Zarneger Content Preparation and Delivery to Support AI

Post on 13-Apr-2017

26 views 1 download

Transcript of Zarneger Content Preparation and Delivery to Support AI

SILVERCHAIR

ContentPreparationandDeliverytoSupportArtificialIntelligence

JakeZarnegarChiefProductOfficer,Silverchair

SILVERCHAIR

SILVERCHAIR

Muchlike“thecloud,”“bigdata,”and“machinelearning”beforeit,theterm“artificialintelligence”hasbeenhijackedbymarketersandadvertisingcopywriters.

Ifthehypeleavesyouasking“WhatisA.I.,really?,”don’tworry,you’renotalone.Iaskedvariousexpertstodefinetheterm andgotdifferentanswers.Theonlythingtheyallseemtoagreeonisthatartificialintelligence isasetoftechnologiesthattry toimitateoraugmenthumanintelligence.

Tome,theemphasisisonaugmentation,inwhich intelligentsoftwarehelpsusinteractanddealwiththeincreasinglydigitalworldwelivein.

OmMalik,TheNewYorkerhttp://www.newyorker.com/business/currency/the-hype-and-hope-of-artificial-intelligence

SILVERCHAIR

GeneralAugmentation

http://www.cnn.com/2017/01/26/health/ai-system-detects-skin-cancer-study/

SpecificAugmentation

TwoTypesofAugmentation

SILVERCHAIR

1:OngoingIndependentLearningFromInteractionwithStimuli

2:ComplexInteractionwithHumans

TwoAI“Augmentation”ConditionsThatMostAgreeOn

SILVERCHAIR

Howcanscientificandscholarlypublishersbestprepareanddelivercontenttoassist(andnothinder)

theadvancementofAIsystems?

Today’sQuestion

SILVERCHAIR

KnowledgeDiscoveryinDatabases(KDD)[1]isdividedinfourmainphases:domainexploration,datapreparation,datamining,andinterpretationofresults.

1. Thefirstphaseisresponsibleforunderstandingtheproblemandwhatdatawillbeusedintheknowledgediscoveryprocess.

2. Thenextphaseselects,cleans,andtransformsthedatatoaformatthatissuitableforaspecificdataminingalgorithm.

3. Inthethirdphase,thechosendataminingalgorithmperformssomeintelligenttechniquestodiscoverpatternsthatcanbeofpotentialuse.

4. Thelastphaseisresponsibleformanipulatingtheextractedpatternstogenerateinterpretableknowledgeforhumans…

SILVERCHAIR

…Mostoftheresearchcarriedoutinthisareafocusonthedataminingphase,whichusesartificialintelligencealgorithmslikedecisiontrees,artificialneuralnetworks,evolutionarycomputation,amongothers[2]todiscoverknowledge.Ontheotherhand, thedatapreparationphase,responsibleforintegration,cleaning,andtransformationofdata,hasnotbeenthesubjectofmuchresearch.Infact,Pyle[3]arguesthat “datapreparationconsumes60to90%ofthetimeneededtominedata– andcontributes75to90%totheminingproject’ssuccess”.

From: PauloM.Goncalves Jr.and RobertoS.M.Barros, "AutomatingDataPreprocessingwithDMPMLandKDDML," 10thIEEE/ACISInternationalConferenceonComputerandInformationScience,2011, DOI:10.1109/ICIS.2011.23.

SILVERCHAIR

“Idownloaded2TBofArxiv contentlastweekbutI can’tbringmyselftoopenitandstartworkingonanalyzingitbecauseIknowIhaveatleast6monthsofpainstakingdatacleanup&preparationaheadofmebeforeIcanbegin.”

--MikeM.,FastForwardLabs

SILVERCHAIR

WhereWeAreNow:ApplyingtheLevelsofCognitiveLearningtoSoftware

SILVERCHAIR Bloom, et al. 1956

SILVERCHAIR

• We’vemasteredthis!• Thefundamentalsofthe

permanentscholarlyrecord(DOI,CLOCKSS,PDF,etc.)

SILVERCHAIR

• Alsostrongincreatinginterfacesthatassistunderstandingfromhumanreaders

SILVERCHAIR

FosteringSoftwareUnderstanding

Insomewayswe'vegotagoodfoundation– detailed,consistentcontenttaggingtoaidwithsoftwareUnderstanding

• StructureunderstandingthroughnormalizedXML:whatisthetitle,authors,abstract, wheretheconclusionsareinthepaper,etc.

Increasedtaggingofnamedentities:understandingwhatisagene,whatisaclinicaltrialID,whatisaperson

• Thiscanstillgoawry: "BethIsrael,""BethIsraelDeaconess"examples

SILVERCHAIR

http://anesthesiology.pubs.asahq.org/article.aspx?articleid=2592740

https://academic.oup.com/rheumatology/article/doi/10.1093/rheumatology/kex082/3101351/Musculoskeletal-manifestations-of-Ebola-virus

SILVERCHAIR

1:Interfacesstillprimarilyvisual,narrative

2:HelpfulunderlyingXMLstructurenotshared

3:Littletonotaggingabove“Understanding”level

3ObstaclestoHigherSoftwareCognition

SILVERCHAIR

WhereWe’reGoing:TheRacetotheTop

SILVERCHAIR Bloom, et al. 1956

SILVERCHAIR

• Full-textnormalizedXML(orJSON)• Separateproduct/subscriptionforsale• Separatedeliverymechanism(nohumaninterface)butcan

piggybackonexistingcontentworkflows• Accessesanewclassofcustomerw/deeppockets

(AIcreatorsorimplementers)• Requiresnewvetting/legalagreements

ConsiderProvidingYourStructuredContentasaNewProduct

SILVERCHAIR

…Mostoftheresearchcarriedoutinthisareafocusonthedataminingphase,whichusesartificialintelligencealgorithmslikedecisiontrees,artificialneuralnetworks,evolutionarycomputation,amongothers[2]todiscoverknowledge.Ontheotherhand, thedatapreparationphase,responsibleforintegration,cleaning,andtransformationofdata,hasnotbeenthesubjectofmuchresearch.Infact,Pyle[3]arguesthat “datapreparationconsumes60to90%ofthetimeneededtominedata– andcontributes75to90%totheminingproject’ssuccess”.

From: PauloM.Goncalves Jr.and RobertoS.M.Barros, "AutomatingDataPreprocessingwithDMPMLandKDDML," 10thIEEE/ACISInternationalConferenceonComputerandInformationScience,2011, DOI:10.1109/ICIS.2011.23.

SILVERCHAIR

• Developyourownsoftware(ordevelopanalysis,applicationandevaluation)higherupthecognitionpyramid

• Ifthat’sthecase,don’tshareyourstructuredcontentwithpotentialcompetitors

OrConsiderCompetingDirectly!

SILVERCHAIR

ThankYou

JakeZarnegarChiefProductOfficer,Silverchair