It may not yet be self-aware, but Google's enterprise search hardware is now capable of learning.

Google on Tuesday announced the availability of 10 new features for the Google Search Appliance (GSA), the company's rack-mountable search server for businesses.

Chief among the additions is a new algorithm called Self-Learning Scorer, which analyzes employee clicks and behavior to improve the relevance of search results.

If, for example, most employees searching for a given term click on the third search result, the GSA will place that term higher on the search results page for future searches.

On the Internet, Google gets highly relevant search results in part because of its PageRank algorithm, which analyzes Web links and treats them as votes for relevance. On corporate intranets, a rich link structure is absent, which makes search relevance harder.

The Self-Learning Scorer system should help compensate for the absence of Web links inside corporate firewalls.

Other new features include native SharePoint integration, which makes indexing SharePoint content 10x faster, Lotus Notes connectivity, through Google's Enterprise Labs, and a new universal login system.

According to IDC's "Worldwide Search and Discovery Software 2009-2013" report, Google's estimated enterprise search revenue was $120.8 million in 2006, $142 million in 2007, and $154 million in 2008, about 7.4% of the market that year. That puts Google in third place, behind Autonomy and Microsoft.

The business search software market grew 19% in 2008, outperforming the general software market and the economy as a whole. This year and in the next few years, IDC expects growth in the 7% to 10% range.

While Google's early versions of the GSA lacked security, permissions, and other features valued by enterprise customers, says IDC analyst Sue Feldman in the report, Google has addressed the GSA's shortcomings and continues to add valuable enterprise features.

"In other words, the GSA is moving upstream, while keeping the price low, and continuing to disrupt the confused search market," says Feldman in the report.

InformationWeek Analytics has published a report on the 10 steps to effective data classification. Download the report here (registration required).