How do you define semantic search? Plenty of companies, of course, are staking a claim to incorporating semantic technology into their search engines – Bing, Google, Hakia, Kosmix, Powerset and Yahoo. That's all welcome, but TextWise CEO Connie Kenneally thinks it might be better for users exploring the landscape to focus less on the mechanics of how semantic technology is powering search and more on what the goals of searches are.

“I tend to not look at semantic search in the technology sense but in the application sense,” says Kenneally, who will be speaking on a panel about the evolution of semantic search at WebMediaBrands’ upcoming Web 3.0 conference. “If someone asked me where to use semantic search for their applications, I would try and understand what they want to do, what is their objective. There are many ways to skin a cat in semantics and [the decision should be a] function of what the application is that you are driving and what approach might best fit with that.”

TextWise, for example, aims to solve for companies the problem of matching internal content to other highly relevant content (within or outside the enterprise), categorizing unstructured information into navigable categories and tagging the content with key concepts from the text to create metadata that supports document search and navigation. Keyword search, born decades ago, can’t cut it for the enterprise in a world where, for example, the breadth and complexity of information in knowledge bases leave customers struggling to answer questions through web self-service portals -- and call center agents equally challenged to help them find the information they need. “Somehow we have to be able to figure out how to return information to people that answers their questions,” Kenneally says. “The problem is we continue to have people narrow their focus to one or two words and those words are supposed to bring back information that deals with customers’ concerns.”

That changes when you look to such customer and enterprise application requirements and determine how semantic tools can best be applied to support answering questions -- and not necessarily and exclusively using information behind the firewall but also what’s out there in the vast world of the web, including the fast-growing social media space. “You need to pull real-time search with social search with semantics and behavioral technology to have a big picture of what something is about and how that something is really important to you now,” she says. “The role semantics plays in that is, rather than pulling in a fire hose of information, you make that fire hose into a garden hose so you get what is much more relevant to you.”

Try it Yourself
While TextWise is focused primarily on the business user set, the vendor has created a prototype application, called Gyzork, that uses its technology to let anyone continually discover new information that is relevant to their interests. Gyzorks are smart bookmarks – you create them for articles you like, and as they appear new articles about the same or similar topics are automatically added to your Gyzork and categorized for easy retrieval.

“Normally with bookmarks you might have a couple of words to describe what the article is,” Kenneally says. “Semantic bookmarks are very different because we consider the text of the entire document that you are looking at and we use our Semantic Signatures to determine the concepts and the weights of those concepts in the document.” The premise is that everything in a document is important, but just in a relative sense.

Another tool leveraging TextWise’s technology is foof, a Firefox add-on that eliminates ads entirely or replaces them with your choice content that is related to the pages you are browsing – be it a YouTube video, a news clip, blog post or even pictures from your FlickR account. “We’re not interested in promoting or blocking ads,” Kenneally notes, “but this is another way to use our semantic technology in order to get more relevant results.” The company also has developed a tool to help simplify patent research processes.