- How Google categorizes misspellings
- Google's previous approach to misspelled searches
- How deep learning changed Google's approach to misspelled words
Ever wondered how Google magically knows what you're looking for, even if your search query contains typing errors?
Because one in ten searches is misspelled and new words are constantly highlighted, Google has a full algorithm dedicated to navigating misspelled words.
How Google categorizes misspellings
The first thing Google's AI does when it comes across a misspelled word is to categorize it:
Fingerprint error – Google sees more than 10,000 variations of fingerprint errors in queries like "Youtube". People know how to spell it, but one letter may be wrong. For example, instead of pressing the "t" in Youtube, tap the letter next to it.
Conceptual error – Also known as "best effort notation". This is the case when a user doesn't know how to spell a word and is putting in their best guesses.
Google's previous approach to misspelled searches
In the past, Google relied on keyboard designs to decipher what word a user was trying to type:
"If you tried to type" u "but made a mistake, our systems determined that you were more likely to type" y "as" z "because" y "was next to" u "on a standard English keyboard "stands."
Google starts with the letter that is closest to the one you entered and works its way out with the next letter and the next one until it finds a suitable letter.
Interestingly, this approach not only finger-solved errors, but also helped pinpoint conceptual errors.
Read on below
How deep learning changed Google's approach to misspelled words
On October 15, 2020, Google introduced a new algorithm, which it names as follows:
"A bigger improvement in spelling than any of our improvements in the last five years."
It was then that the results a user searched for in less than three milliseconds began to appear.
The algorithm executes models with over 680 million parameters in less than two milliseconds.
Instead of using the previous keyboard approach, the new algorithm uses context to figure out what a user was trying to type.
This new approach works through:
- Evaluation of the entire query, not just the wrong word.
- Look for replacement words that match the overall query.
- Providing search results based on the best fit.
Read on below
When Google shows results at what you mean, a note will appear below the search bar letting you know. You can also view search results for the original query:
Have you noticed that sometimes when you search, a small note appears below the search bar asking if you meant a certain term:
Google does this when it has a pretty good idea of what you meant but isn't 100% sure.
Read on below
How you react to these notes directly affects the algorithm as Google uses these signals to further train the AI.
So the next time you make a typo while doing a Google search, keep in mind that the mistake serves a bigger purpose than just being annoying.