The method includes receiving a search query that includes terms in a first language; determining possible translations of the terms of the search query into a second language; locating documents in the first language that match the terms of the search query; identifying documents in the second language that contain references to the first language documents; and disambiguating among the possible translations of the terms of the search query using the second language documents to identify one of the possible translations as a likely translation of the search query.
Assume that a user provides a search query to the server in Spanish, but desires documents to be returned in English. Further, assume that the user desires documents relating to "banks interest." In this case, the query provided by the user may include the terms "bancos" and "interes." To facilitate English-language document retrieval, the server may translate the Spanish query to English.
The query translation engine may perform an initial translation of the terms of the query using, for example, the dictionary. In this case, the query translation engine finds that each of the terms of the query has more than one possible translation. For example, the Spanish word "bancos" could be translated as "banks" or "benches" (among other possibilities) in English. The Spanish word "interes" could be translated as "interest" or "concern" (among other possibilities) in English. The query translation engine disambiguates among the possible translations using documents identified by the search engine.
The search engine performs a search using the original Spanish query (i.e., "bancos interes") to identify Spanish-language documents that include anchors that contain all of the query terms and point to English-language documents. The search engine provides the English-language documents that are pointed to by the anchors to the query translation engine.
The query translation engine analyzes the text of the English-language documents to, for example, compute the frequency of co-occurrence of the various translation possibilities. Specifically, the query translation engine determines how often the word "banks" occurs with "interest," "banks" occurs with "concern," "benches" occurs with "interest," and "benches" occurs with "concern." Presumably, the query translation engine would determine that "banks" and "interest" are the most frequent combination and use these terms as the correct translation for the Spanish query "bancos interes."
"I've been waiting for this for a while," he said at a meeting in London, adding that the significance was boosted as it marked the first time Google had sold the use of the back-end technology of YouTube and GMail.
He said that Google was planning similar deals with other large media firms and content providers. "If we can get this structure right over the next few months and it rolls out, then it becomes the index case for every other country and every other operator."
Microsoft released its own book search service, that includes only out-of-copyright books and books for each Microsoft received permission. The approach is different from Google's initiative, where publishers can only opt-out from Google Book Search.
Ultimately, each site owner is responsible for making sure that their site isn't spammy. If you pick a bad search engine optimizer (SEO) and they make a ton of spammy doorway pages on your domain, Google still needs to take action. Hacked sites are no different: lots of spammy/hacked sites will try to install malware on users’ computers. If your site is hacked and turns spammy, Google may need to remove your site, but we will also try to alert you via our webmaster console and even by emailing you to let you know what happened. To the best of my knowledge, no other search engine confirms any penalties to sites, nor do they email site owners.
Even if Google stands by its promise to protect its users' information, there are no guarantees that mischief-making computer hackers or crusading government agencies won't eventually try to pry into the database, said Lauren Weinstein, co-founder of People for Internet Responsibility.
"When data is sitting on computers other than your own, it becomes a very tempting target," he said. "I have no problems at all with Google's motivation because I really do believe they want to protect their users' privacy. But I think they are creating something that will have the vultures circling."
Despite those concerns, having Google automatically store important documents appeals to absent-minded students like Palo Alto High junior Ryan Drebin. "I am always losing my flash drive anyway," he said, referring to a small portable memory chip.
If you use Picasa Web Albums and want to make your albums more visible, a good idea would be to include a small slideshow into your site. Unfortunately, Google didn't have time for that, so someone created a small JavaScript code that uses the yet unofficial JSON feeds. You just have to enter the RSS feed of a public album.
