Understanding the Information Search Process within a tourism - IFITT

Document Sample
Understanding the Information Search Process within a tourism - IFITT Powered By Docstoc
					Understanding the information search
  process within a tourism domain-
       specific search engine

               Nicole Mitsche
          University of Sunderland

• Research aim
• Background
• Methodologies
  – data set
  – taxonomy for aggregating keywords
• Results
• Limitations
• Conclusions

                  ENTER 2005 Research Track   2
                Research Aim

• investigate different user search patterns and
  the adoption of different strategies to develop
  a typology of search patterns
• examination of keywords used at the
  beginning of the a search and developing a
  framework where they fit in

                  ENTER 2005 Research Track         3
      Information search patterns of
• General models to discover process and influencing
  factors (Fodness and Murray 1997, Vogt and
  Fesenmaier 1998, gursoy and Mc Cleary 2004)
• Multi-faceted factors in choice (Jeng and Fesenmaier
• different orientation of travellers in their search
  (Grabler and Zins, 2004)
   – 6 decision styles (price, destination, accommodation,
     activities, highly-predefined, recommendation oriented)

                       ENTER 2005 Research Track               4
    Information search on the web

• open text queries
  search is conducted by the user in his own words
• categorical search
  uses predefined categories which are developed by
    experts or on the base of user data
• combination search
  using keyword search within categories; mainly used
    within specialised search engines/search services

                   ENTER 2005 Research Track          5
       Simplicity in keyword search

• users submit only few queries per search, most of
  them only one
• these are very short
• additional features as relevance feedback or Boolean
  operators are rarely used, very often they are
  unnecessary and user apply them wrongly
• skewed keyword distribution
  few terms are repeatedly used and a lot of terms only
• searches in the morning are longer, more queries are
                                    Jansen et al 2000, Ozmutlu et al 2003b

                    ENTER 2005 Research Track                            6
     Trend of multitasking searches

• user submit several queries to subsequent
  topics which are suggested to be related to
  each other
• automatic topic identification helps to
  discover switching of topics (but methods
  need still improvement)

       He et al 2000, Ozmutlu et al 2003a, Ozmutlu and Cavadur 2004

                      ENTER 2005 Research Track                   7
Search strategies - external influences

• user goal and search purpose is a major
• influenced by interactive intentions
  identifying, learning, finding, locating, record,
    keeping, accessing, evaluating and obtaining)
• influenced by knowledge
  specific, common, area/location, general

                    ENTER 2005 Research Track         8

• basis data set: internal log file of the search engine
    –   IP adress
    –   language
    –   city code
    –   keywords
    –   kind of task executed
    –   count of links
• September 2003
• 22.844 tasks transferred
  into sessions

                            ENTER 2005 Research Track      9
   Taxonomy for aggregating keywords
Accommodation                                        Information
    Hotels                                                  Information, Brochures
    Other accommodation types                               Guides
    Accommodation named                                     Special Information (e.g. weather)
Attractions                                                  Pictures
      Sights general                                        Languages
      Sights more special (e.g. museum)             Hospitality
      Sights named                                  Activities (Leisure and Sports)
      City cards                                    Events
      City tours                                          Events general
Destination (Places)                                       Evens named
      City name                                           Culture and music
      Country name                                  Target groups
      Other places/surrounding areas                        Business, Jobs, Congress
      Special places named (address, centre)                Shopping
      Map                                                   City breaks
Transport                                                    Diff. target groups

                                    ENTER 2005 Research Track                                  10
      Importance of the homepage

• clicking and re-clicking on the homepage is
  the most important task (32.4%)
• session based: within each session going
  back to the homepage (10%)
• This behaviour suggests, that the homepage
  is not only a starting point but also a
  restarting point

                 ENTER 2005 Research Track      11
        The starting point within a session

    Searches a particular city                           24.6
    Searches a particular keyword                         2.4
    Searches particular city and keyword                  6.3

    Clicks on suggested cities                           15.8
    Clicks on suggested keyword                           9.8
    Clicks on suggested keyword in a particular city      1.2

    Visits a particular webpage                            .9
    Visits a city’s server homepage                      1.4
    Requests cities from A-Z                             27.5
    Requests nearby cities                                 .2

                             ENTER 2005 Research Track          12
        The starting point within a session

    • categorical search
        – largest amount of searches
        – search for particular city, list of available cities
    • inspirational search
        – suggested cities
    • keyword search
        – small proportion
    • combination search

                           ENTER 2005 Research Track             13
         What are users looking for?

• Accommodation
   – “hotels” and “hotel” are the most frequent keywords used
• Maps
   – “map” as a keyword (12.5%)
   – at the beginning of a session 5.1% of users already look for
     a cities map
• “information” (6.5%)
• “guide” (4.4%)
• “events” (3.5%)

                        ENTER 2005 Research Track                   14
    Skewed distribution of keywords

• The 5 most popular keywords comprise half
  of all keywords searched at the beginning
• 19.5% of all keywords are used only once
• keyword frequencies depend if they are
  conducted for all cities or if they are used for
  search within a particular city (p=0.000)

                   ENTER 2005 Research Track         15
   Comparison of keyword topic between search
  within a particular city and open search (p=0.000)
         Keyword topic           All cities                Search within a
                                                            particular city

                         Freq      %                 Freq         %
Destination              140       31.9%             246          27.1%
Accommodation            142       32.3%             253          28.0%
Attractions              27        6.2%              95           10.5%
Information              81        18.5%             165          18.3%
Transport                4         .9%               48           5.3%
Hospitality              1         .2%               23           2.5%
Activities               1         .2%               2            .2%
Events                   22        5.0%              43           4.7%
Target groups            22        5,0%              29           3,2%
Total                    439       100,0%            903          100,0%

                         ENTER 2005 Research Track                            16
      specialised / general search

• adaptation of the keyword taxonomy into
  specialised and general search
• significant result (p=0.000)
  – Specialised keywords are more frequently used
    for searches within a particular city
  – General searches are more frequently used for
    global searches

                  ENTER 2005 Research Track         17
             Patterns over time

• cluster analysis on the first 10 tasks within a
• homepage clicks eliminated
• cluster analysis with best discriminatory result

                  ENTER 2005 Research Track      18
    Description of search patterns derived
          from the cluster analysis
    Search pattern type                                                                       Size

1   Did not do much and this very indifferent                                                 32,9%
2   Searches a particular city                                                                18,0%
3   Requests cities from A-Z and clicks on suggested city                                     15,0%
4   Searches a particular city, clicks on a suggested keyword in particular city and visits
    particular web page                                                                       11,9%
5   Clicks on suggested city, requests cities from A-Z and requests national map
6   Requests cities from A-Z, clicks on suggested city and clicks on suggested keyword in
    particular city                                                                           5,9%
7   Requests cities from A-Z and street map list, clicks on suggested city                    3,8%
8   Clicks on suggested city and on suggested keyword in particular city, visits the city’s
    server homepage and particular web pages                                                  3,4%
9   Searches a particular city, clicks on the suggested city and requests national map and
    nearby cities                                                                             3,0%

                                        ENTER 2005 Research Track                                    19
         Summary and conclusions

• general strong orientation on the city - users look for a
  particular city, browse or are getting inspirations for
  choosing a city
• keyword search highly used in combination with a
  particular city
• accommodation and destinations (places) are the
  categories user are searching for most
• keyword search is dominated by few keywords and
  few categories (accommodation, destination and
  information)  skewed distribution
• The homepage is not only the starting but also a re-
  starting point within the search process

                      ENTER 2005 Research Track           20

• Specific example with specific result
• Ongoing changes, implementation and further
  development within the search engine might
  change results partly
• Time limitations
• User limitations

                ENTER 2005 Research Track   21
     Potential future developments

• Innovative space within inspirational path
• Further developments in the combination
  search of keywords within categories
• Using the knowledge of skewed distribution of
  keywords in improving search engine
• Implementing the knowledge about the
  homepage as a re-starting point in its design

                 ENTER 2005 Research Track    22

Shared By: