Search and Optimal Selection of Recommendation Systems on the Internet: A Comparison between a Knowl

1
Search and Optimal Selection of Recommendation
Systems on the InternetA Comparison between a
Knowledge-based and a Content-mining Approach

• Nicole Mitsche
• Institute for Tourism and Leisure StudiesVienna
  University of Economics and Business
  Administration
• NLTeC, Champaign, 4. September 2002

2
Outline

• Starting the story
• General background
• Prototype Quo Vadis
• Research question and scientific contribution
• Additional background for research questions
• What is going on?
• Future steps

3

• Starting the story

4
Where the story started

• What could they do there
• in their leisure time?
• Soccer game ?
• Bars ?

Spain
Barcelona

• Didnt find the web site
• Long time to get information

No soccer game on that weekend!
5
Problems?

• Did they struggle because of the huge amount of
• information on the internet?
• Was it a good entry point for this search?
• Did they use appropriate words to define their
• search?
• Was it possible to define their search in a way
  that can lead to a result?
• Did the search engine fail?

6

• General Background

7
Searching the Web
Search engines (open text queries) e.g. Google,
Altavista
Portals/Directories (tree format) e.g. Yahoo,
dmoz, Travigator

• Out of date indices
• Invisible web (connected databases, scripts)
• Lower quality than specialized search engines
• Advanced search options

• Limited in size
• Human evaluators
• Subjective judgments
• Inconsistent categorization

8
Travel Web Sites

• Information Variation

• Specialization
• Destination
• Transport
• Accommodation
• Special Interest

9
Travel Choice Decision Making Behavior

• is a temporal, dynamic, successive and
  multi-stage process in which certain decisions
  made in an earlier stage will condition decisions
  made in a later stage.
• (Jeng and Fesenmaier, 1999 134)

10
Travel Choice Decision Making Behavior

• is a temporal, dynamic, successive and
  multi-stage process in which certain decisions
  made in an earlier stage will condition decisions
  made in a later stage.
• (Jeng and Fesenmaier, 1999 134)

Information Search
Information Evaluation
Choice
Post Choice
Booking
11

• Prototype Quo Vadis

12
About Quo Vadis
Quo Vadis Travel Recommender System for Travel
Recommender Systems
to assist travelers in all different
decision stages to guide them to the best
matching web site
13
Quo Vadis System DesignTwo main Databases
14
User Profiles
Question Pool
15
System Profiles
System Profiles
Attribute a1 a2 ... an
Evaluation and Categorization
System A System B ....
Data Repertory
Digitalization

Real World
Supplier 1
2
3
4
5
16
System Design
17

• Research question and scientific contribution

18
Concentration on
User Profiles
Learning
Feedback
Evaluation
System Profiles
Attribute a1 a2 ... an
Evaluation and Categorization
System A System B ....
Data Repertory
19
Research Question

• A Comparison between a knowledge-based
• and a content-mining approach.

Or, does a complex, intelligent approach based
on content mining lead to better recommendation
results than a simple, rule-based expert system?
20

• Additional background for research questions

21
Classification of Travel Web Sites

• Quantitative Methods
• e.g. Web Crawler, Log Analyzer
• Automatically
• Hard facts measurable at site
• Accessibility and Size
• Larger sample in less time
• Content (restricted, insufficient)
• Objective

• Qualitative Methods
• e.g. Human evaluator
• Manually
• Generally more detailed
• Content in every specialization
• Time intensive
• Reliability problem
• Subjective

22

• What is going on?

23
Comparison between a Qualitative and a
Quantitative Evaluation

• Qualitative Assessment
• Contains questions about transport, accommodation
  packages and other general questions
• 175 websites, mainly international sites which
  have a primary focus on information search and
  booking

• Quantitative Evaluation
• Web Crawler, which collects all the links of the
  web site
• Content Analyzer, who collects all content
  information and analyzes it concerning the
  special requirements

24
Qualitative Assessment for Travel Web Sites
25

• Future Steps

26
Future Steps

• Design the two approaches, taking different
  system profiles in account
• Evaluation strategy
• Finalize the prototype(s) and run the system
• Analyze data
• Finish thesis