Search Engines Take More Semantic Baby Steps

By SemanticWeb.com on March 5, 2009 6:29 PM

Jennifer Zaino
SemanticWeb.com Contributor

It’s official — search strings are getting longer as users increasingly try to find the information they want and not be swamped with thousands of references to documents that probably have nothing to do with their requests.

Hitwise, an Experian company, recently released a report that notes that the length of search queries has increased over the past year.

“Longer search queries, averaging searches of 5-plus words in length, have increased 10 percent comparing January 2009 to January 2008,” the report stated. “The same time period showed that shorter search queries, averaging those 1 to 4 words in length, have decreased 2 percent.”

We’ve all lived the web-clutter conundrum. Search by and large still suffers from the fact that — as hard as we try to specify our requirements in some set of keywords that we hope combine together in the exact way that’s going to get us the returns we want — we often miss the mark. Or maybe the better way to say it is that we often hit too many marks.

This is where semantic search has its great promise — if the search engine can understand to some degree both the meaning of the words we search on and their contextual relationship, it’s more likely that we will more quickly get directed to the right information.

Semantic search vendors continue to try to deliver on that promise and expand their potential to connect users with the information they want by aggregating the data they’re interested in new ways. Certainly there’s been a lot of words expended on Google moving toward semantic search, some of it recently fueled by CEO Eric Schmidt’s comments during the company’s earnings call in January speculating that it might be nice if Google understood the meaning of a keyword phrase and not just the words that are in it.

More semantic buzz is circulating around Microsoft building in technology from its acquisition of semantic search engine company Powerset into its Live Search update. When users search for a term, they’ll get semantically related categories. The example mentioned on the web groups items around a search for Taylor Swift into categories such as Taylor Swift songs, bio, albums, etc.

Startups Expand Linking Tools

But here’s how some of other players are less coyly trying to deliver on the goals around using semantic technology to connect users with the information they want:

At the Demo ’09 conference this week, Evri debuted a beta of Collections, a feature of its connections-infused search system for delving into people, places and things. Collections purports to apply social media to the semantic web — you can “follow” entities from Evri’s profile page, such as the cast of a TV show, and build them into a collection that you can annotate and share with others, or that others can discover on Evri or around the web.

Startup search engine Sagoon is going to be using Blinkx’s video search engine that includes language-agnostic speech recognition and visual analysis software to power the rich media search feature on its web site. Blinkx reportedly has indexed more than 35 million hours of audio; video, viral and TV content, and made it fully searchable. Sagoon claims to apply “semantic over lexical search” to get users to the most meaningful results.

Last month Hakia updated its ScoopBar, which highlights search results in the opened Web pages that are found by Hakia, as well as Google, Yahoo, Microsoft Live, or any other search engine. The automatic scrolling and highlighting it offers is aiming at solving that pesky problem of opening a dense web page that is supposed to have the search term you were looking for and then being completely unable to find it in the morass of text.

These may be toddler steps to getting where we really want to go, but it seems we are heading in the right direction.

Comments