Posted:September 10, 2005

Revisiting Limited Keywords in Search

For many years, the apocryphal view (sometimes backed by data) is that most searchers limit their search queries to one or two keywords at most. In my updated search tutorial I cited some references for this, most importantly including this finding has held true since 1994 and that internal analysis at the NEC Princeton research center indicates this limit applies to professional researchers as well.[1]

A just-published study, "Analyzing User Behavior: A Case Study", by Chris Kutler and Ray Devaney, suggests these observations still hold. The authors analyzed more than 6 million searches conducted within The National Archives (TNA) in the UK (which includes public records for England and Wales), and indeed found average searches of from one to two keywords.

However, since the dataset available to the authors spanned from 2001 to 2005, they were also able to discover that search query length did tend to modestly increase to between two to three keywords as users became more familiar with the TNA’ s four underlying databases.

Though an improvement, even 2.5 keywords is hardly adequate and sophisticated query fodder.  Even with ranking adjustments such as for popularity (as Google does) or parts of document weighting (as many search engines do), there is simply no way to square getting accurate results with limited keywords against large document repositories.

Everyone wants magic search bullets, but it is still the case that it is the user that must pull the trigger.  Learning how to construct meaningful and content-rich queries is a skill that should be taught in every public school, much as typing was once done. 


[1] D. Butler, "Souped-up Search Engines," Nature 405:112-115, May 2000.

Schema.org Markup

headline:
Revisiting Limited Keywords in Search

alternativeHeadline:

author:

image:

description:
For many years, the apocryphal view (sometimes backed by data) is that most searchers limit their search queries to one or two keywords at most. In my updated search tutorial I cited some references for this, most importantly including this finding has held true since 1994 and that internal analysis at the NEC Princeton research […]

articleBody:
see above

datePublished:

Leave a Reply

Your email address will not be published. Required fields are marked *