It covers stemming and morphology, then goes onto inferring context and paraphrasing.
Quite interesting, at least I think so. If anybody else is interested I am tempted to put together a talk on domain-specific searching - i.e. building searching based on an application or collection of information.
Things that can't be done in WWW searching because of scale, difference and not having control/knowledge of the information in advance mean that you can't do clever tricks like paraphrasing or lexical analysis, etc that are covered in the article.
N.B. a search for 'perl' at acmqueue gives 0 results. hello? even DDJ manages something with some mention of perl every few months