RealTime IT News

Can Semantic Search Change Our Lives?

Semantic Web

SAN JOSE, Calif. -- Does making search a more natural process and getting relevant results quicker sound life changing?

Experts from different search companies were, well, all over the site map on this question during a panel called "Semantic Search: How Will It Change Our Lives?" at the Search Engine Strategies (SES) conference here Monday.

"We don't think search is a solved problem," said Scott Prevost, general manager and director of products at Semantic search provider Powerset. "The thing about keywords is they are a shallow representation of a document's meaning."

Microsoft (NASDAQ: MSFT) recently purchased the search company.

As one example of what it offers, Powerset has applied its natural-language technology to Wikipedia with the aim of providing more diverse and accessible results from the online encyclopedia.

True semantic search, where you can consistently ask a search engine questions in complete sentences and get an accurate answer "is quite a ways away -- years," said Erik Collier, vice president of product management at Ask.com. He noted that when most search engines are asked "What is the population size of Japan?" they don't produce a relevant Web site or correct result among top results. Remove the word "size," however, and you get accurate results up top.

While he clearly cherry-picked this example, Collier noted that Ask.com gives the correct result with or without the word "size."

Part of the answer of what works and what doesn't depends on how questions are asked and what data is searched. Collier said Ask.com is particularly good at tapping structured content from reliable sources. But a huge part of the Web includes unstructured data that often has the information people are looking for.

Session moderator Kevin Ryan of SES, noted that advances in more natural-language search queries come at a time when a generation of users have trained themselves to use "caveman speak" to get results. Most users would be more likely to enter shorthand like "steak restaurant Chicago" instead of "Where can I find a great steak restaurant in Chicago?" because search engines such as Google rely on keywords, among other criteria such as page rank, to produce results.

"How do you change mind-sets and get people to come back to entering questions in plain English or complete sentences?" Ryan asked.

But that seems to be the least of the challenges," he said. "If users start getting better results from more natural-language queries, the vendors think they will win a greater share of the search market soon enough." < p>One company, BooRah, is tackling very specific markets starting with restaurants. BooRah summarizes online restaurant reviews from multiple sources in local markets, including blogs and those written by critics and consumers. Links to the reviews appear in the results, and users are encouraged to give their "Boo" or Rah" vote for restaurants. Those vote results are shown in query results. The company says the concept can be extended to other service businesses.

"Keyword search is a fairly strong paradigm and will continue to be used," said Nagaraju Bandaru, BooRah co-founder and chief technology officer,. But he said natural-language query technology such as BooRah uses is the best way to quickly find so-called long tail content such as where to find the best calamari in San Francisco.

Amit Kumar, director of product management for Yahoo Search, said the Internet giant has many natural language and semantic search irons in the fire. Geo location is one of the areas Yahoo (NASDAQ: YHOO) is looking into to help produce more personalized results. The thinking is, if the search engine knows your location, it can give more personalized results to queries such as "Where's the nearest barber?" This is one of the promises of geo- or location-aware mobile search.

"For Web search we're looking at all kinds of input to maker sure the relevance of the results is per user, that's the holy grail," Kumar said.

Hakia's gallery of relevant results

Hakia offers a natural-language search engine designed to give users what they are looking in the first attempt. Kartal Guner, a co-chief software architect at Hakia, said more traditional search engines give you a set of results as links you have to follow in hopes of eventually finding your information. "We call that 'second search,'" Guner said. "Our way is to provide you with the answer right on the page."

He cited the example of using Hakia for the query "What drug treats urinary tract infection?"

A later check by InternetNews.com using Google (NASDAQ: GOOG) showed a clearer set of consumer-oriented results than the list of medical links in Hakia that seemed better suited to doctors or specialists. The top of Hakia's results page, however, includes a link to a Hakia "gallery" on the subject. Galleries are Hakia's specially formatted collections of results by topic.

Clicking that link for urinary tract infections gives a set of "Basic information and FAQ" summaries and links on the topic, a related illustration, a separate set of image search results and another section of Symptoms and Diagnosis results.

So while the idea of semantic or a more natural way of querying is coming, the implementation and interpretation varies.

"Semantic search is what you make of it," Yahoo's Kumar said. "Different people have different interpretations."