Artificial Intelligence for Customer Relationship Management: Solving Customer Problems PDF
Preview Artificial Intelligence for Customer Relationship Management: Solving Customer Problems
Human–Computer Interaction Series Boris Galitsky Artificial Intelligence for Customer Relationship Management Solving Customer Problems Human–Computer Interaction Series Editors-in-Chief DesneyTan MicrosoftResearch,Redmond,WA,USA JeanVanderdonckt LouvainSchoolofManagement,UniversitécatholiquedeLouvain, Louvain-La-Neuve,Belgium The Human–Computer Interaction Series, launched in 2004, publishes books that advancethescienceandtechnologyofdevelopingsystemswhichareeffectiveand satisfying for people in a wide variety of contexts. Titles focus on theoretical perspectives (such as formal approaches drawn from a variety of behavioural sciences), practical approaches (such as techniques for effectively integrating user needsinsystemdevelopment),andsocialissues(suchasthedeterminantsofutility, usabilityandacceptability). HCI is a multidisciplinary field and focuses on the human aspects in the development of computer technology. As technology becomes increasingly more pervasive the need to take a human-centred approach in the design and developmentofcomputer-basedsystemsbecomesevermoreimportant. TitlespublishedwithintheHuman–ComputerInteractionSeriesareincludedin Thomson Reuters’ Book Citation Index, The DBLP Computer Science BibliographyandTheHCIBibliography. Moreinformationaboutthisseriesathttp://www.springer.com/series/6033 Boris Galitsky Artificial Intelligence for Customer Relationship Management Solving Customer Problems BorisGalitsky OracleLabs RedwoodShores,CA,USA ISSN1571-5035 ISSN2524-4477 (electronic) Human–ComputerInteractionSeries ISBN978-3-030-61640-3 ISBN978-3-030-61641-0 (eBook) https://doi.org/10.1007/978-3-030-61641-0 ©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringerNature SwitzerlandAG2021 Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher,whether thewholeorpartofthematerialisconcerned,specificallytherightsoftranslation,reprinting,reuse ofillustrations,recitation,broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,and transmissionorinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar ordissimilarmethodologynowknownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbook arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor theeditorsgiveawarranty,expressedorimplied,withrespecttothematerialcontainedhereinorforany errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional claimsinpublishedmapsandinstitutionalaffiliations. ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland Preface to Volume 2 The second volume of the book on Artificial Intelligence for Customer RelationshipManagementpresentsabroadspectrumofapplicationdomainswitha focus on solving customer problems. We design a chatbot to specifically address issues customer experiences, as well as a system that tries to understand the customer complaint, his mood, and what can be done to resolve an issue with the productorservice. In the first volume of the book, we addressed the key issues of modern CRM systems,whichfrequentlycannothandlethehighvolumeofcallsandchats.Themain limitationsofCRMtodayisalackofmachineunderstandingofwhatacustomeris tryingtosay,whataretheissuesheistryingtocommunicateorattemptstoconceal.To attacktheselimitations,weproposedanumberofNaturalLanguageUnderstanding (NLU)techniqueswiththefocusondeeplanguageanalysisandreasoning. Nothingrepelscustomersasmuchasirrelevant,annoying,intrusiveanswersand low-qualitycontent.Theinabilitytofindinformationusersneedisamajorreason theyareunsatisfiedwithaproducerofaprovider.Inthefirstvolume,wesharedthe evaluationsofquestion-answeringcomponents,verifiedtheirexplainabilityfeatures and observed a satisfactory lab performance for major CRM tasks. We concluded that keeping the customers informed is key to their acquisition and retention. For example,inFig.1,fastaccesstoinformationonwhoownsthebirdiscriticalfora smoothtransaction. Inthissecondvolume,wemakethenextsteptotheheartofthecustomerretention issue: solving a customer problem. Now just understanding what the customer is sayingisnotenoughanymore:weneedtoinvolveaformalizedknowledgeabouta productandreasonaboutit.Tosolveacustomerproblemefficientlyandeffectively, wefocusondialoguemanagement.Adialoguewiththecustomerismaintainedin such a way so that the problem can be clarified and multiple ways to fix it can be sought. IftheCRMisnotintelligentenough,itisnotveryusable.Inparticular,CRMsare usedbypeoplewhosell,andassuch,theyareoftentravelingtocustomersites,away fromtheirdesk,infrontofclients,andarepaidtogeneratessales,socommunication withaCRMisoftenanafterthought.Salespeoplearenothiredfortheircomputeror dataentryskillsbutinsteadfortheirabilitytosell.Iftheycloseasignificantdealand v vi PrefacetoVolume2 Fig.1 Communicating specificfeaturesofproducts withabuyer Fig.2 CharlesFillmore’s exampleforjointsentence discourse neverentereditintotheCRM,somethingmaybesaidafterthecongratulationsonthe sale,butnoonewouldcomplainaboutnotusingtheCRM.Thegeneralconsensus amongsalesrepresentativesissimple:noonelikesdumbCRM.Theyarefrequently overly-complex,clunkysystemswithoutaconversationalinterfacethatdonothelp themsellandevenobstructthesalesprocess.Sothepersonnelmayuseunintelligent CRMaslittleaspossible.Morethanathirdofbusinessesfacelowadoptionratesof CRMsystems. Ourdialoguemanagementisbasedondiscourseanalysis,asystematiclinguistic waytohandlethethoughtprocessofanauthor.Discourseanalysisisamethodfor studying the natural language in relation to its social context. Discourse analysis trackshowlanguageisusedinreal-lifesituations.Awell-knownAmericanlinguist CharlesFillmoredemonstratesthattwosentencestakentogetherasasinglediscourse canhavemeaningsdifferentfromeachonetakenseparately(Fig.2). The objects of discourse analysis (texts, conversations, communicative events) are defined in terms of coherent sequences of sentences, propositions and speech acts.Sometimesdiscourseanalysisevenhelpstogetaconsumeroutoftroubleby makingaconversationconvincing(Fig.3).Adialoguestructurecantakeapeculiar formforaconversationbetweentwoapplemaggots(Fig.4). PrefacetoVolume2 vii Fig.3 Findinga contradictionincustomer’s requestandcommunicating it Fig.4 Aconversational structureoftwoapple maggots Wedemonstratehowadialoguestructurecanbebuiltfromaninitialutterance.We alsointroducerealandimaginarydiscoursetreesasameanstorepresentanexplicit and implicitdiscourse of text. Aproblem of involving background knowledge on- demand, answering questions is addressed as well. We outline the Doc2Dialogue algorithmforconvertingaparagraphoftextintoahypotheticaldialoguebasedonan analysisofadiscoursetreeforthisparagraph.Thistechniqueallowsforasubstantial extensionofchatbottrainingdatasetsinanarbitrarydomain. viii PrefacetoVolume2 Fig.5 Anexampleof hypocrisy We compute user sentiments and personal traits to tailor dialogue management andcontenttoindividualcustomers.Wealsodesignseveraldialoguescenariosfor CRMwithrepliesfollowingcertainpatternsandproposevirtualandsocialdialogues forvariousmodalitiesofcommunicationwithacustomer. To detect fake content, deception and hypocrisy (Fig. 5), we need to analyze associateddiscoursepatterns,whichturnedouttodifferfromtheonesforgenuine, honest writing. With discourse analysis, we can zoom in further and characterize customercomplaintswithrespecttothebestwaytoresolvethem.Wesimulatethe mentalstates,attitudesandemotionsofacomplainantandtrytopredicthisbehavior. Having suggested graph-based formal representations of complaint scenarios, we machine-learnthemtoidentifythebestactionthecustomersupportorganizationcan choosetoretainthecomplainantasacustomer. Customer complaints are classified as valid (requiring some kind of compensation) or invalid (requiring reassuring and calming down) the customer. Scenarios are represented by directed graphs with labeled vertices (for communicative actions) and arcs (for temporal and causal relationships between theseactionsandtheirparameters).Theclassificationofascenarioiscomputedby comparing a partial matching of its graph with graphs of positive and negative examples. We illustrate machine learning of graph structures using the Nearest PrefacetoVolume2 ix Neighbor approach as well as concept learning, which minimizes the number of falsenegativesandtakesadvantageofamoreaccuratewayofmatchingsequences ofcommunicativeactions. RedwoodShores,CA,USA BorisGalitsky
The list of books you might like
