Internet search just keeps getting better and better, but intranet search is still in a rut. Google's intranet search product, Google Search Appliance, has gotten good reviews and claims features that let users search not only HTML documents, but relational databases as well. But the fact remains that Google's success on the consumer-dominated Web can't be easily replicated in the enterprise environment, because Google's consumer offering is so dependant on the tracking of links, and links are relatively rare in the documents found on intranets.
From a technology point of view, there are three keys to effective intranet search: metadata, taxonomies and controlled vocabulary. Unfortunately, all three impose huge -- some would say unworkable -- burdens on the people who are meant to use the systems based on these technologies.
Taxonomies are a good example. They require that people agree on what the categories should be -- a problem in itself, e.g., should they be user-created or imposed from the top? With metadata, there are similar problems. Should the tags be applied by unskilled workers after the fact, or by knowledge workers at the time of document creation? In the first case, the quality of the tagging suffers; in the second, it's felt that knowledge workers are wasting valuable time.
When all is said and done, the real problem is simple: Doing what it takes to make effective intranet searches possible is just too much work.