Watson, what s on, what s next? Mikael Haglund CTO, IBM Sweden mikael.haglund@se.ibm.com about.me/mikaelhaglund
2 2013 IBM Corporation
Sinnen Kognitiva System Hjärnliknande chips Cognitive Computing Do not distribute 2013 IBM Corporation 3
Watson Do not distribute 2013 IBM Corporation
IBM Watson, an early example of a cognitive system 1 Understands natural language and human speech 2 Generates and evaluates hypothesis for better outcomes 3 Adapts and Learns from user selections and responses built on a massively parallel probabilistic evidence-based architecture 2013 IBM Corporation
Open-Domain in Business Questions vs. Clues Questions What is the limit of a Roth IRA contribution based on an annual income of $110,000? Clues This is the limit of a Roth IRA contribution based on an annual income of $110,000. What drug has been shown to relieve the symptoms of ADD with relatively few side effects? This drug has been shown to relieve the symptoms of ADD with relatively few side effects. I was running HALO3 and my machine took a dive with a pink screen, what are the possible causes? I was running HALO3 and my machine took a dive with a pink screen, these are the possible causes.
Informed decision making: search vs. Watson Decision Maker Has Question Search Engine Distills to 2-3 Keywords Reads Documents, Finds Answers Finds & Analyzes Evidence Finds Documents Containing Keywords Delivers Documents Based on Popularity Decision Maker Asks NL Question Considers Answer & Evidence Watson Understands Question Produces Possible Answers & Evidence Analyzes Evidence, Computes Confidence Delivers Response, Evidence & Confidence
Some Challenges
Some Basic Jeopardy! Clues This fish was thought to be extinct millions of years ago until one was found off South Africa in 1938 Category: ENDS IN "TH" Answer: coelacanth The type of thing being asked for is often indicated but can go from specific to very vague When hit by electrons, a phosphor gives off electromagnetic energy in this form Category: General Science Answer: light (or photons) Secy. Chase just submitted this to me for the third time--guess what, pal. This time I'm accepting it Category: Lincoln Blogs Answer: his resignation
Broad Domain We do NOT attempt to anticipate all questions and build databases. We do NOT try to build a formal model of the world In a random sample of 20,000 questions we found 2,500 distinct types*. The most frequent occurring <3% of the time. The distribution has a very long tail. And for each these types 1000 s of different things may be asked. Even going for the head of the tail will barely make a dent *13% are non-distinct (e.g, it, this, these or NA) Our Focus is on reusable NLP technology for analyzing vast volumes of as-is text. Structured sources (DBs and KBs) provide background knowledge for interpreting the text.
Different Types of Evidence: Keyword Evidence In May 1898 Portugal celebrated the 400th anniversary of this explorer s arrival in India. In May, Gary arrived in India after he celebrated his anniversary in Portugal. arrived in celebrated Keyword Matching celebrated In May 1898 Keyword Matching In May 400th anniversary Keyword Matching anniversary Portugal Keyword Matching in Portugal arrival in India Keyword Matching India explorer Gary
Different Types of Evidence: Deeper Evidence In May 1898 Portugal celebrated the 400th anniversary of this explorer s arrival in India. On On 27th 27th May May 1498, 1498, Vasco Vasco da da Gama Gama On landed 27th May landed in in Kappad Kappad 1498, Vasco Beach Beachda Gama On the 27 landed in th of May 1498, Vasco da Kappad Beach Gama landed in Kappad Beach Ø Search Far and Wide Ø Explore many hypotheses celebrated Portugal Ø Find Judge Evidence Ø Many inference algorithms landed in May 1898 400th anniversary Temporal Reasoning 27th May 1498 Stronger evidence can be much harder to find and score. arrival in India explorer Statistical Paraphrasing GeoSpatial Reasoning Date Math Paraphrases Geo-KB Kappad Beach Vasco da Gama The evidence is still not 100% certain. 16
Watson s Guts https://www.youtube.com/watch?v=dywo4zksfxw
How Watson works: DeepQA Architecture Learned Models help combine and weigh the Evidence Inquiry Answer Sources Primary Search Candidate Answer Generation Answer Scoring Evidence Sources Evidence Retrieval Deep Evidence Scoring Models Models Models Models Models Models Inquiry/Topic Multiple Analysis Interpretations of a question 100 s sources Inquiry Decomposition Hypothesis Generation Hypothesis and Evidence Scoring Synthesis Final Confidence Merging & Ranking Hypothesis Generation Hypothesis and Evidence Scoring Responses with Confidence
How Watson works: DeepQA Architecture Inquiry Inquiry/Topic Multiple Analysis Interpretations of a question Answer Sources Primary Search Candidate Answer Generation 100 s sources Inquiry Decomposition 100 s Possible Answers Hypothesis Generation Answer Scoring Evidence Sources 1000 s of Pieces of Evidence Evidence Retrieval Hypothesis and Evidence Scoring Deep Evidence Scoring 100,000 s Scores from many Deep Analysis Algorithms Synthesis Learned Models help combine and weigh the Evidence Balance & Combine Models Models Models Models Models Models Final Confidence Merging & Ranking Hypothesis Generation Hypothesis and Evidence Scoring Responses with Confidence
Moving beyond Jeopardy! is a non-trivial challenge Watson at Play 1 User Max. input was two sentences 5+ days to retrain Evidence not present Text-only input Q&A model Basic security Watson at Work 10s of thousands concurrent users Pages of input (e.g. medical record) Dynamic content ingestion Supporting evidence integral Text, tables and images as input Both Q&A + Conversation model High security (e.g. HIPAA)
Watson möjliggör tre klasser av kognitiva tjänster Fråga Absorbera och dra nytta av enorma mängder data Ställ nyanserade frågor för bättre insikter Förstå frågor ställda med naturligt språk Upptäck Hitta de djupa sambanden Begär mer information för att förbättra svaren Gå från sökning till att göra upptäckter Fatta beslut Absorbera och analysera källor från relevanta domäner Fatta evidens-baserade beslut Lär från varje situation, åtgärd och följd
IBM Watson can be applied to many pressing business priorities Watson Decision Advisor Watson Engagement Advisor Watson Discovery Advisor Watson Case Advisor https://www.youtube.com/watch?v=hzspc0h_mtm
IBM Watson can be applied to many pressing business priorities Watson Decision Advisor Watson Engagement Advisor Watson Discovery Advisor Watson Case Advisor
Traditional approaches to engaging with customers come up short 270B Calls made annually to call center costing $600B 1 in 2 incoming calls require escalation or go unresolved 61% of all calls could have been resolved with better access to information 4.6% Market value gain from a single point customer sat gain *Case studies based on Coremetrics, Sterling Commerce and Unica solutions
Customers expect personalization and control You don t know me Intolerance of mass-market, impersonalized approaches You re not connecting with me Demand for interaction on channel of choice You make it too hard Expectations for immediate results
Do not distribute 2013 IBM Corporation
Watson Engagement Advisor - 6 Weeks Deploy / 6 Mo. ROI Ready Build Teach Pilot Run Extend Identify content Develop Q&A training set Upload Q&A pairs Configure & adapt for use cases Validate UX Auto. ingest documents (PDF, HTML, etc.) Add content Test & evaluate Migrate to production Utilize Watson in production Full production Q&A pairs Expand corpus Utilize Watson in new domains Automated content ingestion tools Toolkit allows building custom UIs Q&A pairs for training & accuracy Activity Monitoring Tools Initial Ready/Run Cycle = 6 Weeks WEA End User UI Tight integration with existing infrastructure or stand alone offering
https://www.youtube.com/watch?v=mr-1janairs 2013 IBM Corporation
2013 IBM Corporation
Watson Ecosystem
Fluid, Powered by IBM Watson Jag och familjen tänker åka till svenska fjällen och gå en vandringsled, och sova i tält i sommar. Vi är inte vana vandrare så det kommer att vara ganska korta etapper. Vi är fyra personer. Vad behöver vi? Barnen kan inte bära ett eget tält.
Watson på Svenska? Brisman(?) IBM i Sveriges första chef, 1928
Äh, det är ju bara att översätta... (1) Chandeliers look great but nowadays do not usually use these items from which their name is derived. Bonus: Vad sökte man? What is a...
Äh, det är ju bara att översätta... (2) Chandeliers look great but nowadays do not usually use these items from which their name is derived. Översättning med Google translate: Ljuskronor ser bra ut, men nu för tiden brukar inte använda dessa poster från vilken deras namn härstammar. Borde vara något i stil med: Ljuskronor ser bra ut, men nu för tiden brukar de inte använda dessa föremål från vilka deras namn härstammar.
Sinnen Kognitiva System Hjärnliknande chips Cognitive Computing Do not distribute 2013 IBM Corporation 35
Smarter Computing 1997 Die Mensch Maschine Halb Wesen und halb Ding Die Mensch Maschine Halb Wesen und halb Überding /Kraftwerk
Tack! Mikael Haglund IBM mikael.haglund@se.ibm.com about.me/mikaelhaglund 2011 IBM