Choosing an enterprise search product is often fraught with compromise. If you, say, pick something with a simple search interface to appease users, administrators will likely be restricted in indexing databases or customizing results.
Vivisimo's Velocity hasn't asked organizations to make such concessions. Its search engine indexes content as is, so you don't have to preprocess or reformat documents. The content integrator allows users to perform simultaneous searches through diverse sources -- internal documents, intranets, Web, or syndicated news feeds. Last, the clustering engine organizes all these search results into categories intelligently made from words and phrases in the results, not some arbitrary popularity ranking.
Still, Version 5 has something new for both end-users and system managers, making it an even stronger search solution for Fortune 1000 companies and government agencies. To improve the user experience, Vivisimo added role-based search so that enterprises can target results, for example, to employees in sales or to human resources.
Yet Velocity 5's biggest changes are in the less visible underlying technology and administration. Topping the list are connectors for databases, Microsoft SharePoint portals, and EMC Documentum Docbases. Because these sources usually require authentication, Vivisimo Velocity 5 easily connects to LDAP or Active Directory servers and restricts viewing documents (or content sections) based on a user's rights. Moreover, administration is more straightforward compared with Version 4.5.
Gentlemen, start your engine
In approximately 30 minutes I'd installed Velocity 5 under Red Hat Enterprise Linux 4. Vivisimo consolidated some of the management functions and reorganized parts of the Web administration interface, which compressed down to a few hours the process of creating my test scenarios -- searching an intranet, external Web sites, two SharePoint portals, and several Microsoft SQL databases. More elaborate customizations required a few days; Vivisimo also offers professional services and commits to completing most complex projects within 90 days.
Crawling and indexing sites or documents is as simple as selecting the type of resource (such as a database) and pointing to the server. The control that IT staff has over content extraction and normalization -- without much effort -- is significant.
Using a simple form, for example, I adjusted the HTML converter so that the crawler ignored common navigation that appeared on each page but gave more weight to link and tag density. I also boosted the priority of certain pages that I wanted to appear at the top of results. These tweaks, along with Velocity's own relevance-ranking algorithms (freshness, term proximity, link analysis), generated results that were more accurate than other products I've tested.
Similarly, I adjusted XSL templates to change the appearance and behavior of the search interface and results page. Things got even more interesting when I used "formula-based sorting" to retrieve very specific results, a feature that truly improves the search experience. For instance, based on metadata, I created graphical sliders that allow users to quickly search a Web site's product section and pick servers that employed specific processors. Or, for a real estate site, sliders could allow users to easily select homes in specific price ranges, number of rooms, or land size.