passage answers
• Each pipeline is responsible for reasoning about how to yield the type of answers it is responsible for
• There is a point of diminishing returns for covering absolutely every question type … some questions will never be covered and will be redirected
Jeopardy!
WDA
AlchemyAPI
WEA
WEA
What are your visiting hours?
Can you walk me through
installing a new app on my
phone?
What are the
requirements for car
insurance in Kentucky?
I want to submit a
claim
Do I need a visa to
enter Brazil?
What is the capital
of Texas? Who is known as the
father of modern
psychology?
What level of customer
churn will we see next
quarter?
What is the highest
priced phone offered
by Samsung?
19
Humans excel
at:
DILEMMAS
COMPASSION
DREAMING
ABSTRACTION
IMAGINATION
MORALS
GENERALIZATION
Cognitive
Systems
excel at:
COMMON SENSE
NATURAL LANGUAGE
LOCATING KNOWLEDGE
PATTERN IDENTIFICATION
MACHINE LEARNING
ELIMINATE BIAS
ENDLESS CAPACITY
Cognitive systems are creating a new partnership between humans and technology.
20
20
21
Examples include:
Analyst reports
tweets
Wire tap transcripts
Battlefield docs
E-mails
Texts
Forensic reports
Newspapers
Blogs
Wiki
Court rulings
International crime
database
Stolen vehicle data
Missing persons data
Data, information, and expertise
create the foundation.
Cognitive systems rely on collections of data and information:
22
There are three capabilities that differentiate cognitive systems from traditional programmed computing systems.
Reasoning
They reason. They can understand
information but also the underlying ideas
and concepts. This reasoning ability can
become more advanced over time. It’s
the difference between the reasoning
strategies we used as children to solve
mathematical problems, and then the
strategies we developed when we got into
advanced math like geometry, algebra
and calculus.
Learning
They never stop learning. As a
technology, this means the system
actually gets more valuable with time.
They develop “expertise”. Think about
what it means to be an expert- - it’s not
about executing a mathematical
model. We don’t consider our doctors to
be experts in their fields because they
answer every question correctly. We
expect them to be able to reason and
be transparent about their reasoning,
and expose the rationale for why they
came to a conclusion.
Understanding
Cognitive systems understand like
humans do, whether that’s through
natural language or the written
word; vocal or visual.
© 2015 IBM Corporation 23
What will an inquisitive mind want to know from text?
Rarely has there been a better time for start-ups to raise
money from Silicon Valley venture capitalists. But this
golden age may not last long. At least that is what Bill Gurley
says. Mr. Gurley is a prominent investor who has spent a
career betting on who the winners and losers will be in the
future of technology. Ciao!
Entities /Concepts & Relationships
Expand
Concepts
Exploreinformation
based on Concept
Associations
Sentiment & Sentiment
Targets
Deeper understanding
on author’s personality
Languageidentification
Ask questions & interact in
natural language
Resonance of the text
© 2015 IBM Corporation 24
I want to …
Classify Text Extract Entities & Relationships
Search & Rank Results Understand Utterances & Act
Expand
© 2015 IBM Corporation 25
• To pass the Turing Test and qualify as artificially intelligent, a
machine should be able to do for example: natural language
processing (i.e. communicate with no trouble on a given
language); automated reasoning (using stored information to
answer questions and draw new conclusions) and machine
learning (the ability to adapt to new circumstances and detect
patterns).
© 2015 IBM Corporation 26
© 2015 IBM Corporation 27
© 2015 IBM Corporation 28
So what are the
Characteristics of a
Cognitive System?
Scale in
Proportion
Provide Supporting Evidence
IngestVariety of (Big) Data
Respond with
Degree of Confidence
Learn with Every
Interaction
Offer Contextual
Guidance & Insights
Support for Decision Making
Generate & Evaluate
Hypothesis
UnderstandNatural
Language
Understand personality
at a Deeperlevel
Relatebetween Terms &
Concepts
Engage in a Dialog
© 2015 IBM Corporation 29
Chapter 2
IBM Watson
© 2015 IBM Corporation 30
Watson is creating a new
partnership between people
and computers that
enhances, scales and
accelerates human expertise.
© 2015 IBM Corporation 31
What is IBM Watson?
Cognitive Technology
Read & Understand
Natural Language
Generate multiple
hypothesis with
Evidence
Support for several
usage patterns
Natural extension of
what humans can do
at their best
Learns over time
Engagement Discovery Decision Exploration Platform
© 2015 IBM Corporation 32
Bringing IBM Watson to market
PlatformAlchemyLanguage Concept Expansion Concept Insights Dialog Language Translation
Natural Language Classifier Personality Insights Retrieve and Rank Tone Analyzer AlchemyVision
Visual Insights Visual Recognition AlchemyData News Tradeoff Analytics
Relationship Extraction Document Conversion Speech to Text Text to Speech
Frameworks
• IBM Watson Engagement Advisor
• IBM Watson Explorer
• IBM Watson Discovery Advisor
• IBM Watson Care Manager
• IBM Watson Decision Advisor
• IBM Watson Policy Advisor
Health Solutions
• IBM Watson for Oncology
• IBM Watson for Genomics
• IBM Watson for Clinical Trial Matching
• IBM Watson for Clinical Trial Matching Express
• IBM Watson Health Cloud for Life Sciences Compliance
X Industry Solutions
• Watson Explorer Content Miner
• Watson Curator
• Watson Analytics
• Watson Knowledge Studio
Watson Industry Apps
© 2015 IBM Corporation 33
Chapter 3
Adoption Patterns
© 2015 IBM Corporation 34
IBM Watson enables various adoption patterns
ENGAGE DISCOVEREXPLORE DECIDE PLATFORM
• Engage with
customers in a NL
conversation
• Ask questions for
greater insight
• Answer with context
• Guide consumers
to an outcome
• Merge of enterprise
search with content and
text analytics
• 360-degree Information
apps at the point of
impact
• Use Platform services to
further enhance
cognitive processing
• When finding the
question is as
important as finding
the answer
• Find rationale for
given responses
• Scaling to millions of
documents
• Ingest and analyze
domain sources,
info models
• Evidence-based
decisions with
greater confidence
• Multiple reasoning
strategies to enable
decision support
• Build your application
using callable Watson
Service APIs
• Can be combined with
the 100s of other
available services on
Bluemix
Transform customer
experiences
Search, consolidate &
analyze
enterprise data
Accelerate
Research into
Insight
Evaluate possible
options & choose
the best one
Enrich apps with Cognitive,
NLP, Vision & Machine
Learning capabilities
© 2015 IBM Corporation 35
ENGAGE DISCOVEREXPLORE DECIDE PLATFORM
• Engage with customers in a NL conversation
• Ask questions for greater insight
• Answer with
context
• Guide consumers to an outcome
• Merge of enterprise search with content and text analytics
• 360-degree Information apps at the point of impact
• Use Platform services to further enhance cognitive processing
• When finding the question is as important as finding the answer
• Find rationale for given responses
• Scaling to millions of documents
• Ingest and analyze domain sources, info models
• Evidence-based decisions with greater confidence
• Multiple reasoning strategies to enable decision support
• Build your application using callable Watson Service APIs
• Can be combined with the 100s of other available services on Bluemix
Transform
customer
experiences
Search, consolidate
& analyze
enterprise data
Accelerate
Research into
Insight
Evaluate possible
options & choose
the best one
Enrich apps with Cognitive,
NLP, Vision & Machine
Learning capabilities
IBM Watson enables various adoption patterns
Watson
Engagement
Advisor
Watson
Explorer
Watson
Discovery
Advisor
Oncology,
Clinical Trials
Matching,
Wealth
Advisor
Watson API’s on
Bluemix
© 2015 IBM Corporation 36
Pattern Description Client success stories
EngagementLeading insurance company in US introduced
Watson to answer questions from membersTransform customer experiences
DiscoveryLeading medical research institution, Baylor,
publishes breakthrough in protein targetingAccelerate Research into Insight
DecisionMemorial Sloan Kettering Cancer Center
partnered with IBM to build an Oncology
solution
Evaluate possible options and choose
the best one
Cognitive Exploration
700+ clients using the product across
industries
Search, consolidate and analyze
enterprise data
Cognitive Services
6000+ developers and 200+ partners are
building on Watson today
Enrich new or existing applications with
NLP, Vision and Machine Learning
Capabilities
Watson Adoption Patterns
© 2015 IBM Corporation 37
Engagement
© 2015 IBM Corporation 38
How Watson works: DeepQA Architecture
Multiple interpretations Hundreds of answer
sources• Primary search• Candidate answer
generation
Tens of thousands of evidence
sources and scores• Answer scoring• Evidence retrieval• Deep evidence scoring
Learned models• Combine and weigh evidence
Inquiry analysis Decomposition Hypothesis generation SynthesisHypothesis and evidence scoring
Final confidence merging & ranking
Question
Answer? !
© 2015 IBM Corporation 39
How Watson works: DeepQA Architecture
Multiple interpretations Hundreds of answer
sources• Primary search• Candidate answer
generation
Tens of thousands of evidence
sources and scores• Answer scoring• Evidence retrieval• Deep evidence scoring
Learned models• Combine and weigh evidence
Inquiry analysis Decomposition Hypothesis generation SynthesisHypothesis and evidence scoring
Final confidence merging & ranking
Question
Answer? !
Natural language processing
Information retrieval
Knowledge representation and reasoning
Deep analytics
Parallel and distributed computing
Machine learning
© 2015 IBM Corporation 40
Watson understands natural language
• Uses natural language processing and
DeepQA technology to break down &
understand a customer’s question
• There may be 100 ways to phrase the
same question – Watson is able to
learn them….
© 2015 IBM Corporation 41
Watson understands context
• Questions may only make sense if you
understand context
• E.g.
– Previous question
– Information about the consumer
– Surrounding environment
© 2015 IBM Corporation 42
Watson engages consumers in a conversation
• Watson understands that dialog is
iterative
• Show a personality representative
of the brand
• Clarification questions to carry a
conversation and get user’s intent
• Guide consumers to an outcome
© 2015 IBM Corporation 43
Watson can take action
• Watson is able to integrate
business processes into customer
conversations
– Inject the right offers or promotion
based on context
– Submit an application
– Retrieve information about a
consumer
© 2015 IBM Corporation 44
Chapter 4
Watson Platform
© 2015 IBM Corporation 45
WATSON
EXPLORER
WATSON
DEVELOPER
CLOUD
WATSON
ENGAGEMENT
WATSON
DISCOVERY
WATSON ECOSYSTEM
Web-based APIs
Message, Sentiment, and
Personality insights
Relationship Extraction
Questions&
Answers
Language
Detection
PersonalityInsights
Keyword Extraction
Image LinkExtraction
Feed Detection
VisualRecognition
Concept Expansion
ConceptInsights
DialogSentiment Analysis
Text to Speech
Tradeoff Analytics
Natural Languag
eClassifier
Author Extraction
Speech to
Text
Retrieve&
Rank
WatsonNews
LanguageTranslation
EntityExtraction
Tone Analyzer
ConceptTagging
Taxonomy
TextExtraction
MessageResonance
ImageTagging
FaceDetectio
n
Answer Generation
Usage Insights
Fusion Q&A
Video Augmentation
Decision Optimization
Knowledge Graph
Risk Stratification
Policy Identification
Emotion Analysis
Decision Support
Criteria Classification
Knowledge Canvas
Easy Adaptatio
n
Knowledge Studio
Service
Statistical Dialog
Q&A Qualification
Factoid Pipeline
CaseEvaluation
The Waston that competed on
Jeopardy! in 2011 comprised what
is now a single API—Q&A—built
on five underlying technologies.
Since then, Watson has grown to
a family of 28 APIs.
By the end of 2016, there will
be nearly 50 Watson APIs—
with more added every year.
Natural
Language
Processing
Machine Learning
Question Analysis
Feature
Engineering
Ontology Analysis
Watson Platform
built on Bluemix
© 2015 IBM Corporation 46
Discovery
© 2015 IBM Corporation 47
Published Knowledge
Knowledge-Driven Method Data-Driven Method
ObservationalData
• Longitudinal records
• Claims, Rx, Labs
• Patient reported data
• Scientific papers
• Books
• Guidelines
Closing the translational knowledge gap Personalized Insights from institutional data
From population averages … To insights for individual patient!
Watson for healthcare and life sciences spans all aspects of
knowledge and data
© 2015 IBM Corporation 48
Genes
Chemical
Compounds
Diseases
Patient
s
Animal Models
FDA Orange
Book/Moieties
Cells PatentsDrugs
Plant
Biology
®™
Meaningful insights are only gained when data reveals a
universe of relationships
© 2015 IBM Corporation 49
1New Cancer Drug targeting
p53 per year
5Potential New Cancer
Drugs in 30 days
Leveraging Watson Cognitive Capabilities toAccelerate the Discovery of New Drugs to Fight Cancer
WW Research IBM Watson
Baylor / IBM Study on p53
© 2015 IBM Corporation 50
Most Likely
Targets
Results were “Too Good to be True?” Re-ran experiment with pre 2003 information.
Only 10 known kinases triggered p53 at that time
Of the 9 found in the next decade…
Watson accurately predicted 7
Showing Researchers Where to Look
© 2015 IBM Corporation 51
Decision
© 2015 IBM Corporation 52
Business challenge:
• Clinicians have no easy way to search across eligibility criteria of relevant clinical trials for their patient; affecting research advancement and patient care
• Less than 10% of clinical studies are completed on time
• 30% of sites for clinical trials fail in enrolling even a single patient
Watson solution:
• Use patient data to instantly check eligibility across all relevant clinical trials, which can improve research and treatment outcomes
Two Use Cases
• Point of care: Cognitive Physician assistant to assess patient eligibility against all available trials
• For Pharma: Identify potentially productive sites (potentially eligible patients) help recruit patients for a specific trial
IBM Watson for OncologyTrained by Memorial Sloan Kettering
Initial focus is oncology, Watson is
trained in breast, lung and
colorectal cancer, with other tumor
types to follow
© 2015 IBM Corporation 53
IBM Watson for Clinical Trial Matching
Business challenge:
• Ability to assess quickly the best treatments for an individual patient based on latest evidence and clinical guidelines
Watson solution:
• A tool to assist physicians make personalized treatment decisions
• Analyzes patient data against thousands of historical cases and trained through thousands of Memorial Sloan Kettering MD and analyst hours
• Suggestions to help inform oncologists’ decisions based on over 290 medical journals, over 200 textbooks, and 12M pages of text
• Evolves with the fast-changing field
• Currently supports first line treatment (Breast, Lung, Colorectal cancers)
© 2015 IBM Corporation 54
Exploration
© 2015 IBM Corporation 55
Watson Explorer addresses information challenges across
your enterprise
Scaling expertise
Pressure to increase performance and innovation—while doing more with less
Insights from unstructured data
80% of data is unstructured but only a small percentage leveraged for business insights
Information access
100s of TB of data but employees struggle to get the information they need in context
ExploreInformation at the point of impact,
in 360-degree information
applications
AnalyzeContent analytics to aggregate
and analyze vast amounts of
unstructured data for business
insights
InterpretCognitive capabilities from Watson
Developer Cloud to enhance, scale
and accelerate human expertise
Cognitive ExplorationInformation, analytic insights and cognitive services to improve business outcomes
Watson Explorer V10Challenges Watson Explorer Delivers…
© 2015 IBM Corporation 56
• Discovery and navigation across disparate sources for timely data “findability”
• Merge of enterprise search with content and text analytics
– Sentiment Analytics
– Social Search/Analytics
– Social media crawler
– Domain-adaptive search
– Compound Document support
• Explore: Securely connects and explores data, regardless of format or location
• Analyze: Mines unstructured content to reveal trends, patterns and insights
• Interpret: Delivers cognitive capabilities via the Watson Developer Cloud
• Enables integration of Watson Developer Cloud services to further enhance the cognitive processing of unstructured information
• A single point of access for information and services - On-premise enterprise systems, Public and private cloud, Public and private Internet sources
Watson ExplorerAllow for exploration of easily consumable data across multiple systems
© 2015 IBM Corporation 57
Watson Explorer Use Cases
• Community policing
• Investigation analytics
• Incident management
• Antigang initiatives
• Antiterrorism initiatives
• Cyber crime investigation
• Customer experience
• Customer satisfaction and
survey analysis
• Product and service quality
• Churn prediction
• Marketing campaign
development and execution
• New revenue opportunities
• Product enhancements
• Diagnostic assistance
• Clinical treatment
• Critical care intervention
• Research for improved disease
management
• Fraud detection and prevention
• Voice of the patient
• Claims management
• Prevention of readmissions
• Patient discharge and follow-up care
• Risk assessment
• Fraud detection
• Policy and underwriting analysis
• Claims analysis, payment
validation and loss review
• Reserve trending and
optimization
• Anti–money laundering
• Internet banking fraud
• Operational efficiency
• Risk management and
compliance
Customer
InsightCrime
Analytics Healthcare Insurance Finance
© 2015 IBM Corporation 58
Cognitive Cooking
http://www.ibmchefwatson.com
© 2015 IBM Corporation 59
Watson Health is working to enhance,
scale, and accelerate expertise across the
domains of health and wellness, and to
facilitate collaboration across the
communities of care.
SolutionsPopulation Health
Management
Condition
Specific Care
Health
and Wellness
Social
Programs
Discovery
Solutions
Real World
Evidence
Ecosystem
Population Health
Management
Condition
Specific Care
Health
and Wellness
Social
Programs
Discovery
Solutions
Real World
Evidence
Individual
Social
Programs
Education
Governments
Home Health
Agencies
Practitioners
Hospitals
Therapists
Health
Plans
Family
Public Health
Medical Devices
and Diagnostics
Bio-Pharma
Employers
Payers
Data
Insight
61
As the Watson technology evolves and deepens,
so too are the ways it’s being put to work in the world.
36Countries
50,000Studentsin Melbourne
5.5MCitizensin Singapore
5LanguagesLearned by Watson
160Universitiesoffering Watson courses
400+PartnersPowered by Watson
1.1MPatientsat Bumrungrad
29Industries
80KDevelopers building with Watson
What will you do with Watson?
© 2015 IBM Corporation 63