Evaluating the Online Searching Behaviour and Activity of Users of the NeLI

1
Evaluating the Online Searching Behaviour and
Activity of Users of the NeLI
Gemma Madle, Patty Kostkova, Jane
Mani-Saada Healthcare Digital Libraries Workshop
2004
City eHealth Research Centre
2
Talk outline

• Introduction to transaction log analysis
• Methods
• Results
• General usage statistics
• User statistics
• Most common topics viewed
• Search/browse behaviour
• Other results
• Future research
• Contacts

3
1 Transaction log analysis

• What data do we get?
• Path of user through a site
• IP address of user
• What conclusions can we draw?
• How users are navigating/searching the site
• Most popular pages
• Linked pages, relationships between pages
• What are the limitations?
• Identification of crawlers/spiders
• Problems relating IP addresses to users, use of
  proxy servers
• Time spent on last page not recorded
• Use of back/forward buttons
• Limited information about the user
• Periods of inactivity within sessions

4
1 Transaction log analysis

• Why is this useful for NeLI?
• Can identify how people are using the library,
  navigation paths they prefer
• Can identify which pages are being viewed and
  therefore which pages are not
• Can link patterns in use to current epidemics/
  infections in the news
• Can gain a picture of where users are from to see
  whether we are attracting our target audience

5
2 Methods

• 1. Collect access logs with the following fields
• (1) The IP or hostname of the origin of the
  request, (2) date and time of the request, (3)
  the type of request, (4) the page requested, (5)
  the returned status of the page and (6) the
  number of bytes transferred
• 2. Clean data
• Remove hits on picture files, remove identifiable
  robots, separate time and data into separate
  fields
• 3. Import data into Microsoft Excel by month
• Look up IPs hostnames (identifying accesses
  from medical establishments)
• Group accesses into sessions (30 minutes of
  inactivity or change of IP starts a new session)
• 4. Analyse to gain statistics of use

6
3 Results General Use Statistics
7
3 Results User Statistics
8
3 Results User Statistics (Hospital/Medical
based Users)
9
3 Results Most common topics viewed (All users)
10
3 Results Most common topics viewed (Hospital
Users)
11
3 Searching/browsing behaviour
12
3 Other results

• Use of NeLI may be linked to Infectious diseases
  in the news e.g. hits to legionella pages
  increased during an outbreak in August 2002
• Average no. pages accessed per visit 2.11
• Entry page is home page for only 37 sessions so
  users may be bookmarking or entering via search
  engines directly to topic pages

13
3 Future Research

• Need to compare results with questionnaire survey
  for more detailed results (now underway August
  2004)
• New free text search engine added August 2003
  (analysis underway August 2004)
• Adaptation to structured search page to enable
  collection of keyword data
• THANKS PATTY FOR PRESENTING ?

14
4 Contacts
Gemma Madle g.c.madle_at_city.ac.uk City eHealth
Research Centre http//www.city.ac.uk/cerc Nationa
l electronic Library of Infection
(NeLI) http//www.neli.org.uk