Transcript Best Bets: Improving Search to High Demand Resources

Best Bets:
Improving Search to High
Demand Resources
Tito Sierra
NCSU Libraries
Code4Lib 2007
Best Bets Concept
1. Analyze your search logs to see what
people are search for most
2. Build a custom index for these items
3. Integrate it with your website search to
make your users happy!
Search Behavior at NCSU
Libraries
Top Search Terms
1.
2.
3.
4.
5.
6.
citation builder
circulation
copyright
refworks
web of science
jstor
Search Stats
• Top 100 search queries account for
30.2% of all searches
• Top 50 search queries account for
26.3% of all searches
Top 20 search queries =
21.8% of all searches
Best Bets Implementation
Best Bets at NCSU
Best Bets at NCSU
Best Bets at NCSU
• Quick Search module focused on high
demand resources
• 50 total Best Bets
• Mix of
• Popular databases (‘Web of Science’)
• Services (‘Course Reserves’)
• Library information (‘Library Hours’)
Implementation
• Small XML files
•
•
•
•
<title>
<description>
<link>
<keywords>
• Indexed with SWISH-E
Key Benefits
• Force popular resources to the top of
search results
• Best Bets for items that aren’t included
in your Web search crawl
• Custom <keywords> field enables
matching on misspellings or synonyms
(e.g. Lexus Nexus)
• Custom description displays
Best Bets Usage
Quick Search click tracking by module
Best Bets Usage
Best Bets account for 16.6% of all search
results clicks in the last 12 months
Advanced Analytics
• Local logging of all
serves and clicks
• High click % is good
• Low click % is bad
• Use click % data to
improve Best Bets
performance