User profiles

hurriedtinkleAI and Robotics

Nov 15, 2013 (3 years and 8 months ago)

84 views

User Modeling,
Adaptation,
Personalization

Part 2

ΕΠΛ 435:

Αλληλεπίδραση Ανθρώπου
Υπολογιστή

In this lecture

User profiles




Main techniques for acquiring user profiles



Examples for representing user profiles



01/11/2013

Τμήμα Πληροφορικής

2

General Architecture of

User
-
Adaptive Systems

USER PROFILE

USER MODEL

ACQUISITION

USER MODEL

APPLICATION

INFORMATION ABOUT
U

ADAPTING TO
U

01/11/2013

Τμήμα Πληροφορικής

3

User Profiles

Include general information about the users



Demographic information


Name


Age


Country


Education level





User interests


List of key words


List of topics





User preferences


Disabilities


Preferred interaction style


Preferred media




01/11/2013

Τμήμα Πληροφορικής

4

General Architecture of

User
-
Adaptive Systems

USER PROFILE

USER MODEL

ACQUISITION

USER MODEL

APPLICATION

INFORMATION ABOUT
U

ADAPTING TO
U

Identify the user

Collect information about the user

01/11/2013

Τμήμα Πληροφορικής

5

Methods for User Identification


Cookies


Easiest and most widely deployed


Accuracy can be poor



Data aggregators


Example: Acxiom (www.acxiom.com)


Provide demographic information about customers



Software agents


Small programs that reside on the user’s computer


Collect information about the user and share with a
server via some protocol


More control over the implementation

01/11/2013

Τμήμα Πληροφορικής

6

Methods for User Identification


Logins


Users identify themselves upon login


Can access information from different computers


Accuracy pretty good



Enhanced proxy servers


User registers their computer with a proxy server


User identification via identification of the computer


Usage of several computers (all should be registered
to the proxy server)



Session IDs


Store information about the user per visit


Short
-
term user profile, not suitable for long term

01/11/2013

Τμήμα Πληροφορικής

7

Methods for User Information
Collection


Explicit information collection


Information entered by the user


Most common


use forms




Fairly reliable


Complies with privacy regulations




Requires time and willingness to contribute


Can be obtrusive


Dynamic changes can be missed

01/11/2013

Τμήμα Πληροφορικής

8

Methods for User Information
Collection


Implicit User Information Collection (
Gauch

et al., 2007)

Collection technique

Information collected

Information breath

Browser cache

Browsing history

Web sites

Proxy servers

Browsing activity

Web sites

Interaction agents

Interaction activity

Any personalised
application

Desktop agents

All user activity

Any personalised
application

Logs (most commonly
web logs)

Interaction activities
(browsing activity)

Logged applications
(web sites)

Search logs

Search

Search engine site

01/11/2013

Τμήμα Πληροφορικής

9

Explicit vs Implicit Information
Collection


Studies are inconclusive



Earlier research


support for explicit



Most recent research


support for implicit



Agreement


combined performs reasonably well

01/11/2013

Τμήμα Πληροφορικής

10


User Profile: example


3D shopping mall:
www.activeworlds.com

Luca
Chittaro

and Roberto
Ranon
, Adaptive 3D Web Sites, P.
Brusilovsky
, A.
Kobsa
, and

W.
Nejdl

(Eds.): The Adaptive Web, LNCS 4321, pp. 433


462, 2007.

Demographic data


User preferences


Usage data sensing

-
Seen products

-
Clicked products

-
Cart products

Interest ranking

01/11/2013

Τμήμα Πληροφορικής

11


User profile: data base table

01/11/2013

Τμήμα Πληροφορικής

12


User profile: xml

01/11/2013

Τμήμα Πληροφορικής

13

General Architecture of

User
-
Adaptive Systems

USER PROFILE

USER MODEL

ACQUISITION

USER MODEL

APPLICATION

INFORMATION ABOUT
U

ADAPTING TO
U

CBF

01/11/2013

Τμήμα Πληροφορικής

14

Why do we need to build cognitive
models of users


Examples:


User
-
adaptive mortgage consultant


User
-
adaptive computer store


Ticket assistant (e.g. COLLAGEN)


Recommender systems (hybrid models)



Employing cognitive models of users to improve
adaptation


Decide when explanations/help are needed


Tailor help to the user’s level of understanding


Decide what content to include and how to structure it


Filter recommendations and tune recommender algorithms

01/11/2013

Τμήμα Πληροφορικής

15

Expert
-
based user modelling

Expert knowledge

User model

Maintaining the user model

System
-
user interface

01/11/2013

Τμήμα Πληροφορικής

16

Representation of knowledge (1)


Frames


Frame: country


capital (town)


location


language


currency


climate


popular resorts
(resort)


Frame: town


population


nearest airport


location


Frame: resort


nearest airport


activities


cultural events

01/11/2013

Τμήμα Πληροφορικής

17

Representation of knowledge (2)


Frames


Country: Spain


capital: Madrid


location: Europe


language: Spanish


currency: Euro


climate: warm
continental


popular resorts:
Barcelona, Madrid,
Malaga, Toledo


Town: Madrid


population: 3
mln


nearest airport: Madrid


location: central Spain


Resort: Malaga


nearest airport: Malaga


activities: swimming, golf,
beaches


cultural events: Cathedral,
Museum of Picasso

01/11/2013

Τμήμα Πληροφορικής

18

Representation of knowledge (3)


Semantic networks


Powerful representation but difficult to infer

Spain

Country

Capital

Language

Continent

Town

isa

part
-
of

has

has

isa

has

Location

has

01/11/2013

Τμήμα Πληροφορικής

19

Representation of knowledge (4)


Logical systems


most often belief modal logic

Belief set


B(U,p1)

B(U,
¬
p2)

B(U,p3)

B(U,p3=>p4)

B(U,p5=>p1)

B(U,p6=>
¬p1)

Example Reasoners applied by the user


R1: (

)







Modus Ponens

R2: (

)








Modus Tolens

R3: (

)








R4: (

)










Apply the reasoners to derive additional beliefs of the user, called
DERIVED BELIEFS.

01/11/2013

Τμήμα Πληροφορικής

20

Beliefs, knowledge, misconceptions


Beliefs inferred based on user
behaviour


Correct beliefs considered knowledge


Incomplete and erroneous beliefs considered misconceptions

User belief set

B(u,p1)

B(u,
¬
p2)

B(u,p3)

B(u,p3=>p4)

B(u,p5=>p1)

B(u,p6=>
¬p1)

System belief set

B(s,p1)

B(s,p2)

B(s,p2=>p4)

B(s,p5=>p1)


Identify knowledge and misconceptions

01/11/2013

Τμήμα Πληροφορικής

21

Schema of User
-
Adaptive Systems

USER MODEL

USER MODEL

ACQUISITION

USER MODEL

APPLICATION

INFORMATION ABOUT
U

ADAPTING TO
U

01/11/2013

Τμήμα Πληροφορικής

22

Two steps



Content adaptation



what content is most
appropriate for the current user based on the user
model



Content presentation



how to most effectively
present the selected content to the user

01/11/2013

Τμήμα Πληροφορικής

23

Page
-
based approaches


Pre
-
defined pages


The adaptation mechanism selects the most
appropriate page


UM

Select pages

Show to user

Advantages and disadvantages?

01/11/2013

Τμήμα Πληροφορικής

24

Example: KBS Hyperbook


Adaptive
Information
Resources



Adaptive
Navigational
Structure



Adaptive Trail
Generation



Adaptive Project
Selection



Adaptive Goal
Selection

http://wwwis.win.tue.nl/asum99/henze/henze.html

01/11/2013

Τμήμα Πληροφορικής

25

Example: AHA


Navigation frame
(generated by the
system)



Content frame


combines fragments
prepared by authors




Inclusion/exclusion
of links;



Inclusion/exclusion
of detail

http://aha.win.tue.nl/

01/11/2013

Τμήμα Πληροφορικής

26

Dynamic approaches


Content adaptation:


Dynamic selection of content


Dynamic structuring of the content



Content presentation


Defining relevance and focus


Dynamic media adaptation

01/11/2013

Τμήμα Πληροφορικής

27

Dynamic content adaptation


Content automatically selected from:


Knowledge base, relevance measures

(e.g. ILEX, STOP)


Bayesian networks expressing causal probabilistic
relationships between variables from the domain (e.g. NAG)


User preferences model, importance measures

(e.g. GEA, RIA)


Content automatically structured:


Task
-

accomplished planners


Argumentation models


Conversation theories

01/11/2013

Τμήμα Πληροφορικής

28

Example: ILEX

http://www.hcrc.ed.ac.uk/ilex/


Domain Model



The Content
Potential


Text Structure


Syntactic Structure



Presentational
Forms



Representation of
Context



01/11/2013

Τμήμα Πληροφορικής

29

Example: GEA

(
Carenini

& Moore, 2001)


User preferences in
a hierarchical model
(e.g. house, location,
number of bedrooms)



Argument structure
tailored to user
preferences (uses
measure of relevance)



Level of detail will
differ for users or for
the same user at
different stages



01/11/2013

Τμήμα Πληροφορικής

30

Example: RIA

http://www.research.ibm.com/RIA/

Two different responses to the same query depending on user preferences

01/11/2013

Τμήμα Πληροφορικής

31

Dynamic content presentation

Maintaining focus and context



Focus


emphasise

the content that has been found
most relevant to the user



Context


allow access to less relevant content to
preserve context


Stretch text


Scaling fragments


Dimming fragments


Summary thumbnail

01/11/2013

Τμήμα Πληροφορικής

32

Follows the

“fish eye”

visualisation


Technique



Adaptation
of

an online guide

about cultural

events in Toronto:

http
://whatsuptoronto.com/



Example: scaling approach

01/11/2013

Τμήμα Πληροφορικής

33

Dynamic content presentation

Media adaptation: factors



User
-
specific features


Information
-
specific features


Contextual information


Media constraints


Limitations of technical resources

01/11/2013

Τμήμα Πληροφορικής

34

Dynamic content presentation

Media adaptation: approaches



Rule
-
based approaches


Using rules to define how to take into account the
media factors in media selection



Optimisation

approaches


Given the media factors, find the media
combination that produces the most optimal result

01/11/2013

Τμήμα Πληροφορικής

35

Example: RIA

Optimisation

adaptation

http://www.research.ibm.com/RIA/




The optimization procedure deals with: (1) suitability of the information to the
media; (2) increase
recallability
; (3) maintain presentation consistency

01/11/2013

Τμήμα Πληροφορικής

36

01/11/2013

Τμήμα Πληροφορικής

37

Καλό Βράδυ