Syntax's built in search engine has had search suggestions for a while now, you can see it in action on NCLR's site. These "best bet" recommendations are a simple match for the user's search query in the name field of each object. This makes sense since for relevant content the keywords are often within the title of a press release, publication, or person.

Search tools has looked into search suggestions, and come up with eight principles for good search suggestions.

Search suggestion systems complement search engine results with human judgment, providing links to specific landing pages for common search terms where the algorithmic results may be confusing or unavailable.

The list is an excellent set of recommendations that has me thinking about adding some intelligence to our search engine. Particularly the idea of of logging search queries and using that data to understand what users are searching for and fine tune recommendations. I could see a system that:

1. Tracks search term frequency by month, incrementing the count for a search term in a simple table with fields for the search query, the month, and count. You could then use that data to see what are popular search terms in a given month and also analyze how terms change from month to month.
2. Use the log to provide suggested search terms as the user types. For example, if the type in "Leg", a small drop down appears with "Legal" or "Legislation" as possible matches. Using jquery and its autocomplete plugin, this wouldn't be technically difficult to implement
3. Provides a way for site administrators to view search query data. A simple log can provide some insight into what user's are looking when they visit.

You'd probably need some sort of filter so that profanity and offensive terms don't get logged in the system and shown to other search users. Since search is a common navigation tool, improving the search interface would improve the user experience.