I recently spoke on Tuning Enterprise Search at the Gilbane SF show (you can view the tuning enterprise search slides here). In fine-tuning my slide deck, I found myself comparing the tools that the Google Search Appliance and Microsoft’s Search Server Express provide for tuning search results.
I’ll be posting my observations in three parts this week:
This one looks at how these two systems allow you to address the most popular searches.
The second looks at how these search solutions allow you to tilt the landscape of search results to meet the needs of your specific content ecosystem
The final post discusses how you can broaden search queries
As a general observation, Microsoft provides more power - particularly programmatic power - while Google opts for a simpler but less comprehensive approach.
Addressing the Short Tail of Search
I’ve written before about the shape of search and the leverage you can obtain by focusing on the most frequent search phrases – the short tail of search.
Both the Google Search Appliance and Microsoft Search provide simple interfaces for overriding algorithmically produced search results, and selecting a result - or group of results - to appear in first, second or third position (etc). Google’s KeyMatch interface allows you to select a search phrase and define a URL (and its title) to be returned. You have the option of returning the result if someone searches for the exact phrase you’ve defined, or a partial match.
Microsoft Search Server also lets you define “Best Bets” to achieve the same outcome, which provides a few additional capabilities
Defining synonyms to the keyword phrase
Defining the text to appear below the link
Specifying start and end dates for the Best Bet
If you do nothing else to improve the effectiveness of your intranet search engine, use these tools:
Review the 100 most frequent searches this month
Work with subject matter experts to define the “best” results
Do it again next month (in other words, be consistent)
In a very short period, you will be delivering much more effective results to at least one-third of your searchers.
Language of posts
Posts by date