DOC PREVIEW
CMU CS 15492 - Speech Processing

This preview shows page 1-2-15-16-31-32 out of 32 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 32 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Speech Processing 15-492/18-492Spoken Dialog Systems- Details of Olympus modules- Dialog Task DesignThe Olympus ArchitectureBackendKnowledgeSourcePhone /DesktopInterpretationPHOENIXDialog ManagerRAVENCLAWGenerationROSETTARecognitionAUDIOSERVERSynthesisKALLIOPEInteraction MgrAPOLLOSynth. Engine(SAPI/FLITE)Recog. Engine(SPHINX)RavenClawPlanPlan--based dialog managerbased dialog managerTaskTask--independent engineindependent enginecore Olympus librarycore Olympus librarymanage dialog by executing task specificationmanage dialog by executing task specificationprovides generic domainprovides generic domain--independent behaviorindependent behaviorHelp, repeat, …Help, repeat, …Confirmation, nonConfirmation, non--understandings…understandings…Dialog Task SpecificationDialog Task Specificationdialog plandialog planinterpretation contextinterpretation contextRavenClaw ArchitectureManages dialog by Manages dialog by executing the dialog executing the dialog task specificationtask specificationProvides many Provides many domaindomain--independent independent conversational conversational strategiesstrategiesStandard for most Standard for most applicationsapplicationsNo need to modify, just No need to modify, just link shared librarylink shared libraryTask-independent Dialog Engine Dialog Task (Specification)Captures all domainCaptures all domain--specific dialog (task) specific dialog (task) logic using a logic using a hierarchical descriptionhierarchical descriptionUnique to each Unique to each applicationapplicationMust be created for Must be created for each applicationeach applicationLinks to dialog engine Links to dialog engine librarylibraryRavenClaw: Dialog Task SpecificationMadeleineE:LoadSymptoms GeneralFeelR:HowAreYou? I:Glad I:SorryDiagnoseFever TravelR:AskFever E:MeasureTemp I:InformFeverI:Welcomegeneral_feelinghave_feverdiagnosticTree of dialog agentsTree of dialog agentsTerminals: Inform, Request, Expect, ExecuteTerminals: Inform, Request, Expect, ExecuteNonNon--terminals / Dialog agency: plans execution of child nodesterminals / Dialog agency: plans execution of child nodesHierarchical Task Execution Network; each agent:Hierarchical Task Execution Network; each agent:PreconditionsPreconditionsSuccess & failure criteriaSuccess & failure criteriaTrigger (focus) criteriaTrigger (focus) criteriaEffectsEffectsSample Task Specification Code// /Madeleine/GeneralFeelDEFINE_AGENCY(CGeneralFeel,DEFINE_CONCEPTS(STRING_USER_CONCEPT(general_feeling, none))DEFINE_SUBAGENTS(SUBAGENT(HowAreYou, CHowAreYou)SUBAGENT(Glad, CGlad)SUBAGENT(Sorry, CSorry))SUCCEEDS_WHEN(COMPLETED(Glad) || COMPLETED(Sorry)))// /Madeleine/GeneralFeel/HowAreYouDEFINE_REQUEST_AGENT(CHowAreYou,REQUEST_CONCEPT(general_feeling)GRAMMAR_MAPPING("![Yes]>good, ![FeelingGood]>good, ""![FeelingSoSo]>soso, ![FeelingBad]>bad")))// /Madeleine/GeneralFeel/GladDEFINE_INFORM_AGENT(CGlad,PRECONDITION(C("general_feeling") == CString("good"))PROMPT("inform glad_youre_good")ON_COMPLETION(FINISH(/Madeleine)))// /Madeleine/GeneralFeel/SorryDEFINE_INFORM_AGENT(CSorry,PRECONDITION(C("general_feeling") != CString("good"))PROMPT("inform sorry_youre_bad"))R:HowAreYou?GeneralFeelI:Glad I:Sorrygeneral_feelingRavenClaw Task Specification Language (RCTSL)(Pseudo(Pseudo--)declarative language)declarative languageDefines concept typesDefines concept typesDescribes the task treeDescribes the task treeSet of C++ macrosSet of C++ macrosConcept types and agents are classesConcept types and agents are classesCan use pure C++ code if necessaryCan use pure C++ code if necessaryNeed to be recompiled when modifiedNeed to be recompiled when modifiedRCTSL ConceptsConcepts are effectively RCTSL variablesConcepts are effectively RCTSL variablesStore values for later use and manipulationStore values for later use and manipulationStandard typesStandard typesString, integer and String, integer and boolboolUserUser--defined typesdefined typesStructures and arraysStructures and arraysTwo main categories:Two main categories:SystemSystemconceptsconceptsStore internal values, database results, etc.Store internal values, database results, etc.UserUserconceptsconceptsCapture entities obtained from the userCapture entities obtained from the userHow User Concepts get ValuesGRAMMAR_MAPPINGGRAMMAR_MAPPINGdirectivedirectiveDefines which grammar Defines which grammar slot(sslot(s) from Phoenix ) from Phoenix are assigned to an expected conceptare assigned to an expected concept// /MyBus/PerformTask/GetQuerySpecs/RequestOriginPlaceDEFINE_REQUEST_AGENT( CRequestOriginPlace,REQUEST_CONCEPT(origin)PROMPT("request origin_place")GRAMMAR_MAPPING("[origin_place], ![Place]") )••Maps parsed value from grammar (slotMaps parsed value from grammar (slot[[origin_placeorigin_place]]) to concept ) to concept originoriginSpecifying Binding ScopeInitiative can be controlled via binding scopeInitiative can be controlled via binding scopeSystem vs. Mixed initiativeSystem vs. Mixed initiativeGrammar mappings encode binding scope:Grammar mappings encode binding scope:Special character before grammar slot nameSpecial character before grammar slot nameStrict (!): bind only when request agent is activeStrict (!): bind only when request agent is activeOpen (@): bind alwaysOpen (@): bind alwaysDefault (Ø): bind only when request agent’s Default (Ø): bind only when request agent’s subtask is activesubtask is active// /MyBus/PerformTask/GetQuerySpecs/RequestOriginPlaceDEFINE_REQUEST_AGENT( CRequestOriginPlace,REQUEST_CONCEPT(origin)PROMPT("request origin_place")GRAMMAR_MAPPING("[origin_place], ![Place]") )RavenClaw ExecutionDialog StackMadeleineE:LoadSymptoms GeneralFeelR:HowAreYou? I:Glad I:SorryDiagnoseFever TravelR:AskFever E:MeasureTemp I:InformFeverI:WelcomeExpectation Agendageneral_feelingcharthave_feverdiagnosticRavenClaw ExecutionDialog StackMadeleineMadeleineE:LoadSymptoms GeneralFeelR:HowAreYou? I:Glad I:SorryDiagnoseFever TravelR:AskFever E:MeasureTemp I:InformFeverI:WelcomeExpectation Agendageneral_feelingcharthave_feverdiagnosticRavenClaw ExecutionDialog StackMadeleineWelcomeMadeleineE:LoadSymptoms GeneralFeelR:HowAreYou? I:Glad I:SorryDiagnoseFever


View Full Document
Download Speech Processing
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Speech Processing and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Speech Processing 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?