Pat Langley Pat Langley Institute for the Study of Institute for the Study of Learning and Expertise Learning and Expertise Palo Alto, California Palo Alto, California and and Center for the Study of Language Center for the Study of Language and Information and Information Stanford University, Stanford, Stanford University, Stanford, California California http://cll.stanford.edu/~langley http://cll.stanford.edu/~langley [email protected][email protected]Adaptive User Interfaces Adaptive User Interfaces for Personalized Services for Personalized Services D. Billsus, M. Chen, C.-N. Fiechter, M. Gervasio, M. Goker, W. Iba, D. Billsus, M. Chen, C.-N. Fiechter, M. Gervasio, M. Goker, W. Iba, n, and J. Yoo. n, and J. Yoo.
32
Embed
Pat Langley Institute for the Study of Learning and Expertise Palo Alto, California and
Pat Langley Institute for the Study of Learning and Expertise Palo Alto, California and Center for the Study of Language and Information Stanford University, Stanford, California http://cll.stanford.edu/~langley [email protected]. Adaptive User Interfaces for Personalized Services. - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Pat LangleyPat Langley
Institute for the Study of Learning and ExpertiseInstitute for the Study of Learning and ExpertisePalo Alto, CaliforniaPalo Alto, California
andandCenter for the Study of Language and Center for the Study of Language and
InformationInformationStanford University, Stanford, CaliforniaStanford University, Stanford, California
Adaptive User InterfacesAdaptive User Interfacesfor Personalized Servicesfor Personalized Services
Thanks to D. Billsus, M. Chen, C.-N. Fiechter, M. Gervasio, M. Goker, W. Iba, S. Rogers,Thanks to D. Billsus, M. Chen, C.-N. Fiechter, M. Gervasio, M. Goker, W. Iba, S. Rogers,C. Thompson, and J. Yoo. C. Thompson, and J. Yoo.
We now have more information and choices available than We now have more information and choices available than ever before, and we need help to handle them effectively. ever before, and we need help to handle them effectively.
The Need for Personalized AssistanceThe Need for Personalized Assistance
This has led to This has led to recommendation systemsrecommendation systems, which , which help users locate and select relevant items.help users locate and select relevant items.
But often we want But often we want personalizedpersonalized assistance that assistance that takes into account our individual preferences. takes into account our individual preferences.
However, such personalized response requires a user However, such personalized response requires a user modelmodel or or profile profile that is constructed in some manner.that is constructed in some manner.
Approaches to User ModelingApproaches to User Modeling
Hand-craftedHand-craftedProfilesProfiles
Adaptive UserAdaptive UserInterfacesInterfaces
Data-MiningData-MiningMethodsMethods
Hand-craftedHand-craftedStereotypesStereotypes
IndividualIndividualProfilesProfiles
StereotypicalStereotypicalProfilesProfiles
ManualManualConstructionConstruction
AutomatedAutomatedConstructionConstruction
Definition of an Adaptive User InterfaceDefinition of an Adaptive User Interface
that reducesuser effort
by acquiringa user model
based on pastuser interaction
a softwareartifact
Definition of a Machine Learning SystemDefinition of a Machine Learning System
that improvestask performance
by acquiringknowledge
based on partialtask experience
a softwareartifact
Applications of Adaptive User InterfacesApplications of Adaptive User Interfaces
Inferring Individual User ProfilesInferring Individual User Profiles
Our work focuses on Our work focuses on content-basedcontent-based approaches to adaptive user approaches to adaptive user interfaces, rather than on interfaces, rather than on collaborativecollaborative approaches. approaches.
Tasks that requireTasks that require a user decisiona user decision
A descriptionA descriptionfor each taskfor each task
Traces of theTraces of theuser’s decisionsuser’s decisions
Mapping from taskMapping from taskfeatures ontofeatures ontouser decisionsuser decisions
FindFind
Navigation aides already exist in both vehicles and on the World Navigation aides already exist in both vehicles and on the World Wide Web.Wide Web.
One decision-making task that confronts drivers can be stated One decision-making task that confronts drivers can be stated as:as:
However, they do not give However, they do not give personalizedpersonalized navigation advice to navigation advice to individual drivers.individual drivers.
• Given: Given: The driver’s current location The driver’s current location C;C;• Given: Given: The destination The destination DD that the driver desires; that the driver desires;• Given: Given: Knowledge about available roads (e.g., a digital map); Knowledge about available roads (e.g., a digital map); • Find: Find: One or more desirable routes from One or more desirable routes from CC to to D.D.
The Task of Route SelectionThe Task of Route Selection
The Adaptive Route AdvisorThe Adaptive Route Advisor
The Adaptive Route Advisor represents the driver model as a The Adaptive Route Advisor represents the driver model as a weighted linear combination of route features. weighted linear combination of route features.
Training cases: [x0, . . . , xn] is better than [y0, . . . , yn].Training cases: [x0, . . . , xn] is better than [y0, . . . , yn].
TimeTimeDistanceDistance
IntersectionsIntersectionsTurnsTurns
CostCostw0w1w2w3
Generating Training CasesGenerating Training Cases
The system uses each training pair as constraints on the weights The system uses each training pair as constraints on the weights found during the modeling process.found during the modeling process.
Personalized user models produce better results than generalized Personalized user models produce better results than generalized models, even when the latter are based on more data. models, even when the latter are based on more data.
Experimental Results on Route AdviceExperimental Results on Route Advice
Many online news services are available on the World Wide Many online news services are available on the World Wide Web, but few offer personalized selection.Web, but few offer personalized selection.
Another service that would benefit drivers can be stated as:Another service that would benefit drivers can be stated as:
Moreover, they are ill suited for use in the driving environment, Moreover, they are ill suited for use in the driving environment, where visual attention is a limited resource. where visual attention is a limited resource.
• Given: Given: Topics and events that interest the driver; Topics and events that interest the driver; • Given: Given: Recent news stories available on the Web;Recent news stories available on the Web;• Given: Given: Knowledge about stories the driver has heard; Knowledge about stories the driver has heard; • Find: Find: Stories to read the driver during the current tripStories to read the driver during the current trip..
Adaptive News Readers in the MarketplaceAdaptive News Readers in the Marketplace
Many recommendation systems are available on the World Wide Many recommendation systems are available on the World Wide Web, including ones that suggest restaurants.Web, including ones that suggest restaurants.
A third type of service that would assist drivers can be stated as:A third type of service that would assist drivers can be stated as:
However, they are not designed for use by drivers, who must However, they are not designed for use by drivers, who must keep their eyes on the road. keep their eyes on the road.
• Given: Given: Some goal the driver wants to achieve (e.g., eat lunch);Some goal the driver wants to achieve (e.g., eat lunch);• Given: Given: Constraints on destinations that achieve this goal; Constraints on destinations that achieve this goal; • Given: Given: Knowledge about available nearby destinations; Knowledge about available nearby destinations; • Find: Find: One or more destinations that satisfy the constraintsOne or more destinations that satisfy the constraints..
The Task of Destination SelectionThe Task of Destination Selection
The Adaptive Place AdvisorThe Adaptive Place Advisor
Results,Attribute Information
User Operatorsand Values
System Operatorsand Values
Dialogue Manager
Speech Generator Speech Recognizer
User ModelingSystem
RecognitionGrammarsPrompts
ItemDatabase
UserModels
Conversation History Updated QueryInitial Query
Retrieval Engine
System Output(Voice)
User Input(Voice)
DomainModel
Inquirer:Inquirer: Where do you think I should eat tonight? Where do you think I should eat tonight?
Advisor: What type of food would you like?Advisor: What type of food would you like?
Inquirer:Inquirer: What types are there? What types are there?
Advisor: You can say things like Chinese, Indian, and Advisor: You can say things like Chinese, Indian, and Mediterranean.Mediterranean.
Inquirer:Inquirer: Oh, maybe a French place with free parking. Oh, maybe a French place with free parking.
Advisor: What quality rating would you like?Advisor: What quality rating would you like?
Inquirer:Inquirer: I don't care, as long as it's cheap. I don't care, as long as it's cheap.
Advisor: I’m sorry, I don’t know any restaurants like that. Advisor: I’m sorry, I don’t know any restaurants like that. Would you like to search for any price?Would you like to search for any price?
Inquirer:Inquirer: No, I think I'd like Chinese instead. No, I think I'd like Chinese instead.
Advisor: In what city would you like to eat?Advisor: In what city would you like to eat?
Inquirer:Inquirer: How about Berkeley? How about Berkeley?
Advisor:Advisor: How does the Long Life Vegi House on 2129 University How does the Long Life Vegi House on 2129 University Avenue sound?Avenue sound?
Inquirer:Inquirer: Sure, that sounds fine.Sure, that sounds fine.
Adaptive Place Advisor (Before Adaptation)Adaptive Place Advisor (Before Adaptation)
Inquirer:Inquirer: Where do you think I should eat tonight? Where do you think I should eat tonight?
Advisor: What type of food would you like?Advisor: What type of food would you like?
Inquirer:Inquirer: Cheap Chinese. Cheap Chinese.
Advisor: Advisor: What city do you prefer? What city do you prefer?
Inquirer: How about Palo Alto? Inquirer: How about Palo Alto?
Advisor: How does Jing Jing Szechuan Gourmet Advisor: How does Jing Jing Szechuan Gourmet on 443 Emerson sound?on 443 Emerson sound?
Inquirer:Inquirer: Sure, that sounds fine.Sure, that sounds fine.
Adaptive Place Advisor (After Adaptation)Adaptive Place Advisor (After Adaptation)
Speech Acts Per ConversationSpeech Acts Per Conversationwith Adaptive Place Advisorwith Adaptive Place Advisor
0
2
4
6
8
10
12
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
dialogue number
num
ber o
f spe
ech
acts
modeling
control
INCA: An Adaptive SchedulerINCA: An Adaptive Scheduler
Personalized Music DeliveryPersonalized Music Delivery
A Personalized Travel AgentA Personalized Travel Agent
An Adaptive Apartment FinderAn Adaptive Apartment Finder
An Adaptive Stock TrackerAn Adaptive Stock Tracker
Alternative Presentation StylesAlternative Presentation StylesSe
quen
tial
Sequ
entia
l
Cla
ssifi
catio
nC
lass
ifica
tion
Twea
ked
Set
Twea
ked
Set
Ran
ked
List
Ran
ked
List
suggestinitialize
short-termprofile
initialize/retrieveprofile
specifyquery
modify
present
respond
decideDecision
Long-termprofile
Item database
User query
Suggestion User
Response
Updateprofile
Short-termprofile
A Flexible Framework for Adaptive InterfacesA Flexible Framework for Adaptive Interfaces
Challenges in Developing an Adaptive InterfaceChallenges in Developing an Adaptive Interface
Formulatingthe Problem
Engineering theRepresentation
CollectingUser Traces
Utilizing ModelEffectively
Gaining UserAcceptance
ModelingProcess
Contributions of the ResearchContributions of the Research
Our research program on adaptive user interfaces has produced: Our research program on adaptive user interfaces has produced:
Although some issues remain, we understand adaptive interfaces Although some issues remain, we understand adaptive interfaces well enough to apply them in practical services. well enough to apply them in practical services.
• a variety of artifacts that learn user preferences unobtrusively;a variety of artifacts that learn user preferences unobtrusively;
• evidence that this approach to user modeling is a general one;evidence that this approach to user modeling is a general one;
• experimental support for the effectiveness of these systems;experimental support for the effectiveness of these systems;
• an analysis of presentation styles possible for such systems;an analysis of presentation styles possible for such systems;
• a flexible framework for constructing them efficiently; anda flexible framework for constructing them efficiently; and
• clarification of issues that arise in their effective design.clarification of issues that arise in their effective design.
Directions for Future ResearchDirections for Future Research
Despite clear progress on adaptive user interfaces, we must still:Despite clear progress on adaptive user interfaces, we must still:
Together, these advances will lead us toward a society in which Together, these advances will lead us toward a society in which personalized computational aides are a regular part of our lives. personalized computational aides are a regular part of our lives.
• design methods to combine stereotypes and individual profiles;design methods to combine stereotypes and individual profiles;
• create approaches that transfer user profiles across domains;create approaches that transfer user profiles across domains;
• apply these techniques to an ever wider range of problems;apply these techniques to an ever wider range of problems;
• utilize new sensors to collect data even less obtrusively; andutilize new sensors to collect data even less obtrusively; and
• develop complete physical environments that adapt to users.develop complete physical environments that adapt to users.
Dialogue Operators for Adaptive Place AdvisorDialogue Operators for Adaptive Place Advisor
System OperatorsSystem OperatorsAsk-ConstrainAsk-Constrain Asks a question to obtain a value for an attributeAsks a question to obtain a value for an attributeAsk-RelaxAsk-Relax Asks a question to remove a value of an attributeAsks a question to remove a value of an attributeSuggest-ValuesSuggest-Values Suggests a small set of possible values for an attributeSuggests a small set of possible values for an attributeSuggest-AttributesSuggest-Attributes Suggests a small set of unconstrained attributes Suggests a small set of unconstrained attributes Recommend-ItemRecommend-Item Recommends an item that satisfies the current constraintsRecommends an item that satisfies the current constraintsClarifyClarify Asks a clarifying question if uncertain about latest user operatorAsks a clarifying question if uncertain about latest user operator
User OperatorsUser OperatorsProvide-ConstrainProvide-Constrain Provides a value for an attributeProvides a value for an attributeReject-ConstrainReject-Constrain Rejects the proposed attributeRejects the proposed attributeAccept-RelaxAccept-Relax Accepts the removal of an attribute valueAccepts the removal of an attribute valueReject-RelaxReject-Relax Rejects the removal of an attribute valueRejects the removal of an attribute valueAccept-ItemAccept-Item Accepts the proposed item Accepts the proposed item Reject-ItemReject-Item Rejects the proposed itemRejects the proposed itemQuery-AttributesQuery-Attributes Asks system for information about possible attributesAsks system for information about possible attributesQuery-ValuesQuery-Values Asks system for information about possible attribute valuesAsks system for information about possible attribute valuesStart-OverStart-Over Asks the system to re-initialize the searchAsks the system to re-initialize the searchQuitQuit Asks the system to abort the searchAsks the system to abort the search