Beyond Search Results—Get the Whole Picture and Reap the Rewards
Search engines today are built with an innately document-centric view of the world. This is hardly surprising, given how the technology has evolved, but it is often not the paradigm best utilized for providing an answer to a user's search request. When users search, they are searching for lots of different kinds of things.
In many situations, a search is performed not to identify a document, but to identify another kind of object: a product, a person, a project, etc. A superior search system enables users to quickly identify documents, but also to identify these other kinds of entities. This is accomplished by incorporating a view of the world that contains more objects than just documents.
How do search systems that contain representations of different kinds of objects differ from the traditional search systems on the market today?
Consider a commerce website that sells a particular kind of product—in this case let's say fasteners, or ways of joining two pieces of material together. Each product (fastener) has a title and a short description. These represent a kind of document that traditional search systems can index, search and, hopefully, find when keywords contained in the query are also contained in the description.
But suppose that you, as the user, are looking for a particular kind of fastener—a nut and bolt—to join two pieces of metal together. You're looking for that particular kind of fastener not because you need to use a nut and bolt, but because that is what came to mind. In a traditional search system, when you search for "nut and bolt" and variations thereof, nuts and bolts will be returned, but not screws, glues or rivets. If you can't find an appropriate nut and bolt, you're at a loss, and the site loses your business.
Now consider what would happen if the site deployed a more advanced search system. The site can avoid losing your business by returning a suitable substitute, in this case perhaps metal screws or epoxy. To do so, the search system needs to be able to have an idea of (a.) what products belong to the fastener class; (b.) what fasteners are useful for joining metal; and (c.) what the prices of the items are. One way to solve this problem would be to add this information to every document, a time-consuming process that results in high maintenance costs as inventory and prices change over time. Another way is for the system to contain an understanding of a fastener object, building a virtual representation of such an object from the various systems (product catalog, joining FAQ, product inventory and pricing) involved. In this way, as pricing and recommended joining methods change over time, the information the search system returns remains current—at a low maintenance price tag.
In the example above, more than just the document is being searched on. Information is being pulled from other systems to represent the "fastener" object and substitutes can be provided whose own descriptions might not contain any of the keywords contained in the first product's description. A substitute product might be suggested because it is also a fastener, can be used for metal, and is about the same price, for example.
The benefit of the more advanced search system building a virtual fastener object is readily apparent. This is only the most basic case, however. In a customer-facing environment, like an online store or a customer service center, many of the relationships described above will be outlined explicitly—a fastener class will be explicitly created, for example. The benefits of reaching beyond simple document search become even greater in many internal search applications, where these object classes have grown up organically and are implicit in the way the organization works.
Implicit objects are all around us in organizations. One common search we're all familiar with is to identify someone in your organization who can help with a particular issue. Your colleagues have written documents, written e-mails, billed time, worked on particular accounts and done myriad other things that provide a virtual representation of who they are. All of this information can be exploited in the manner described above to identify a person to help with a particular task, by the search system pulling all of these disparate documents from different systems into a virtual person object. As a result, when trying to find who was involved in that Enron gas deal, searching for "Enron gas deal" can help you identify all those involved, in an explicit manner. The same information can then be used to identify other deals that were similar, perhaps one in Europe done by Gazprom. By knowing that the European deal was similar, you can identify people in the London office who can help with a new gas deal you're working on.
Whether finding the right product, finding the right person, or finding a similar deal is your aim, when considering your information infrastructure, make sure your search system can get the whole picture and you will reap the rewards.
Recommind's (
www.recommind.com)MindServer product suite delivers search results in full context, enabling users to retrieve information on a product or project basis or to identify colleagues with particular expertise. With faster access to the right information, organizations such as Bertelsmann, DuPont, NIH, and leading law firms Morrison & Foerster and Cravath, Swaine & Moore are using MindServer to save time, increase sales, and increase the value of information assets.