I went to Google's Searchology event Tuesday morning, which is where it unveils it's latest search technologies. This year, there was a lot said about trying to understand the meaning of user queries, which is a semantic analysis. Udi Manber, vice president of core search, said it was Google's responsibility to understand what a searcher meant, and not what they typed.
This is all part of improving the "Is this what you meant?" line that you often see at the top a page of search results. Mr Manber said there was a considerable amount of "rocket science" search technology involved behind the scenes. It's a very difficult computational problem.
The demos were impressive. Google is using a lot of factors, including the user's location to try to determine what is meant by a search query. For example, "sushi" would return a result of local sushi restaurants. And "sfll playoffs" would turn up San Francisco Little League playoffs rather than "NFL playoffs."
Google said that improving search by understanding meaning, was still in its early days, and that there is so much more to be done.
It's great that Google is increasingly relying on semantic analysis of user searches to get better at delivering the right results in the top few search results. My question to Mr Manber was how will Google monetize this technology? Google sells keywords to advertisers that serve up related ads. But keywords are very specific. Could Google offer a "semantic keyword" if there's such a thing?
Mr Manber wasn't able to directly answer my question, saying that the term "semantic" meant different things to people. [I noted the wonderful irony that "semantic" was itself difficult to understand.]
I asked Danny Sullivan, the search expert at SearchEngineLand, who was sitting right behind me, about this. He said I shouldn't expect Mr Manber to know anything about selling keywords, that he was focused on search technologies.
However, I'd still like to know how Google, or for that matter, how other search, and advertising companies, propose to make money from a semantic approach rather than a specific keyword approach. We hear so much about the "semantic web" and about "semantic search," how would "semantic advertising" work?
Since search is becoming less about what I typed, and more about what I meant but didn't articulate, surely Google will have to change its business model. Currently, it makes money by selling specific keywords tied to specific search terms. Will businesses now have to buy larger numbers of keywords to try to cover the semantic spread? Will this increase the cost of search engine marketing?
Or can the ambiguity of a search term, or a page of content, be measured and sold to advertisers in some way?
There could be a good business opportunity here for a startup that has figured out "semantic advertising."