User Modeling for Personalized City Tours

1
User Modeling for Personalized City Tours

• Josef Fink Alfred Kobsa

Robert Whitaker
2
Problem Discussed

• develop a user modelling server that offers
• services to personalized systems with
• regard to the analysis of user actions, the
• representation of assumptions about the
• user, and the inference of additional
• assumptions based on domain knowledge
• and characteristics of similar users

3
What is a User Modelling System

• Application that acquires and stores evidence
  about a person
• Identifies consistencies and inconsistencies held
  about a user
• Answer queries of an application about the user
• Assist in the personalisation of the users
  experience

4
Background

• Deep Map Project
• Group of projects aimed at developing
  personalised and mobile tourist guides
• Web Guide Sub Project
• Gives personalised tour recommendations for the
  city of Heidelberg

5
Web Guide Project
No Personalisation
Personalisation
6
Work Performed

• Explored the characteristics of user modelling
  systems already developed
• Developed a generic User Modelling Server
• Demonstrated a customisation of the server for
  the tourist domain

7
User Modelling System Background Study

• Generic Requirements
• Learn interests and preferences of users based on
  their usage of the application
• Predict interests and preferences of users based
  on those of similar users
• Infer additional interests using domain knowledge
• Supply an authorised application with current
  information about the user

8
User Modelling System Background Study

• Additional Requirements need for domain
• Need to take into account demographic data
• A priori knowledge about the user is generally
  not available
• Explicit acquisition of data at runtime is
  restricted to a short interview
• Adapting to change should be relatively quick

9
User Modelling Server Architecture
10
External Clients

• Type of client is endless
• Type of client was important when designing
  storage of data
• Uses directory management system instead of
  conventional database

11
Directory Component

• Four types of models
• User Models
• Usage Models
• System Models
• Service Models

12
User Modelling Components

• What makes it generic
• Components that allow the system to be customised
  to the specific domain
• Examples shown
• User Learning Component (ULC)
• Mentor Learning Component (MLC)

13
User Learning Component

• Uses Univariate Significance Analysis

14
Mentor Learning Component

• Uses Spearman Correlation
• Three Steps
• Findings similar users
• Selecting Mentors
• Computing Predictions

15
Summary
Problem Solution
Handle user modelling requirements of the domain Developed user modelling plug in
Network Connectivity Directory Storage and distributed storage
Act on user events Through event notification
No training data initially Use demographic details learning and modelling