Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation
Precision, Accurate Confidence and Speed What s Next? Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation
The Future Watson Extend Watson technology Moves beyond question-in & answer-out to always learning evidence-based decision support Lead in new domains Addresses the enterprise need to convert growing volumes of information into actionable knowledge Demonstrates business value in critical problem spaces, starting with Healthcare Enable efficient adaptation Efficiently adapting and scaling Watson to new domains requires a novel blend of engineering and research Global Technology Outlook 2012 - IDo Not Distribute 2011 IBM Corporation
Watson s real value proposition: Efficient decision support over unstructured (and structured) content Deeper Understanding, Higher Precision and Broader, Timely Coverage at lower costs Open-Domain Question-Answering Key Word Search Relevance Ranking Jeopardy! Challenge Existing BI Inference/Rules SQL/XQuery Shallow Understanding Low Precision Broad Coverage Unstructured Data Broad, rich in context Rapidly growing, current Invaluable yet under utilized Deeper Understanding but Brittle High Precision at High Cost Narrow Limited Coverage Structured Data Precise, explicit Narrow, expensive Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 4
Taking Watson beyond Jeopardy! Understanding Interacting Explaining Learning Specific Questions Question-In/Answer-Out Precise Answers & Accurate Confidences Batch Training Process The type of murmur associated with this condition is harsh, systolic, and increases in intensity with Valsalva From specific questions to rich, incomplete problem scenarios (e.g. EHR) analysis and look-ahead, drive interactive dialog to refine answers and evidence Move from quality answers to quality answers and evidence Scale domain learning and adaptation rate and efficiency Input, Responses Answers, Corrections, Judgements Entire Medical Record Dialog Refined Answers, Follow-up Questions Responses, Learning Questions Rich Problem Scenarios Interactive Dialog Teach Watson Comparative Profiles Continuous Training & Learning Process Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 5
Moving from specific questions to high-value problem scenarios Question Answer Specific, short questions produce a narrow, targeted range of hypotheses. Much Harder Multi-Dimensional Input and Multi-Dimensional Confidence Merging Problem Scenario........ Any subset of factors may provide some indication of an answer (e.g., Diagnosis or Treatment) Factor 1 Factor n Combine Combine Combine How to combine confidences across factors and factor sets Answer Set Answer i Answer j Answer k Which subsets of factors represent fruitful hypotheses and which may be sufficiently evidenced by the content? There are ~55K fields in an EMR any subset may predict one or more of some several 100K diseases or conditions. And the clinicians will want to know WHY. Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 6
Teach Watson technology generates questions to enhance understanding and acquire new knowledge used in dialog or crowd-sourcing opportunities Watson considers What would have to be true for seizure disorder to be correct? What neurological condition contraindicates the use of bupropion? Automatically generates Learning Questions Does contraindicates the use of bupropion mean should not use bupropion? Q A Patients These with disorders preexisting can stop seizure the disorder nerves and should muscles not in use bupropion your esophagus due to from a higher-than-proportional working right. This can increase cause in food the possibility to move slowly of seizure or even as the get dose stuck is in the increased. esophagus. Does contraindicates the use of mean should not use in general? Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 7
Dialoguing to an answer Present Factors Red, painful eye Blurred vision Family history of arthritis Q: What diagnosis explains the patient s condition? Medical Record Absent Factors Circular rash Fatigue Headache The first symptom of Lyme disease (also called Lyme s disease) Lyme disease for about can 50% affect of people different is body a small, systems, red bull s-eye such rash, the called nervous erythema system, migrans, joints, skin, at the and site of an heart. infected Symptoms tick bite. are often described as Other early, happening acute Lyme in three symptoms stages (although are flu-like not everyone fatigue, achy experiences muscles all or three): joints, fever, chills, stiff neck, swollen 1.A circular glands, rash, and typically a headache. within 1-2 weeks of infection, often is the first sign of infection. Lyme disease 2.Along is with is caused by the rash, the a bacterium person Borrelia may have flu-like burgdorferi and symptoms is is transmitted such to as swollen to humans through lymph nodes, the fatigue, bite of of infected headache, blacklegged ticks. and muscle Typical aches. symptoms include high temperature, headache, fatigue, and a characteristic skin rash called erythema migrans. Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 8
It s all about the evidence Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 9
Watson 2.0: From Jeopardy! to Clinical Decision Support Profiles explain answers and confidences, citing key evidence Follow-Up and Learning Questions most likely to impact Profiles Scenario/Case e.g., Entire Medical Record Dialog Inference Chaining Question Generation Scenario Analysis Hypothesis Generation Discovery Hypothesis & Scoring Analysis Final Confidence Merging & Ranking Multi-Dimensional Input Formulate meaningful questions and discover subsets of factors that independently contribute to answers Extending beyond text evidence to images and speech Multi-Dimensional Merging Merge confidences across independently contributing factors Intermediate hypotheses result in recursive calls to system (Chaining) Factors present in evidence but absent in hypotheses lead to follow-up questions Ambiguities and missing knowledge result in learning questions Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 10
Domain learning: Training and adapting Watson to new domains Y Start New Algorithms Positive Impact? Algorithm Development Systems Engineers Research Plan Idea Generation & Prioritization Update/Compile System Researchers N Watson2 Biweekly Build Experimental Research Process Systems Engineers Researchers Learning Statistical ML Train System Training Data Indices & Derived Resources Content Analysis Prep Ideas Content Evaluation & Prep ML Models Learning Knowledge Extraction Original Content Test Data System Run Answers & Domain Experts Test Data Input & Output Learning Interactive Teach Watson Expert Annotation Answer & Vetting and Enrichment Headroom Analysis Researchers Headroom Analysis Estimate Potential Impact of Addressing Failures Research Opportunities Accuracy Analysis Identify & Classify Key Accuracy Failures Vetted Output & Enriched Training Data Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation 11
The Future Watson Extend Watson technology Moves beyond question-in & answer-out to always learning evidence-based decision support Lead in new domains Addresses the enterprise need to convert growing volumes of information into actionable knowledge Demonstrates business value in critical problem spaces, starting with Healthcare Enable efficient adaptation Efficiently adapting and scaling Watson to new domains requires a novel blend of engineering and research Global Technology Outlook 2012 - IDo Not Distribute 2011 IBM Corporation
Global Technology Outlook 2012 - Do not Distribute 2011 IBM Corporation