|
Challenge |
Overview
FindLaw, the most frequently visited legal site on the Internet, provides a comprehensive set of legal resources for legal professionals, businesses, students, and the public. These resources include LawCrawler®, a full-text Web search engine powered by Google and geared specifically to legal research, as well as cases and codes, legal news, an online career center, and community-oriented tools.
Fast, accurate search results are essential to giving FindLaw's users access to the data they need — a significant challenge, since new legal resources are added to the site daily.
The Challenge
According to Cicely Wilson, FindLaw's librarian, relevant search results are extremely important to FindLaw users. "The previous search application we used had high recall but low precision. We would get a lot of search results, but they were not spot-on in usability. Most users will not go through more than one or two pages of search results — and if they don't find what they're looking for, they give up. We needed to find a better balance."
Flexibility and ease of configuration were also key concerns. FindLaw is searchable not only from FindLaw.com but also from a number of related portal sites, each with its own orientation and area of emphasis. In order to provide search results tailored to the specific requirements of these portals, FindLaw senior software engineer Jeff Rogers explains, "we wanted fine-grained control of what's searched and what comes up in the search results in any particular case." And Rogers wanted that control without having to filter searches at the script level, so as to avoid the significant performance degradation that such filtering entails.
Finally, the search solution had to meet the site's operational needs. FindLaw's extensive and frequently updated resource index contains millions of pages, primarily in HTML and PDF formats. FindLaw needed an efficient way to index and selectively crawl specific sections of the site to meet the needs of FindLaw's broad user base. The site's content resides in an NFS-based networked storage architecture, with all content cross-mounted across all webservers. The production network, situated in a single cage at a colocation facility, includes numerous frontline webservers serving static content pages and additional servers providing personalized content. The FindLaw site delivers more than 50 million pageviews a month to more than 2.5 million unique users.
The Solution
Choosing Google's GB-1001 enabled FindLaw to provide more precise results for its users. "The Google Search Appliance provides better balanced results, which usually appear within the first page of results," says Cicely Wilson. "With the algorithms Google uses to do their indexing, you find what you're looking for within the first few pages. It's a combination of high recall and high precision. This is a delicate balance, but when you find it, the product is extremely good for the user."
The Google Search Appliance also allowed senior software engineer Jeff Rogers to tailor search results to the specific needs of FindLaw's portals without incurring the overhead — and the attendant performance costs — imposed by script-based search filtering. Rather than specifying complex control parameters in scripts that would run once for each search conducted, the Google Search Appliance enabled FindLaw to use subcollections to define these restricted searches, giving Rogers an easy way to specify the searches coming in from FindLaw's various portals while giving users faster results.
The Google Search Appliance provided other benefits as well, according to Rogers. FindLaw needed to keep certain parts of the site off-limits to search users. Unlike FindLaw's previous search application, "the Google Search Appliance enables us to exclude that content from the crawl." It also made it easy for Rogers to time the crawls to correspond to FindLaw's publishing schedule, enabling him to capture new content for FindLaw's search users without crawling the site unnecessarily.
End Result
"One of the great features of the Google Search Appliance is that it's fairly seamless," says Wilson. "It performs great searches and you get the results you want. Just type a word in and it works. We trust Google. It works well and our users are satisfied."
About the Google Search Appliance
The Google Search Appliance is an integrated corporate search solution that extends Google's award-winning search technology to intranets and websites. The Google Search Appliance is available in three models: the GB-1001 for departments and mid-sized companies, the GB-5005 for dedicated, high-priority search services such as customer-facing websites and company-wide intranet applications, and the GB-8008 for centralized deployments supporting global business units.
Contact sales for more information.
|
