SBA switches to a vertical search engine

The Small Business Administration will upgrade its Business.gov Web portal this week by installing a self-learning search engine that finds information relevant to its target audience of business owners. The specialty search engine, which performs what some call vertical searches, takes into account the needs of a select user group and the context of those users' searches. IT.com developed the SBA's vertical search engine. Mark Cordover, founder and chief executive officer of IT.com, said installation requires an understanding of the user domain. For the Business.gov project, SBA provided IT.com with an extensive list of business compliance resources. IT.com's proprietary search algorithm and other machine-learning algorithms will use SBA information and data continuously collected from user queries to learn which rules and forms are most relevant to Business.gov's users. 'We know the intention or the context in which someone is searching, and we take advantage of that,' Cordover said. Nancy Sternberg, program manager of SBA's Business Gateway program, said her colleagues asked focus group participants, who represented businesses of all sizes, how SBA could improve its e-government Web site. Participants overwhelmingly said they wanted more information on regulatory compliance and faster access to it. SBA wanted to satisfy business information needs that are not met by other search portals, Sternberg added. 'We needed something that would set us [apart] from Google or FirstGov,' she said. Besides serving the business community, the vertical search capability might help SBA and its partner agencies' employees keep up with the public's demand for information, Sternberg said. 'We do hope that it allows us to alleviate some of the burden on federal staff,' she said. 'They may not have to answer as many phone calls.' With IT.com's search algorithms, search results for any given query on Business.gov will be much narrower than what a user would see after entering the same query on Google or FirstGov. For example, if a Google search for 'retirement plans' retrieves 3 million hits, Business.gov's search would return closer to 300 hits, Sternberg said.She added that scaled-down search offerings allow entrepreneurs to spend less time searching for information and get back to running a business.'The bottom line is that small businesses and medium businesses really bear an enormous amount of the burden of compliance,' Sternberg said. The new search engine will find federal regulations and forms from nearly 100 agencies. The Labor Department possesses the majority of compliance information related to businesses, such as laws on equal opportunity employment and immigration documents. Sternberg said she wants to add personalized features, such as mybusiness.gov, which would let users set up a password, log in, save favorite searches and create calendars to remind them when forms are due. Sternberg said she will make changes based on business owners' suggestions. 'We're not out to do stuff just to do stuff,' she said. 'We will anchor back functionalities to federal requirements or things we hear in our focus groups.' She added that she doesn't intend to try to outdo commercial sites, but she will copy familiar features of popular commercial Web sites. ''A' for originality doesn't necessarily get you anything,' Sternberg said. 'We feel that our [visual] design is very similar to Google, and that's a great thing. If they're comfortable with that, why should we try to do anything differently?' Search experts say any method of categorizing data that improves the relevance of search results is a good thing. But each agency must determine which tool is the best for targeted users, said Kevin Lee, co-founder and executive chairman of Did-it Search Marketing.'Any government agency that has a narrow focus will benefit from a search engine system that can classify, tag and search documents,' Lee said. 'Nearly all federal and state agencies serve a narrow mandate and may benefit from a vertical search system,' he added.But agencies with a broad focus, such as the General Services Administration, would probably not benefit as much from that type of search, he said. The biggest advantage of vertical search is that it solves the data overload problem, Lee said. 'The government generates a huge volume of documents,' he said. 'Sorting, classifying them and making them accessible both internally and externally is a challenge met by vertical search, as well as some newer user classification systems, such as tagging and wikis.'Beyond machine-learning search solutions, government agencies should also explore tapping the knowledge of employees and the public, Lee said. 'Wikis and tagging by users are ways that existing data can be refined and made more accurate and relevant using the smartest systems of all: humans,' he said.