User Modeling in IR LIS678 CoTeaching

1
User Modeling in IRLIS678 Co-Teaching

• Pei-Chia Chang
• Dr. Luz M Quiroga
• LIS678 Personalized Information Delivery

2
Required Readings

• - Belkin, Nicholas J. (2000). "The human
  elements Helping people find what they dont
  know." Communications of the ACM 43(8), 58-61.
• - Daniels, P. J. (1986). "Cognitive models in
  information retrieval An evaluative review."
  Journal of Documentation 42(4), 272-304.
• - Rich, Elaine A. (1983). "Users are individuals
  individualising user models." International
  Journal of Man - Machine Studies 18, 199-214.
• - Luz M. Quiroga, Martha E. Crosby Marie K.
  Iding (2004) "Reducing Cognitive Load". In
  Proceedings of the 37th Hawaii International
  Conference on System Sciences.

3
Additional Readings

• H Liu, P Maes (2004) What would they think? a
  computational model of attitudes - Proceedings
  of the 9th international conference on
  Intelligent User Interfaces
• Plinio Thomaz Aquino Junior, Lucia Vilela Leite
  Filgueiras (2005) User modeling with personas -
  ACM Proceedings of the 2005 Latin American
  conference on Human-computer interaction

4
Additional References

• Marvin Minsky(1988) The Society of Mind, Simon
  Schuster
• Amanda Spink and Charles Cole(2005) New
  Directions in Cognitive Information Retrieval,
  Springer
• Constantine, L. L and Lockwood, L. A. D.(1999),
  Software for Use, Addison Wesley

5
Why User Modeling?

• Belkin claims user modeling for Information
  Retrieval to help people find what they dont
  know.
• Quiroga et al. advocate user modeling for
  Information Filtering to reduce the cognitive
  load.

6
Mental Models, Marvin Minsky 1988

• Knowing facts, opinions, beliefs.
• Freedom of will

BODY
MIND
CAUSE
CHANCE
Free Will
7
Space of User Models, Rich, 1983

• One model of a single, canonical user vs. a
  collection of models of individual users.
• Models specified explicitly either by the system
  designer or by the users themselves vs. models
  inferred by the system on the basis of users'
  behavior.
• Models of fairly long-term user characteristics
  such as areas of interest or expertise vs. models
  of relatively short-term user characteristics
  such as the problem the user is currently trying
  to solve.

8
Canonical vs. Individual Models, Rich, 1983

• Canonical homogeneous user communities.
• Individualizedheterogeneous user communities
• Differenceflexibility

9
Explicit vs. Implicit Models , Rich, 1983

• Explicit manually configure the system
  parameters
• Implicitfeedback based personalizationintellige
  nt modeling

10
Long-term vs. Short-term Models , Rich, 1983

• Short-termcurrent goal
• Long-terma series of interactions

11
Rich, 1983
12
User Modeling Techniques , Rich, 1983

• Inferring individual factspatterns of user
  behaviorscondition-action rules
• Using stereotypes to infer many things at a time

13
Cognitive Information Retrieval Techniques,
Editor Amanda Spink and Charles Cole (2005)

• Implicit feedback
• Knowledge domain visualization
• Learning and training to search

From Amanda Spink and Charles Cole - New
Directions in Cognitive Information Retrieval,
2005, Springer
14
Case studies

• Grundythe use of stereotypes
• User Modeling with Personasuser modeling
  techniques
• What would they think (WWTT)affective user model
• Recommender systems

15
Grundy, Rich1983

• Individual user
• Implicitly constructed
• Longer-term
• Use stereotypes

16
Grundy , Rich1983

• User is asked to provide a few words and these
  words trigger appropriate stereotypes.
• If the system has enough information, books are
  recommended. Otherwise, it asks user for more
  words.

17
Rich1983
18
Rich1983
19
User Modeling with Personas, Plinio Thomaz
Aquino Junior, Lucia Vilela Leite Filgueiras
(2005)

• A persona is a user representation intending to
  simplify communication and project decision
  making by selecting project rules that suit the
  real propositions.

20
User Modeling Techniques Plinio Thomaz Aquino
Junior, Lucia Vilela Leite Filgueiras (2005)

• User roles
• User profiles
• User segments
• Marketing segments
• Extreme characters
• Persona

21
User Role, Constantine, L. L and Lockwood(1999)

• A collection of attributes that characterize
  certain user populations and their intentional
  interactions with the system

22
User Profile Plinio Thomaz Aquino Junior, Lucia
Vilela Leite Filgueiras (2005)

• Fictitious biographical summary, adding
  motivation, goals, and personalities. 1
• The understanding of user individual
  characteristics might be achieved by the user
  profile, including information related to age,
  gender, skills, education, experience, and
  cultural level.

1, from Brusilovsky, P. Methods and techniques
of adaptive hypermedia. User Modeling and
User-Adapted Interaction 6, 2-3 (1996) pp.
87-129. And Shneiderman, B., Designing the User
Interface, 1990.
23
User Segments Plinio Thomaz Aquino Junior, Lucia
Vilela Leite Filgueiras (2005)

• Groups of people who will use the services or
  product.

24
Marketing Segments Plinio Thomaz Aquino Junior,
Lucia Vilela Leite Filgueiras (2005)

• Establishes the marketing portions which reveal
  personal involvements with particular
  characteristics in common, according to the
  segmentation initial objective.

25
Extreme characters Plinio Thomaz Aquino Junior,
Lucia Vilela Leite Filgueiras (2005)

• The modeling of radical personalities will help
  cover all kinds of users.

26
Personas Plinio Thomaz Aquino Junior, Lucia
Vilela Leite Filgueiras (2005)

• The personas technique is based on data gathered
  through user research, mapping user archetypes,
  that represent a few important classes of users
  whose goals and needs a specific digital products
  or services.

27
Personas Implementation Plinio Thomaz Aquino
Junior, Lucia Vilela Leite Filgueiras (2005)

• Personal information
• Technical information
• Relationship information
• Opinion information

28
Persona Representation Plinio Thomaz Aquino
Junior, Lucia Vilela Leite Filgueiras (2005)
29
Persona Representation Plinio Thomaz Aquino
Junior, Lucia Vilela Leite Filgueiras (2005)

• His salary is R1200,00 per month.
• He has modest experience with computers and
  infrequent access
• to the Internet. Although he is not an expert
  user, he does not
• present a dodging behavior towards technology and
  computers.
• He reacts favorably when asked to used a computer
  for internet
• service.
• His use of governmental services is highly
  occasional, driven by
• some extreme motivation the need of being
  regular with
• government obligations, such as his income
  declaration to save
• money, like the electronic licensing of vehicles
  or to make more
• money, applying for a better job through civil
  service exams.
• When using the government websites, he has
  problems with the
• mouse and the printer, with the concept of URL
  and with the
• understanding of texts in the website

30
Persona Representation Plinio Thomaz Aquino
Junior, Lucia Vilela Leite Filgueiras (2005)
31
Persona Representation Plinio Thomaz Aquino
Junior, Lucia Vilela Leite Filgueiras (2005)
32
What Would They Think, H Liu, P Maes,04

• Generate a model of a persons attitudes from
  automated analysis of personal texts.
• Use affective memory system to build digital
  persona.
• Mining attitudes through natural language
  processing and commonsense-based textual affect
  sensing.

33
Computing a Persons Attitudes , H Liu, P Maes,04

• A bipartite affective memory system
• Mining attitudes from personal texts
• Predict attitudes using the model
• Enriching the basic model.

34
Affective Memory System , H Liu, P Maes,04

• Concepts, topics, and episodes are extracted
  from text and associated with their respective
  affective valence scores.
• Each pair constitutes a single exposure of an
  attitude, which accumulate in an affective memory
  system.
• Affective long-term episodic memory (LTEM)
• Affective reflexive memory

35
Affective Memory System , H Liu, P Maes,04

• Affective long-term episodic memory (LTEM)
• An episode is a basic unit of memory.
• Compute an affective LTEM as an episode frame.
• Episodes are content-addressable.

36
Affective Memory System , H Liu, P Maes,04

• Affective long-term episodic memory (LTEM)John
  and I were at the park. John was eating an ice
  cream. I asked him for a taste but he refused. I
  thought he was selfish for doing that.

37
Affective Memory System , H Liu, P Maes,04

• Affective reflexive memory

38
Mining Attitudes from Personal Texts , H Liu, P
Maes,04

• Digital persona can be automatically acquired
  from suitable personal text using natural
  language processing and textual affect sensing.
• Suitable text first-person, opinion-rich,
  well-balanced, explicitly episodic.
• Coping strategies were employed for dealing with
  erroneous appraisals.

39
Predict Attitudes using the Model, H Liu, P
Maes,04

• Predict the attitude by offering some affective
  reaction.
• Point-of-view agree/disagree

40
Predict Attitudes using the Model, H Liu, P
Maes,04
41
Enhancing the Basic Model , H Liu, P Maes,04

• Imprimer someone whose goals and attitudes we
  admire and hope to emulate.
• Heuristic approach was implemented to identify
  imprimers from a persons affective memory.
• Attach the imprimers affective memory to
  supplement the persons own affective when
  appraising new textual episodes.

42
Recommender Systems, Belkin, 2000

• Explicit term suggestion is a better way for
  recommender system supporting query
  reformulation.
• Understanding the contents of database
• Learning about effective vocabulary
• Being able to evaluate the relevance of an
  information object quickly and accurately