Search with fewer keystrokes and better spelling

We spend a lot of time thinking about search results, but we also spend a lot of time thinking about search queries. Today we’re announcing three enhancements to help you input your searches more quickly and easily: more localized Google Suggest, improved spell correction for names and auto-correction for 31 languages.

Feel at home with Google Suggest
Last year we launched localized Google Suggest by country, offering relevant popular search queries tailored for different regions. However, just as people in the U.K. often look for different things than people in U.S., we’ve found that people in Seattle tend to look for different things than people in Dallas. So last week, we rolled out a version of Google Suggest that is tailored to specific metro areas in the U.S. You may notice that the list of queries beneath the search box will seem more locally relevant than it used to:

• In San Francisco [bart] is probably not Bart Simpson; it’s probably Bay Area Rapid Transit:

• In Chicago it’s easy to find out about your local NBA team:

Spelling enhancements for names
While Suggest can help you find good queries, sometimes you can get stuck because of misspellings. That’s why for years we’ve offered corrected spellings for mistyped searches (with the “Did you mean” link). We’ve steadily improved this spelling technology over time, but recently we made some big strides in correcting misspelled names.

People often search for people’s names — and not just celebrities and old friends. They look for doctors, horse trainers, hang-gliding instructors... the searches are just as diverse as the personalities in your hometown. We’ve noticed that people sometimes struggle to correctly spell names, and it’s not surprising. Names can be complicated and often there are multiple common spellings.

Our new technology is based on the concept that people often know something else about the person besides the approximate spelling of his name. People often include other terms such as "composer" or "lawyer sparta wisconsin" in their search query, which provides valuable context to help us narrow the range of possibilities for the spelling correction. We use these additional descriptive words to offer you better suggestions. Some examples: [matthew devin oracle], [yuri lehner stanford], [simon tung machine learning]. With these improvements you’ll start seeing more useful spell corrections for names.

For now this enhancement is available in our English spelling system in the U.S. We'll be rolling out the change to other parts of the world and other languages in the coming months.

Spelling auto-correction in 31 languages
Another improvement we made recently to the spelling system is auto-correction. If you search for [aiprt], rather than showing you a link on your results page that says “Did you mean: airport” we’ll take you straight to the results for the corrected search. We auto-correct when we’re highly confident in our correction in order to get you the information you’re looking for that much faster. In the past week we’ve expanded auto-correction to 31 languages across over 180 domains, with more to come.

Did you make a typo while looking for [chocolate strawberries and cream] in Italian? The right word is so close you can taste it:

While saving you that unnecessary click, we make search that much faster.

Posted by Pandu Nayak, Member of Technical Staff